A Revolution in R&D

A Revolution in R&D
HOW GENOMICS AND GENETICS ARE TRANSFORMING
THE BIOPHARMACEUTICAL INDUSTRY
BCG REPORT

The Boston Consulting Group is a general management consulting firm
that is a global leader in business strategy. BCG has helped companies
in every major industry and market achieve a competitive advantage by
developing and implementing unique strategies. Founded in 1963, the
firm now operates 51 offices in 34 countries. For further information,
please visit our Web site at www.bcg.com.

A Revolution in R&D
HOW GENOMICS AND GENETICS ARE TRANSFORMING
THE BIOPHARMACEUTICAL INDUSTRY
PETER TOLLMAN
PHILIPPE GUY
JILL ALTSHULER
ALASTAIR FLANAGAN
MICHAEL STEINER
www.bcg.com
NOVEMBER 2001

2
© The Boston Consulting Group, Inc. 2001. All rights reserved.
For information or permission to reprint, please contact BCG at:
E-mail: imc-info@bcg.com
Fax: 617-973-1339, attention IMC/Permissions
Mail: IMC/Permissions
The Boston Consulting Group, Inc.
Exchange Place
Boston, MA 02109
USA
Credits: Left cover photo by Bob Waterston, Washington University, St. Louis, Missouri. Used by permission.
The photo shows a bird’s-eye view of one room in the DNA sequencing facility at the Whitehead Institute Center for
Genome Research.

Table of Contents
ABOUT THE AUTHORS 4
FOREWORD 5
EXECUTIVE SUMMARY 6
INTRODUCTION 9
CHAPTER 1: THE IMPACT OF GENOMICS 11
Preface 11
The Opportunities 12
The Challenges 18
A Final Word 21
CHAPTER 2: THE IMPACT OF GENETICS 24
Preface 24
Disease Genetics 27
Pharmacogenetics 33
A Final Word 39
CHAPTER 3: MANAGERIAL CHALLENGES 41
Preface: Looking Back and Looking Forward 41
Strategy—Searching for Genomic Competitive Advantage 41
Putting the Strategy into Operation 49
A Final Word 56
CONCLUSION 57
METHODOLOGY 59
3

4
About the Authors
About the Authors
Peter Tollman is a vice president in the Boston office and leads BCG's biopharmaceutical R&D business.
Philippe Guy is a senior vice president in the Paris office and leads the worldwide Health Care practice. Jill
Altshuler is a manager in the Boston office and a key contributor to BCG’s genomics initiative. Alastair
Flanagan is a vice president in the London office and leads the U.K. Health Care practice. Michael Steiner
is a senior vice president in the Munich office and leads the German Health Care practice.
Acknowledgments
Sarah Cairns-Smith (Boston) pioneered BCG’s investigation of genomics. Samantha Gray (Boston) has made
significant contributions throughout the research and writing phases of the report.
The authors would like to thank the advisory team: Oliver Fetzer (Boston), Hamilton Moses (Washington,
D.C.), Niko Vrettos (Düsseldorf), and Craig Wheeler (Boston). The authors would also like to acknowledge
the contributions of the project team: Dierk Beyer (Frankfurt), Markus Hildinger (Boston), Raphael Lehrer
(Washington D.C.), Nancy Macmillan (Boston), Jonathan Montagu (London), and Joanne Smith-Farrell
(Washington, D.C.).
For Further Contact
The authors welcome your questions and comments. For inquiries about this report or BCG’s Health Care
practice, please contact:
Alastair Flanagan, London e-mail: flanagan.alastair@bcg.com
Philippe Guy, Paris e-mail: guy.philippe@bcg.com
Mark Lubkeman, Los Angeles e-mail: lubkeman.mark@bcg.com
Michael Steiner, Munich e-mail: steiner.michael@bcg.com
Martin Reeves, Tokyo e-mail: reeves.martin@bcg.com
Peter Tollman, Boston e-mail: tollman.peter@bcg.com

Foreword
To meet growth targets, pharmaceutical companies are going to have to increase R&D productivity. By a fortunate
coincidence, that crisis in expectation is being counterbalanced by a surge of opportunity. Recent
years have seen astonishing advances in technology and explosions of data, which are driving two waves of
change through the industry—a genomics wave and a genetics wave—and radically reshaping R&D methods
and economics in the process. Biopharmaceutical R&D is moving into a new era: almost every link in the
value chain has the potential for tremendous boosts in efficiency or success.
But these advances are not assured. Technological hurdles have yet to be overcome, particularly in the genetics
wave. Moreover, because the productivity boosts are likely to be unequal and uncoordinated, the value
chain itself will demand reconfiguring. And so too, in consequence, will many traditional operational procedures
and organizational structures. The repercussions of genomics, in other words, are going to reach the
furthest recesses of corporate constitution and culture. A true revolution, in short—and one that is already
well under way.
BCG has evaluated deeply the economic and business implications of these disruptions. To bolster our internal
understanding, we gathered information and perspectives in an extensive program of interviews with
leading R&D scientists and executives. Our findings—based on the combination of these interviews, economic
modeling, and client casework—form the substance of this report. Its three sections are devoted
respectively to the impact of genomics, the impact of genetics, and some of the strategic and operational
implications for biopharmaceutical firms.
The first two sections have already been published separately. They generated considerable publicity, and—
more important—considerable comment. We now look forward to your further responses to the report as a
whole.
Philippe Guy
Senior Vice President
5

6
Executive Summary
In the pharmaceutical industry’s struggle to reach
the levels of growth expected of it, one of its key
aims will be to increase R&D productivity. And a key
means of meeting this challenge is to adopt some of
the new technologies and approaches broadly
defined as genomics. 1 That is bound to be a complicated,
perilous, and often painful process, but if
companies get their strategy right and overcome
the obstacles, they could, in the best case, as much
as halve the cost of drug development.
The report is divided into three parts.
The Impact of Genomics
The first great advance of the genomics era is in
technology—above all, the integration of new highthroughput
techniques with powerful new computing
capabilities. The new technologies are active
throughout R&D, most immediately at the drug discovery
stage, and promise to enhance productivity
by boosting efficiency.
The staggering investment needed to develop a
drug—$880 million and 15 years is the pregenomics
average—could be reduced by as much as
$300 million and two years by applying genomics
technologies. Productivity gains would be realized
at every step in the value chain. Potential obstacles
abound, however. In particular, two broad challenges
must be met to realize the savings:
•Target quality must be maintained. Pursuing new
target classes could involve unfamiliar costs initially,
and these could delay the rewards—though
only temporarily. But to jeopardize target quality
by withholding that early investment would be to
risk higher failure rates downstream, and that
would involve far greater costs in the end.
• Bottlenecks must be eased. Owing to the unevenness
of the efficiency gains at different steps in
the value chain, the pipeline’s flow will be
impeded at various chokepoints. If the requisite
action is taken, an even flow should be restored
and the promised rewards should be safeguarded.
The Impact of Genetics
The second great advance of the genomics era is in
the quantity and quality of data. From the data,
invaluable information about individuals’ genetic
variation can be extracted and exploited. In pharmaceutical
R&D, genetics will be applied particularly
to two tasks: identifying genes whose carriers
are susceptible to specific diseases (disease genetics);
and subdividing patients in clinical trials
according to variations in drug response (pharma-
1. Genomics in its narrow sense contrasts with genetics. Roughly, the former concerns itself with the common “standard” genetic makeup, the latter with
the distinctive genetic makeup of individuals. But in its broader sense, genomics includes genetics. In this report, the context makes clear which sense is
intended.

cogenetics). The productivity gains will be realized
mostly in later phases of the value chain, through
the boosting of success rates.
This genetics wave is still gathering strength, but in
due course could make an even greater impact on
R&D than the genomics wave. In an ideal scenario,
the savings would exceed half a billion dollars per
drug. Several troubling hurdles would have to be
negotiated first, however. These include:
• Scientific and technical hurdles. For genetics
approaches to work, the disease susceptibility or
drug response has to be genetic in nature. The
gene in question has to be identifiable and must
lead to a drugable target and/or be found in time
to streamline trials.
• Economic and market hurdles. The cost of conducting
genetics studies will need to drop, and
the opportunity cost of a restricted label could
offset the potential market upside of pharmacogenetics.
Beyond these hurdles, other challenges will need to
be addressed:
• Difficult investment decisions will have to be
made, weighing high risk against potentially high
rewards. Companies will need to decide exactly
how to participate in genetics—whether to invest
in genetics approaches, and how deeply, consistent
with their level of risk tolerance.
• Unprecedented coordination between marketing
and R&D will be necessary. Marketing will need to
have a say in deciding which markets and which
genetic diseases R&D should concentrate on, and
will need to become involved earlier than ever.
• Careful attention will need to be given to ethical
considerations. Companies will have to ensure
privacy of genetic material, and be prepared to
address any concerns the public may have.
Managerial Challenges
With the new wealth of options and the increased
interdependencies across the value chain, strategic
issues will prove more complex than in the past.
Likewise operational issues: many traditional ways
of doing business will be disrupted by genomics
technologies, and companies may need to restructure
fairly drastically.
The range of strategic options available to a
company will be dictated by the company’s starting
position—its size, beliefs, aspirations, and capabilities.
Given the magnitude of the opportunities and
7

8
the risks involved, momentous investment decisions
will need to be made, and at the very highest levels
of the organization. And R&D executives will face a
daunting new set of management responsibilities
and challenges. These include:
• Selecting an appropriate research focus—no
longer just the therapeutic area or disease state of
interest, but also such dimensions as target class
and treatment modality
• Choosing which technologies to implement and
when and how to implement them—in-house, or
through partnering or licensing
• Rebalancing the value chain—partly by reallocating
resources but mainly by redesigning processes
and more actively planning and managing capacity
• Establishing a unified informatics infrastructure—including
a centralized knowledge management
system
• Establishing the new organization—creating new
interfaces within the R&D department, between
departments, and even between corporations
• Revising decision-making procedures—fully
exploiting the latest data in order to select the
most promising targets and compounds to move
through the pipeline and to optimize their relative
resourcing
• Reinforcing these various reforms by engaging
the emotional and behavioral issues as keenly as
the operational ones
All things considered, companies cannot stand
aside. Certainly there are risks in signing up for the
revolution, but there is also a great risk in ignoring
it—the risk of becoming uncompetitive. The revolution
is real, and will leave no one untouched.

Introduction
Throughout the pharmaceutical industry, executives
are worried. They fear they will not be able to
meet the double-digit annual growth expectations
implied by high market capitalizations. The requisite
new drugs will not be forthcoming: R&D just
cannot deliver them all.
One standard response to this problem is to scale
up—that has been the basis of many a recent
merger—but while scale can pay off in commercialization,
global development, marketing, and distribution,
it is unlikely that scale alone can solve the
R&D problem. Another standard response is to buy
in drug candidates. Such a Band-Aid approach cannot
work indefinitely, and is a risky one anyway,
given that the price of these deals will continue to
rise as demand for them grows.
The only sure way to address the problem is to
increase R&D productivity. And the way to ensure
that is either to increase efficiency (lower cost or
higher speed) or reduce failure rates along the
value chain. Many companies have increased productivity
over the past decade, specifically by
reengineering the development phase. That optimization
may be reaching its limits, however. As for
the discovery phase, it has long been less amenable
to such improvements. So the problem of produc-
tivity persists. Traditional approaches cannot provide
an answer, but genomics can. (See Exhibit 1.)
It will not be easy, of course. There are some difficult
obstacles en route—difficult, but not insurmountable.
By making informed strategic choices,
companies can overcome the obstacles and reap the
productivity rewards. Those that embrace the revolution
most boldly could potentially halve the cost
and time it takes to develop a new drug—if they
meet certain challenges successfully.
EXHIBIT 1
GENOMICS IMPROVES R&D PRODUCTIVITY
Cost (time)
spent
Failed targets/
candidates
Successful
drug
SOURCE: BCG analysis.
Total cost to develop a drug
Reduce
failure
Improve
efficiency
Total cost (time) per step
9

Chapter 1: The Impact of Genomics
Preface
As the science of genomics has advanced, so has the
definition. When the term was coined in 1986, it
referred mainly to the study of the mammalian
genome—specifically, the mapping, sequencing,
and analyzing of all its genes. The scope soon
expanded, focusing not just on the genes’ structure
but on their function as well. More recently, the
scope of the term has broadened further, focusing
no longer just on knowledge of the genome but also
on the exploitation of that knowledge, especially
for health care.
Going beyond dictionary definitions, our interest is
in what genomics means for the economics of pharmaceutical
R&D. On the basis of our extensive
research and many discussions with prominent people
throughout the industry, we suggest characterizing
genomics, for the purposes of this study, as the
confluence of two interdependent trends that are
fundamentally changing the way R&D is conducted:
industrialization (creating vastly higher throughputs,
and hence a huge increase in data), and informatics
(computerized techniques for managing and
analyzing those data). The surge of data—generated
by the former, and processed by the latter—is
of a different order from the data yields of the pregenomics
era.
To elaborate. The new high-tech industrialization
has increased the efficiency of certain activities
beyond recognition. Instead of assigning individual
scientists to work manually on modest individual
experiments, companies now invoke automation
and parallel processing to conduct experiments
much larger in scale and complexity, and at a much
faster pace.
Look around this lab—you have to search high
and low to find a human heartbeat. Now robots
can do the menial things we did in grad school.
—Research leader,
leading biotech company
The data that emerge are immensely greater both
in quantity and in richness. Enormous databases—
detailing gene expression, for example, or homologous
genes across species, or protein structures—
afford unprecedented comprehensive views of
biological processes. Increasingly, researchers can
understand properties of the system rather than
just individual parts, and that holds out the promise
of a more rational approach to drug discovery.
The new technology of informatics serves to handle
and process all these data. Without it, the data
would remain raw material. Informatics was nurtured
by several coinciding factors: the ever-accelerating
power of computers, refined algorithms, the
integration of data and technology platforms, and
the versatility of the Internet. The effect is that
overwhelming masses of information can now be
marshaled, managed, and analyzed as never before.
Data are transformed into knowledge.
We could never have achieved drug development
that fast with traditional techniques. No way—
without the computers we didn’t have a chance.
—VP of chemistry,
biotech company
11

12
The Opportunities
What is the impact of genomics on the economics
of R&D? To what extent will genomics improve productivity
overall, and what will its effects be when
applied at various points of the value chain? What
other incidental advantages might genomics bring
in its wake?
These crucial questions have received a great deal
of attention of late, and a wide variety of responses.
To address the questions in a rigorous, fact-based
way, we built an economic model of the entire R&D
value chain, grounded in a program of discussions
within the industry (more than 100 meetings with
more than 60 scientists and executives from nearly
50 companies and academic institutions.) (See the
methodology section at the end of this report.)
Realizing Savings
Before genomics technology, developing a new
drug has cost companies on average $880 million,
and has taken about 15 years from start to finish,
that is, from target identification 2 through regulatory
approval. (See Exhibit 2.) Of this cost, about 75
percent can be attributed to failures along the way.
By applying genomics technology, companies could
on average realize savings of nearly $300 million
and two years per drug, largely as a result of efficiency
gains. That represents a 35 percent cost and
15 percent time savings. (And those are the savings
possible with technologies that are available today;
when new or improved genomics technologies
emerge, the savings will be even greater.) If companies
wish to stay competitive, they have no choice:
they must implement genomics technologies. (See
Exhibit 3.)
Doing so, however, will hardly produce such huge
savings immediately, or automatically. It will take a
few years, and many deft decisions, for the savings
to be realized. The early years of implementation
may in fact involve an increase in costs as the learning
curve is negotiated for novel targets—specifi-
EXHIBIT 2
DRUG R&D IS EXPENSIVE AND TIME-CONSUMING
Cost: $880 million total
Approximate cost ($M)
165
205
Time: 14.7 years total
Approximate time (yrs)
1
Biology
2
Target ID Target Validation
40
0.4
Chemistry
cally, as the necessary quality controls are established—and
as major strategic decisions (about
personnel and processes, for instance) are confirmed
or revised.
More on these challenges later. But first, we will
take a closer look at the long-term upside, detailing
the savings at various steps along the value chain.
120
2.7
Screening Optimization
90
1.6
Development
260
7
Preclinical Clinical
SOURCES: BCG analysis; industry interviews; scientific literature; public
financial data; Lehman Brothers; PAREXEL’S Pharmaceutical R&D
Statistical Sourcebook 2000.
NOTE: Cost to drug includes failures. Target identification includes initial
experiments that companies may have outsourced to academic research
institutions.
2. Includes initial experiments to identify potential targets. Traditionally, companies have sourced much of this research from academia.

EXHIBIT 3
GENOMICS CAN YIELD SIGNIFICANT SAVINGS
Cost to drug
Pre-genomics
Post-genomics
target ID
Plus in silico
chemistry
Plus preclinical and
clinical advances 1
Time to drug
Pre-genomics
Post-genomics
target ID
Plus in silico
chemistry
Plus preclinical and
clinical advances 1
ID Biology
Target ID Target Validation
0
0
200
Cost ($M)
Target Discovery/Biology
The identification of targets is being industrialized—through
the use of technology such as gene
chips to perform gene expression analysis, for
example—and then further enhanced by bioinformatics.
Scientists can now use a single gene chip to
compare the expression of thousands of genes, in
diseased and healthy tissue alike, all at once, and
can then use informatics technology to find follow-
5
Chemistry
400
Screening Optimization
600
10
610
590
800
740
13.0
12.7
880
13.8
15
Time (years)
Development
Preclinical Clinical
SOURCES: BCG analysis; industry interviews; scientific literature; public
financial data; Lehman Brothers; PAREXEL’S Pharmaceutical R&D
Statistical Sourcebook 2000.
1,000
14.7
1Includes surrogate marker savings from early elimination of unpromising
candidates, not from early FDA approval; does not include potential savings
from pharmacogenetics.
up information, on these or related genes, in databases
around the world. (Target validation, however,
seems difficult to industrialize, owing to the
“slow” biology of whole-animal systems still
involved, and is not yet showing significant productivity
gains.)
In all, the potential savings per drug are on average
about $140 million and just under one year of time
to market, achieved entirely through improved efficiency.
That would add about $100 million in value
per drug (assuming an “average” drug with peak
annual sales of $500 million). So for this step in the
value chain, productivity would increase vastly: it
would be six times as high as before, assuming the
same level of investment. A sixfold increase in the
number of potential targets!
Several companies have already benefited handsomely
from this windfall. Take the case of
Millennium, which was an early adopter of industrialized
biology. The company, anticipating an overabundance
of targets, established a business model
in which it sells off much its output and uses that
income to fund internal research. Starting from its
early genomics platform, Millennium has strategically
acquired or partnered with other platform
companies to establish an integrated drug discovery
value chain. From the other perspective, pharmaceutical
companies such as Bayer and Aventis
have made deals with Millennium, in the expectation
of profiting from the new abundance of targets
they can choose to pursue.
Lead Discovery/Chemistry
Chemistry is being revolutionized by in silico (that
is, computer-aided) technology—specifically, virtual
screening supported by chemoinformatics. In
virtual screening, potential lead chemicals are
assessed with computer algorithms to test how likely
they are to interact with a target. Chemoinformatics
provides the necessary platform for virtual screening,
using data and analysis from high-throughput
screening (HTS) and other chemistry activities.
This approach increases efficiency by focusing compound
synthesis, reducing the number of assays,
increasing the parallelization of screening steps,
13

14
and generally helping to optimize screening. The
power of this approach is expected to increase dramatically
with the availability of larger data sets for
refining the predictive algorithms. (At the moment,
however, in silico chemistry has one notable shortcoming:
it looks as if it will be suitable for only
about 30 percent of targets—the rest fail to yield
the requisite structural information—and even
then might prove difficult to apply until lead optimization.
Our savings are calculated for those targets
where in silico technology can be applied.)
The potential savings are on average about $130
million and nearly one year per drug. That would
add about $90 million in value per drug. For this
step of the value chain, then, productivity would
double, assuming the same level of investment.
As a beneficiary of these advances, a good case in
point is Vertex. Starting from an IDD (in silico drug
design) platform in chemistry, the company has
gone on to develop an integrated value chain in its
own right. In silico models have allowed more efficient
design of small-molecule drugs than a purely
traditional approach, and the company’s discovery
focus has been on certain target classes that benefit
most from proprietary in silico technologies. This
approach has met with considerable success, culminating
in one of the biggest biotech alliances so far
(with Novartis, and worth $813 million). Vertex can
fairly claim to have the strongest small-molecule
drug pipeline within the biotech industry. With one
drug on the market and twelve candidates in development,
it compares favorably with some of the big
pharmaceutical pipelines.
Serious money can be saved for the target classes
where in silico chemistry works.
—Director of chemistry,
major pharmaceutical company
Development
Three key genomics advances look set to increase
capacity here. In silico ADME/tox (absorption, distribution,
metabolism, and excretion/toxicity) and
high-throughput in vitro toxicology are revolutionizing
the preclinical phase through their power to
predict drug properties. And surrogate markers
(physiological markers that correlate with elements
of drug response), applied in both preclinical and
clinical trials, evaluate drug effects more efficiently
than before: they are quick to identify failing compounds,
and once regulatory approval is granted,
will be used to identify passing compounds too.
In combination, the potential savings available in
the short term are on the order of $20 million and
0.3 years per drug. That would add about $15 million
in value per drug. But these approaches will
become even more valuable as clinical data on the
relationship between genes, gene expression, and
disease accumulate and regulatory agencies begin
to accept clinical-marker data: the potential savings
could rise to $70 million.
These technologies are being adopted by forwardlooking
chemistry companies, and are enabling
them to pull certain preclinical activities into the
chemistry part of the value chain. For example,
ArQule has recently acquired Camitro to incorporate
an integrated in vitro and in silico ADME/tox
platform into its own set of capabilities.
These are not the only advances likely to transform
productivity during the development phase. Pharmacogenomics—through
its power to identify subgroups
of patients who respond differently to a
drug under study—offers the promise of streamlining
clinical trials; we explore this topic in more
detail later. Beyond genomics (and beyond the
scope of the current report), “e-technologies,” such
as electronic patient recruitment and monitoring
via the Internet, are expected to speed up the
launch and completion of clinical trials.
Beyond the Traditional Value Chain:
Chemical Genomics
The various productivity gains just outlined occur
within specific steps of the value chain. But suppose
you could transcend the traditional value chain, or
refashion it to streamline R&D. That is one of the
revolutionary prospects now opening up. The key is
chemical genomics, and the way it will dissolve the
old boundaries is by introducing into the value
chain a kind of parallel processing. (See sidebar,
“Chemical Genomics—Forward or Reverse.”)

CHEMICAL GENOMICS—FORWARD OR REVERSE
When companies say they are pursuing chemical
genomics, they are usually referring to large-scale
reverse chemical genetics. (That is how the term is
used in our report.) This approach involves finding
chemical compounds that bind to a known target.
Companies often perform this task for entire target
classes; it is especially popular for protein classes
that are known to be highly drugable, such as Gprotein
coupled receptors (GPCRs). The assay for
binding does not need to provide functional information
relevant to a specific disease state—biological
function can be assessed in validation experiments.
The alternative is forward chemical genetics. This
approach begins with functional knowledge. A
library of compounds is screened in an assay that
tests for changes in a specific biological function.
One immediate result would be to process the glut
of identified targets more quickly: instead of joining
the logjam at the validation stage, a great many
of them can now be diverted directly to screening.
If they fail there, they can be discarded right away,
and thus simply bypass most of the validation stage
altogether. In other words, screening moves up the
value chain to rest alongside validation, in a parallel
rather than consecutive position. By bracketing
the industrialized steps of target identification and
chemical screening, chemical genomics has given
the value chain a remarkable makeover.
The key is to move lengthy, messy biology far downstream
where you know it’s worth pursuing. Many
targets aren’t drugable, so just validate the smaller
drugable subset.
—SVP of discovery,
leading biotech company
The effect of this new value chain is dramatic: time
to drug is cut by a further two years (that’s on top
of the year already saved by using genomic target
identification). On the other hand, there is a large
3. In July 2001, Aurora Biosciences was acquired by Vertex Pharmaceuticals.
The intention is to screen a library against all expressed
genes in the system under investigation.
This approach has the tremendous advantage of allowing
the identification of targets without any presumptions
as to their function. Additionally, these
targets can help to elucidate the mechanism of disease,
thereby revealing other potential targets in relevant
pathways. The drawback is that forward chemical
genetics has not yet been industrialized, and
throughput levels are therefore very low. According
to our model, implementing it today would increase
costs to more than $1 billion per drug, owing to the
use of “slow” biology, which is needed to set up the
screening assays in chemistry. The expert estimate is
that forward chemical genetics is still as much as
five years away from being economically feasible.
increase in cost, offsetting all cost savings from target
identification. But the tradeoff is still positive.
In a highly competitive market, where new entrants
are continuously eroding share, chemical genomics
can add more than $200 million in value per drug.
(In less competitive conditions, the value added
may be as little as $20 million.)
No doubt chemical genomics costs more—but you
take the loss to gain the speed. Time is money.
—SVP of discovery and technology,
major pharmaceutical company
One important drawback of chemical genomics is
this: it is limited mainly to known target classes.
With targets of unknown function, results become
very difficult to interpret. The proxy assays used for
screening—heat-stability assays, for instance—tend
to yield both false positives and false negatives.
Nevertheless, chemical genomics is already being
pursued throughout the industry. Several big pharmaceutical
companies have adopted it, and genomics
companies such as Aurora Biosciences 3 and
15

16
Cellomics are well positioned to exploit the expected
resulting demand for screening resources.
Aurora is a likely winner in the race to resolve chemical
genomics-related bottlenecks, since it boasts
some of the most advanced screening and assay
technologies in the industry. It has an unusual business
model, in that it provides tools and discovery
services but does not engage in any drug discovery
of its own.
* * *
So much for the imminent efficiency savings across
the R&D value chain. They are hardly the end of
TECHNOLOGIES IN WAITING—OTHER TECHNOLOGIES EXAMINED,
BUT OMITTED FROM OUR REPORT
In this report we have focused on the technologies
and approaches that are having the greatest impact
on R&D economics today. Several other exciting advances
appear likely to make a comparable impact
beyond the next three to five years (too far ahead for
inclusion in our analysis for this report), in particular,
the use of proteomics in target identification,
conditional gene inhibition in target validation, and
industrialized structural biology in screening and
drug design.
Proteomics is the study of protein expression and
protein-protein interactions. Its aim is an understanding,
and ultimately exploitation, of protein function.
Identifying proteins through sequence or structure
homology has recently become much more efficient,
thanks to bioinformatics’ role in analyzing largescale
experiments. One example of a genomics company
applying proteomics is Oxford Glycosciences,
which is engaged in identifying targets and surrogate
markers, both in collaboration with pharmaceutical
companies and in an independent pipeline. But proteomics
is not really industrialized yet, and has high
hurdles to overcome before it is.
We examined the economics of proteomic expression
studies using two-dimensional gel analysis, followed
the story, of course. Other technological advances
are bound to improve R&D productivity further in
due course. Important emerging technologies
include proteomics, partial target inhibition, and
structural biology. (See sidebar, “Technologies in
Waiting.”)
Improving Decision Making
The economics of R&D hinge on success rates, and
success rates depend largely on a cascade of decisions
that have to be made again and again:
whether or not to pursue a target or lead, and if so,
how—to what extent and with what approach.
by identification of interesting proteins through mass
spectrometry.
Under optimal conditions today, this approach has
the potential to save about as much in cost as
genomics-based approaches do, though not as much
in time (about six months less). As the technology
becomes industrialized, proteomics could well surpass
genomics-based approaches, but that is still
several years away.
The aim of the second promising technology we
investigated, conditional gene inhibition, is to overcome
a common problem in target validation. Here
is the background. A standard technique for target
validation uses “target knockouts.” The potential target
is removed, or “knocked out,” from an animal at
conception; this results in the total inhibition of the
target’s function from embryo to adult. The trouble is
that drugs work differently. Very seldom do they
inhibit target function fully, and they are taken only
after genes have already fulfilled their developmental
role in utero. So the use of target knockouts as a target
validation technique does run the risk of creating
false negatives (in some cases indicated by death,
because of the unnatural disruption of embryonic
development). What is needed instead of total gene

Genomics may offer an opportunity for companies
to make the correct decision more often than
before. For one thing, genomics can ultimately provide
more, better, and earlier information, and
good information translates ultimately into high
success rates. For another, the implementation of
genomics approaches will force companies to
rethink their internal decision-making processes.
Genomics-based information, together with the
ability to mine it productively, gives a company an
enormous advantage. Such a company will now be
able to make and execute decisions on targets and
inhibition, therefore, is conditional gene inhibition,
which mimics the partial inhibitory effect of a drug.
Several promising approaches have emerged, including
forward genetics, chemical genomics, and
molecular switches that modulate gene expression,
but their practicality has still to be proved.
Examples abound of genomics companies engaged in
developing these target-validation techniques. Lexicon
Genetics, Exelixis, and Ingenium, for instance,
are using mass mutagenesis on animals such as mice
and zebrafish. In a more focused project, Hypnion is
using forward genetics and other approaches to
understand sleep-wake disorders in mammals.
What benefits lie in wait? By eliminating the false
negatives associated with the current knockout technique,
these new technologies could double or even
triple the number of validated targets, and in that
way save up to $200 million per drug. At the
moment, however, these new kinds of validation
(with the exception of certain chemical genomics
approaches, discussed in the main text) are still
mainly limited to “slow” biology.
Finally, structural biology is used for generating and
analyzing the three-dimensional structure of targets
leads with greater speed and consistency than
before. Guided by more rigorous selection criteria,
the company should go on to improve its success
rates and hence its productivity.
A mere 10 percent improvement in accuracy of
decisions at any stage would confer disproportionately
large benefits. Consider, for example, all the
target/lead pairs that fail just before clinical trials:
if a company were able to decide in just one out of
ten such cases against pursuing the target in the
first place, it would save as much as $100 million
per drug on average. As for INDs that fail clinical
for virtual screening, and is essential to in silico drug
design. Unfortunately, it currently entails protein
crystallization (to prepare the proteins for visualization
by X-ray diffraction), which is a difficult, laborintensive
manual process. Speedier alternatives,
such as NMR spectroscopy, cannot predict overall
protein shape adequately, being restricted to protein
subsegments. As a result, in silico modeling remains
limited in its applicability: the algorithms cannot
boast really high precision for target classes where
no example structures are available.
Several projects, both public and private, are under
way to upgrade structural biology platforms to the
point where they will achieve industrialized scale.
Among the private endeavors is the Novartis Institute
for Functional Genomics, founded by Novartis to
identify and characterize targets using high-throughput
technologies. In the biotech field, Structural
Genomix aims to become a platform provider and
generate revenues by selling protein structures; the
company may also decide to exploit its data inhouse,
and extend into in silico drug design. But it
might be several years before technologies have
developed far enough for the necessary scale effects
to be realized.
17

18
trials, if the company were able to decide in just one
out of ten such cases to abandon development earlier,
it could save an additional $100 million per
drug.
Improving decision making to that extent will take
more than simply acquiring and implementing the
new genomics technologies and approaches. It will
take some serious strategic rethinking too, and possibly
major organizational changes. Whether to keep
all activities in-house, or seek partners, or buy in targets
or leads. How to redistribute resources, reassign
personnel, and revise lines of communication and
chains of command. Such operational and organizational
quandaries will be addressed in detail in the
final chapter of this report.
We implemented a fast-in/fast-out decision policy
about projects—if we didn’t have optimal conditions
met in 18 months, we killed it. That made all
the difference.
—Former executive,
leading pharmaceutical company
Even the basic business skill of decision making,
then, is not immune to the influence of genomics
technology. Whatever other benefits it brings,
genomics serves as a wake-up call across the industry,
even for companies trying to shelter from the
genomics revolution.
The Challenges
Although implementing genomics offers companies
great opportunities, it also presents them with
formidable challenges. One of these is to ensure
that the quality of the pipeline remains uncompromised.
Another is to put the new technologies into
efficient operation.
Maintaining Quality
If the potential productivity gains are to be fully
realized, the post-genomics R&D pipeline will need
to retain or improve its pre-genomics quality. Any
decline in quality—the quality of targets and
leads—would obviously have an adverse effect on
productivity. The main threat to quality derives
from the unorthodoxy, the unfamiliar nature, of so
many new targets. Entire target classes, previously
unknown, will need investigating. The temptation
to pursue leads prematurely is bound to arise, and
quality control will need to be rigorously enforced
to uphold the pipeline’s usual success rates.
In any given experiment, 70 percent of what I see is
completely new. It could be a gold rush, or it could
be junk—-there’s no way to tell until I sit at the
bench and do more work.
—Director of research,
leading biotech company
To appreciate the threat accurately, we need a
proper definition of the term quality.
The “intrinsic quality” of a target or lead amounts
to its likelihood of success, which is based on factors
such as clinical relevance and drugability.
Companies can do little to alter this type of quality.
The “provisional quality” (or “informational quality”)
of a target or lead is based on the amount of
data available on it at any given time—how much is
known about its clinical relevance, drugability, and
so on. (This informational quality helps to predict
success rates, but does not influence them.)
Companies can alter this type of quality, by spending
appropriately, and in that way can improve their
ability to predict downstream success rates.
This distinction is crucial. But it has at times been
overlooked, resulting in some confusion in the
industry. A widely publicized concern has been that
novel targets identified through genomics would
tend to be of inherently lower quality than pregenomics
targets, and thus more likely to fail at
some costly phase downstream. That inference is an
oversimplification, and is misleading.
Certainly genomics proposes many more novel targets
(as much as 60 to 70 percent of potential targets,
in our interviewees’ experience, may belong
to previously unknown target classes), and their
informational quality at that early stage is duly modest.
But that says nothing about their intrinsic quality.
Any prudent company, no matter how bold, will
strive to learn more about novel targets before
deciding to pursue them downstream. In our analysis,
investments made to raise a novel target’s infor-

mational quality to the level of a known target’s
would be more than recouped in due course.
The overall cost of these novel targets—raising
their informational quality and then pursuing them
down the value chain—is bound to rise initially.
However, within three to five years from the initial
discovery of a target in a novel class, according to
our model, the overall cost increase per novel-class
drug could return to average.
Where do the added costs come from? And what
must happen to offset them?
The Cost of Quality Control
Our model predicts that the typical increase will be
about $200 million and more than one year per
drug (that is, a total cost of $790 million versus
$590 million, and a total time to drug of 13.8 years
versus 12.7 years). The increase is mainly attributable
to the extra time needed to understand target
function and develop appropriate assays in target
validation and screening; also, to the need to screen
a higher proportion of compounds, since an appropriate
subset of a larger library cannot be selected
in advance.
Chemical optimization costs would increase only if
the novel target required a novel compound (by no
means a necessary requirement, though certainly a
possible one occasionally). Our model examines
this worst-case scenario explicitly. If a novel target
does happen to require a novel compound, or a
compound unfamiliar to the medicinal chemists,
the potential efficiency loss causes a further
increase of $290 million and more than two years
per drug (that is, a total cost of about $1.1 billion
versus $590 million, and a total time to drug of 15
years versus 12.7 years). The additional increases
here would be due to the extra time needed now for
medicinal chemists to learn how to modify the compound
and attain specific properties through trial
and error. But this worst-case scenario should not
be very common.
Moving further still down the value chain, to the
preclinical and clinical phases, costs are not
expected to increase. The downstream success rate
for novel compounds or targets should turn out to
be much the same as that for known compounds or
targets, as long as the same standards are applied.
There should be no significant increase in toxicity
or decrease in efficacy, other than in very unlikely
circumstances—for instance, if existing animal
models somehow proved less suitable, or if drugs
for novel target classes were to interact with metabolic
pathways in utterly unfamiliar ways.
Offsetting the Costs
Raising the informational quality of novel targets
involves a heavy investment, but it is a wise investment.
And a fairly quick one: knowledge about
one novel target quickly elucidates other potential
targets in the same class. Thanks to feedback
loops, knowledge increases geometrically. As more
is learned, the level of investment can tail off
accordingly.
In any case, the alternatives to making that early
investment in informational quality are far from
attractive. On the one hand, dropping the targets
would be terribly short-sighted: companies would
be forgoing the opportunity to discover and exploit
untapped sources of revenue. On the other hand,
pushing novel targets onward without adequate
information on them would almost certainly result
in a higher failure rate downstream, with all the
associated implications for cost. An increased failure
rate of just 10 percent across chemical optimization
and all of development would on average
increase costs by about $200 million per drug.
To sum up, then: costs incurred early in the value
chain (by information gathering) look preferable
to those that would otherwise be incurred later (as
the result of a higher downstream failure rate). All
the more so, given that the early costs should soon
begin falling (investment in information is almost
always associated with an experience curve): as
novel target classes become increasingly familiar, it
will become increasingly efficient and economical
to pursue new targets within those classes. So with
proper handling, the burden of that early cost
increase is just a short-term one, and the productivity
of genomics-driven R&D should soon return
19

20
almost to that of more familiar target classes. We
estimate the time required for this is about three to
five years from the discovery of a novel target,
which is the amount of time it should take to complete
validation and early screening (assay development).
(See Exhibit 4.)
Putting the New Technology into Operation
It is one thing to acquire and install new capabilities
and another to get them to function as they are
meant to. The challenge of making genomics technologies
operational has two major components:
easing the bottlenecks that will develop, and resolving
the personnel conundrums that are sure to
arise.
The Problem of Bottlenecks
The bottlenecks result, in effect, from the unevenness
of the efficiency gains at different points in the
value chain. (See Exhibit 5.)
Consider the sixfold increase in target identification
described above. This escalating quantity of
targets could turn out to be not so much a glorious
profusion as an exasperating glut. Unless there is
some corresponding increase in the capacity to
process them downstream, these targets will simply
EXHIBIT 4
IMPACT OF QUALITY ON COST TO DRUG
Pre-genomics Post-genomics
Mix of novel and
known targets
Novel targets only
Novel targets and
novel compounds only
Benefits of experience
over 3-5 years
SOURCES: BCG analysis; industry interviews.
590
590
790
880
1,080
Approximate cost ($M)
EXHIBIT 5
UNEVEN PRODUCTIVITY GAINS CREATE IMBALANCE
Number today
to get one drug
Not to scale
2,400 1
Potential
targets
400
Potential
targets
108
Validated
targets
Increased
productivity
138
Lead
candidates
loiter at their source in a wasteful logjam. Or consider
chemical genomics. Implementing this
approach will build up huge pressure on screening
resources: sending unvalidated targets for screening
could involve a 120-fold increase. So too with
efficiency gains at other points in the value chain:
without the necessary downstream adjustments,
bottlenecks will inevitably develop.
Our capacity to do functional experiments was
completely choked by potential targets.
—VP of discovery,
major pharmaceutical company
But the problem is a dynamic one, and accordingly
very awkward to deal with. Ease one bottleneck and
you often create another downstream. Or ease it
too much and you convert it into a bulge—over-
72
Drug
candidates
Required
productivity 2
Targets Compounds
30
7
INDs Drug
ID Biology
Chemistry
Development
Target ID Target Validation
Screening Optimization
Preclinical Clinical
SOURCES: BCG analysis; industry interviews.
NOTE: Does not include impact of pharmacogenetics, to be addressed in
next installment.
1Number of targets identified by investing same resources post-genomics as
pre-genomics.
2 Productivity required to exploit all potential targets from target identification.

esourced in relation to the flow from upstream,
and hence wasteful once again. It will take some
adroit adjustment of resources and processes along
the value chain to restore a smooth flow.
This imbalance will affect incumbents—integrated
companies with established value chains—worst of
all. They have resources and processes in place;
changes are likely to be difficult and disruptive. To
implement the new genomics technologies is troublesome
enough, but then to have to redistribute
resources along the entire value chain will take real
determination. (To other companies, by contrast,
bottlenecks might represent very favorable opportunities.
In particular, genomics companies could
benefit. (See sidebar, “Upstart Start-ups—the
Competitors Classified.”)
The Human Factor
To flourish in the new genomics era, and possibly
even to survive, companies are going to have to
engage the new realities. It will not be easy. Some of
the new technologies will tend to overstretch or
even defy existing capabilities and organizational
structures. All along the value chain, processes and
resources are going to have to be adjusted.
The resources in question include human resources,
and retrenching, reassigning, or supplementing talented
personnel is a far from straightforward procedure.
But it will have to be done. Organizational
restructuring is likely to entail distressing upheavals
for corporate culture and personnel alike. The
strategies adopted for managing it will require constant
monitoring and fine-tuning. New modes of
cross-functional collaboration may need to be instituted,
new incentives offered, and so on.
I spend half my time looking for talent that isn’t
out there, and the other half worrying where they
would fit if I found them.
—Research director,
leading biotech company
* * *
In sum, implementing genomics technology will be
very tricky. It will almost certainly require a holistic,
cross-value-chain perspective. We will discuss potential
solutions to these operational challenges in the
third chapter.
A Final Word
By engaging affirmatively with the brave new
genomics world, companies are making it possible
to increase R&D productivity substantially. They will
bring to bear an array of industrialized processes,
informatics, and rich data sets—a formidable combination
that promises to boost efficiency, and even
improve success rates, all along the value chain.
Here we have discussed both the opportunities and
the challenges that arise when a company adopts
and implements genomics technologies that are
available today.
The opportunities add up to potential savings of
nearly $300 million per drug—about one-third of
the cost—and the prospect of bringing each drug
to market two years sooner. The challenges include
managing quality control and dealing with unfamiliar
operational predicaments: bottlenecks along the
pipeline and a host of organizational difficulties.
But for companies that choose not to meet the
genomics revolution head on, the challenge is even
greater: they will be unable to compete. These companies
do more than leave money on the table.
They face the inevitability of being left behind.
To reap maximum benefit from the new technologies,
companies will need to scrutinize their resources,
processes, and policies throughout the
value chain. Pharmaceutical and biotech managers
will need to ask themselves some taxing questions as
they begin to formulate their genomics strategy:
• Which specific genomics technologies and
approaches make the most sense for our company?
What investments and capabilities would be
needed to integrate these new technologies and
approaches successfully?
• What capabilities do we already have? What
investment are we prepared to make to acquire
those we lack?
21

22
UPSTART START-UPS—THE COMPETITORS CLASSIFIED
As the e-commerce revolution has demonstrated,
disruptive technologies tend to spawn start-ups that
aim to exploit the disruption, either as suppliers to,
or as replacements for, incumbents disoriented by a
changing world. In the case of genomics, the incumbents
are easy to identify: the traditional pharmaceutical
companies and the larger fully integrated
biotech companies. But who are the start-ups?
Genomics companies can be classified in three
broad groups, on the basis of how closely or distantly
they are related to the actual developing and
marketing of drugs.
The group furthest away contains the companies
that supply enabling technologies, in the form of
hardware or software. These companies resemble
the merchants of the California gold rush who sold
pickaxes to the miners, or more recently, companies
such as Cisco and Sun Microsystems that have been
providing the necessary infrastructure for the multitude
of e-commerce practitioners. Examples of such
companies are PE Biosystems, the supplier of highthroughput
sequencing machines, and Affymetrix,
the preeminent gene-chip manufacturer and supplier.
The second broad group contains the companies
supplying information and knowledge, including
those companies that generate proprietary databases
and offer access to them through subscriptions or
fee-per-use business models. One of the best-known
examples is Celera, which sells subscription-based
access to human and animal model-sequence data.
The group also includes companies that are attempting
to integrate and exploit those databases to conduct
in silico R&D. An example is LION Bioscience,
which integrates information from public and private
sources into a single platform to make targets and
leads easier to identify and analyze.
Finally, there is the group of companies that develop
and sell more traditional “physical” drug intermediates—targets
and lead compounds. We call these
platform and orchestrator companies.
Platform companies deploy proprietary technology in
the quest for promising targets and leads. One such
company is Aurora Biosciences, which has developed
proprietary high-throughput screening technology
to exploit an opportunity in screening and chemical
genomics. Another example is MorphoSys,
which has developed a platform for rapid development
of high-affinity antibodies, for use in target validation
and therapeutic antibody discovery.
Going one step further are orchestrator companies,
which string together adjacent platforms to create
optimized segments of the R&D value chain. As the
orchestrators extend their value chain, they can sell
drug candidates that have progressed further and
further downstream—and have thus become more
and more valuable. Although these companies are
still selling only intermediates, they show every sign
of graduating into fully integrated drug companies.
Millennium has already made that transition: concentrating
initially on genomics target discovery, it
has subsequently developed a full R&D pipeline in
its own right.
What is the outlook of each of these groups? The
first two (the pure suppliers, either of enabling technologies
or of information and knowledge) would
appear to be well positioned if they target areas of
scarcity (that is, bottlenecks) with proprietary prod-

ucts, or if they have enough clients to achieve scale
efficiencies reachable only by supplying multiple
companies. But so far, most of these companies
have struggled to find a sustainable, profitable business
model.
Meanwhile, the third group seems in the most promising
position. The pharmaceutical business remains
attractive, with margins averaging more than 80 percent,
so it is easy to see why so many genomics
companies aspire to become drug companies. But
• Where is industrywide scale to be found rather
than just company-level scale? Which capabilities
should we therefore develop in-house, and which
through partners?
• How will any new technologies affect the rest of
the value chain? How can we optimize decision
making and information flow up and down the
value chain?
•What are the implications for the organization of
the changes we wish to make? How feasible is the
necessary restructuring? And what would be the
most efficient way to carry it out?
there are many hurdles en route, and overambitious
companies risk tripping over them. Although many of
these companies may fail, those that succeed will
have a transformational impact on the industry.
Moreover, the traditional drug companies seem to be
moving in the opposite direction, increasingly outsourcing
portions of their R&D value chains. What is
going on? We will try to answer that question in the
third chapter, when we examine these trends in
more detail.
These questions can be addressed by thorough,
thoughtful analysis. Key investment decisions will
be required, as well as a carefully planned implementation
program to ensure that the value of
those decisions is captured.
In the next chapter, we turn to genetics and analyze
its likely impact on R&D productivity. In the final
chapter, we will examine more closely the strategic
choices and operational implications of the various
changes in prospect.
23

24
Chapter 2: The Impact of Genetics
Preface
Having discussed the genomics wave in the previous
chapter, and the way that it promises to enhance
R&D productivity, we now turn to the genetics wave.
Several broad differences suggest themselves immediately.
Where the genomics wave is technologydriven,
the genetics wave is better viewed as datadriven,
exploiting the known details of the human
genome and individual variations within it. Where
the genomics wave brings benefits mainly at the
drug-discovery and preclinical phases, the genetics
wave will prove its worth in both the earliest phase
and the later phases of the value chain—target discovery
and the clinic. Where the genomics wave
enhances R&D productivity mainly by securing
great improvements in efficiency (with only modest
improvements, if any, in success rates), the genetics
wave could boost success rates dramatically as well.
One further difference should be mentioned:
where our model for the genomics wave was put forward
with considerable confidence, our model for
the genetics wave is more tentative. At this early
stage, any assessment of genetics’ impact on the economics
of R&D is bound to be provisional. Certainly
genetics has huge potential: if all goes according to
plan, it will change R&D productivity beyond recognition.
But between that potential and its full realization
lie several years and many obstacles.
The potential consists in tremendous savings. First,
genetics can bring about great efficiency gains by
making it possible to shorten or even bypass various
steps in the value chain. Second, genetics holds the
prospect of transforming success rates: failures in
the R&D pipeline currently account for 75 percent
of the total cost to drug. But offsetting such opportunities,
dangers loom large. Riding the genetics
wave involves a greater risk than riding the
genomics wave alone—though it is more exhilarating
and, if the risks are successfully negotiated, ultimately
more rewarding. How to choose between discretion
and valor is a crucial strategic decision that
companies will have to make.
In analyzing the economic implications of genetics,
this chapter of our report considers the effect only
on pharmaceutical R&D. But genetics is likely to
affect health care far beyond R&D, in both the
short and the long term. In the short term, new
market opportunities should arise in the formerly
sleepy diagnostics sector. (Drug companies may or
may not be able to exploit these opportunities: see
sidebar, “Diagnostics—an Opportunity Too Good to
Miss…and Perhaps Too Good to Grasp.”) In the
longer term, genetics is likely to transform the
delivery of health care. Increasingly, diseases will be
redefined into various subtypes—a refinement that
should facilitate more appropriate care and more
“rational” drug design. The combination of new
diagnostics, new disease definitions, and new tailored
drugs should prove a winning one, and may
well usher in an era of individualized medicine.
R&D remains the focus of our analysis here, however:
specifically, the wide range of economic reactions
that R&D might show under the impact of the
new genetics information. We discuss the tremendous
opportunities as well as the accompanying

isks inherent in genetics-based R&D, and explore
various ways of managing them.
Two Kinds of Genetics Approaches
There are two relevant approaches to consider
when assessing the economic impact of genetics on
DIAGNOSTICS—AN OPPORTUNITY TOO GOOD TO MISS…
AND PERHAPS TOO GOOD TO GRASP
It will be several years before genetics fulfills its
promise. In the meantime, however, companies
might begin to enjoy a preliminary reward, in the
form of diagnostics—essentially a byproduct of their
broader genomics research programs. Certainly diagnostics
is the subject of great expectations, though
whether and how soon it will meet them remains to
be seen.
Many research projects in genomics and genetics
will devise diagnostic tests as a matter of course—in
parallel with research or simply as a preliminary
step, perhaps—without portraying them that way.
Diagnostic tests can be understood in a fairly broad
sense here. Disease genetics, for example, in identifying
a target, is in effect finding a marker of disease
susceptibility. Expression profiling, in identifying the
molecular differences characterizing a disease’s different
subtypes, is pointing the way to differentiated
and fine-tuned therapies. And pharmacogenetics, in
identifying variations in drug response among various
patients, could be helping to suggest the most
suitable drug for them.
The opportunities inherent in diagnostics will appeal
to drug companies at several levels. First, costs are
low. The intellectual capital needed to develop a diagnostic
test comes free, courtesy of existing research
in drug discovery and development; validation
studies can be run in parallel with drug efficacy studies,
or perhaps can even simply borrow their results
and extrapolate from them; and as for safety studies,
diagnostic tests don’t need any. All in all, then, the
incremental spending required to develop a marketable
diagnostic test is, relatively speaking, paltry.
R&D: disease genetics and pharmacogenetics. They
operate at different stages of the value chain.
Disease genetics is invoked earlier, during the discovery
phase: it involves the search for genes that make
people susceptible to particular diseases, with the
aim of then finding targets. Pharmacogenetics is the
Second, rewards are prompt. Diagnostics, in bypassing
most of the traditional steps of pharmaceutical
R&D, can be brought to market not only far more
cheaply than drugs, but far more quickly too. Drug
companies are thereby able to realize some unusually
fast payback on their R&D spending.
Third, the market outlook is favorable. As new therapies
proliferate, more diagnostics will be demanded;
and as technologies advance, new types of diagnostics
will become available. The signs are good.
These opportunities are to some extent offset, however,
if not by risks, then at least by challenges.
There is the challenge of novelty, for instance: for
many traditional companies, diagnostics would
involve manufacturing an unfamiliar kind of product—
a kit—and that in turn would involve developing new
capabilities, or else partnering with a dedicated diagnostics
company. Companies that have an in-house
diagnostics division, such as Hoffmann-La Roche,
Abbott, and Bayer, will have an advantage here.
Then there is the challenge of intellectual-property
rights: companies might find it more difficult to
assert those rights over diagnostics than over their
findings in pharmaceutical research.
Perhaps the most daunting challenge is timing: diagnostic
tests will tend to emerge too speedily, becoming
available sooner than the therapies they indicate.
So the chief appeal of investing in diagnostics—its
prompt availability—may be undercut. Drug companies
may have to delay marketing their diagnostics
(and thus delay capitalizing on the opportunities)
until their drug R&D pipelines catch up.
25

26
genetics-based form of pharmacogenomics (see
sidebar, “Pharmacogenomics—Some Definitions”),
and comes into play later, in the development
phase: it involves predicting the efficacy and side
effects of candidate drugs.
The data explosion detonated by genomics technology
has created vast amounts of genetic information,
ready for sifting. The findings of the
Human Genome Project and related endeavors are
merely the starting point. The ultimate goal is to
elucidate the genetic basis of human disease and
drug response. In the short term, genetics research
will enable scientists to predict disease susceptibility
and likely drug response in individuals; in the
longer term, it should help to improve the quality
of pharmaceuticals and medical diagnoses.
PHARMACOGENOMICS—SOME DEFINITIONS
Pharmacogenomics is the use of genomics
approaches to elucidate drug response. There are
three relevant approaches: via DNA, via RNA, and
via proteins, and three corresponding forms of pharmacogenomics:
pharmacogenomics using genetic
approaches (or pharmacogenetics), expression profiling
(or expression pharmacogenomics), and proteomics
(or proteomic pharmacogenomics).
Pharmacogenetics predicts patients’ drug response
by analyzing the genetic variations in their DNA. It is
the form of pharmacogenomics discussed in the
main text here.
Expression pharmacogenomics predicts patients’
drug response by analyzing their RNA levels—specifically,
by comparing the amounts of RNA found in
different samples to determine which genes are
expressed at different levels. An example: a research
group at The Whitehead Institute studying two very
similar leukemias (AML and ALL) has observed a
distinct difference in expression levels of specific
genes, and thereby provided a quick and reliable
method for differentiating them. Patients are now
less at risk of being misdiagnosed and being given
Attaining the short-term goal is, conceptually, simple
enough. The genetic codes of individuals differ
in tiny, but sometimes decisive, details. By comparing
an individual’s genetic variations against the
“standard” genome, scientists should be able to predict
whether that individual is at risk for a specific
disease, and, if so, how well suited he or she is to a
particular drug therapy—the work respectively of
disease genetics and pharmacogenetics.
The two approaches benefit R&D economics in different
ways. Disease genetics will improve efficiency
in target discovery and, by leading to the discovery
of particularly high-quality targets, will bring about
improved success rates in validation and downstream.
Pharmacogenetics, by enabling scientists to
select patients more appropriately for clinical trials,
an incorrect, and possibly lethal, drug treatment: in
effect, the test screens for adverse drug response.
Expression pharmacogenomics seems to be moving
from academic studies and biotechs into more mainstream
pharmaceutical R&D. Witness the recent purchase
by Merck and Co. of Rosetta Inpharmatics, a
biotech founded specifically to develop expression
pharmacogenomics.
Finally, proteomic pharmacogenomics predicts
patients’ drug response by analyzing their protein
levels—specifically, by comparing protein readings
in different tissue samples to identify proteins that
differ either in structure or in expression levels.
Consider the example of an aberrant fusion of two
proteins called Bcr and Abl, which occurs in more
than 95 percent of patients with CML (chronic
myeloid leukemia, which accounts for about 20 percent
of all cases of adult leukemia). This aberrant
fusion protein is present only in cancer cells. It distinguishes
itself from its normal counterparts by its
increased size. It can be used not only to monitor the
progression of the disease but also to test whether
Gleevec, a revolutionary new drug, would provide an
effective therapy.

will not only help to make the trials faster and
cheaper, but also allow some drugs to pass that
would otherwise have failed owing to poor efficacy
or side effects.
High-Risk, High-Reward Research
In the near-to-medium term, genetics promises to
reduce R&D costs dramatically. Or does it? Our
model shows that the application of disease genetics
and pharmacogenetics together could, in the very
best case, save as much as two-thirds of the current
cost to develop a drug and nearly two years. Of
these potential savings, the vast majority comes
from disease genetics. Of the remainder, some cannot
be clearly apportioned to either disease genetics
or pharmacogenetics but must be credited to
their joint efforts, being the product of synergies
realized when disease genetics information is incorporated
into pharmacogenetics-driven clinical trials.
(Although genetics can be combined with the
genomics technologies described in the previous
chapter, the potential savings are not additive. In
the next chapter, we will assess the total savings realistically
achievable through the application of genomics
and genetics.)
Our model also shows that realizing this potential
will be far from straightforward and is far from
guaranteed. The high-end savings estimate assumes
a positive resolution of several scientific, technical,
and market risks. With every setback, the savings
diminish. Too many setbacks, and the savings fall to
zero. Implementing genetics could even turn out to
have a negative impact on value, owing to adverse
market dynamics.
Such a double-edged sword is awkward to wield. If
companies fail to grasp it at all, they forfeit the
opportunity to reap enormous rewards—nearly
double the savings possible with genomics technologies.
If they do grasp it, they put themselves in
some danger, and will need to develop a genetics
strategy aligned with their risk profile and with specific
market conditions. As an approach to research
and development, genetics remains risky, but full of
promise, too. This chapter assesses the promise and
the risk alike.
Disease Genetics
Underlying many diseases are genetic variants, or
polymorphisms—alterations in the DNA sequence
of a given gene that influence individual risk of disease.
(The term polymorphism usually refers to a
common variant—one found in more than 1 percent
of the population.) Often the genetic difference
consists of just a single altered letter in the
genetic code (the variant is then known as a single
nucleotide polymorphism, or SNP). Such a tiny alteration
can have fatal consequences; for example, the
substitution of an A for a T in the hemoglobin ß
gene is responsible for sickle cell anemia. The goal
of disease genetics is to identify such DNA alterations;
so far, this goal has proved elusive for all but
the most strongly inherited conditions. (See sidebar,
“Disease Genetics—Various Approaches to
Variants.”) But researchers persevere, since finding
the causal variants can be a major step to finding a
treatment—and potentially a cure.
Uncertainty persists: the technologies have yet to
prove their full worth by producing the necessary
quota of practical results. But if all goes according
to plan, the savings realized by pharmaceutical
companies will be enormous.
The Potential
The savings would come partially from improved
efficiency in discovery, but principally from
improved success in target validation and clinical
trials. The improvements in efficiency result from
the consolidation of target discovery into a single
step; the improved success rates result from the
refining of target identification, making it possible
to pinpoint the targets associated with disease susceptibility
in humans.
Once the targets have been located, they can boast
a collective superiority in three particular respects.
First, their relevance to human disease is certain.
They have, after all, been validated in humans,
showing directly that modulation of gene activity
leads to alteration in the intensity or duration of
disease. (By contrast, other targets are usually iden-
27

28
DISEASE GENETICS—VARIOUS APPROACHES TO VARIANTS
When disease genetics is used to identify or test
variant genes, the investigation can take a variety of
forms. There are three key dimensions in the design
of such an investigation: narrow/broad, linkage/association,
and direct/indirect. These dimensions all
have a bearing on cost and the chances of success.
First, researchers can examine some of the genes (in
a narrow study, or candidate-gene study) or all of
the genes (in a broad study, or genome-wide scan).
Second, these genes can be examined through inheritance
patterns in families prone to the disease (in a
linkage study) or by comparing individual patients
with healthy individuals in the population at large (in
an association study).
Finally, within association studies, each variant gene
can be studied directly or indirectly: researchers can
test each variant individually for any involvement in
the disease (in a direct study) or test clusters of
closely positioned variants for the presence of a culprit
among them (in an indirect study).
The earliest disease genetics investigations, conducted
prior to the 1980s, were association studies,
and direct, and as narrow as could be—studying just
a single gene, which had been selected on the basis
of biological knowledge about disease mechanisms.
The researcher would seek polymorphisms in the
gene, and compare their frequencies in patients and
controls. This early approach did achieve some
notable successes, including, in 1956, the first discovery
of an inherited genetic variation found to
cause disease—the variant underlying sickle cell
anemia. The approach had a serious limitation, however:
it allowed very few genes to be examined, and
it required a prior hypothesis.
In the 1980s and 1990s, attention turned to families
showing an inherited pattern of disease. The
investigations took the form of broad linkage studies,
and were tremendously successful in identifying
some genes responsible for single-gene disorders,
notably the cystic fibrosis gene in 1989. Such studies,
being genome-wide, were now unbiased and
comprehensive. But the actual identification of
genes remained a slow, painstaking laboratory
process. So the early versions of such studies were
really suitable only for monogenic diseases. Hopes
were raised in the early 1990s, when companies
such as Millennium, Sequana, and Myriad were set
up to develop and exploit these techniques in the
quest to identify the genes implicated in common
polygenic diseases. Their initiative seems to be
stalled at the moment: although the localizing of
disease-related genes has become more efficient,
the actual locating of them remains discouragingly
difficult. That task would be better suited to association
studies.
Given the limitations in genome-wide linkage studies,
association studies have recently come back into
fashion, fortified by the efforts of the Human
Genome Project, Celera, and the SNP Consortium.
These studies have high-throughput technologies to
undergird them, as well as comprehensive databases
of gene sequences and SNPs.
Of the two possible approaches here, direct and indirect,
the former looks like a very formidable task.
The researcher conducting a direct association study,
and aiming to find the actual polymorphism underlying
the specified disease, is confronted by the entire
set of common variants in the genome—expected to
number some ten million. To examine such a horde
of variants with current technology would be inordinately
time-consuming and expensive.
Hence the hopes—and funds—now being invested in
indirect studies. Since variants in close proximity
tend to form clusters (known as haplotypes), it may
be possible to track down the disease-related polymorphisms
using only a small proportion of all
SNPs. In light of recent research findings, indirect
association studies of this kind do look practicable,
if not quite imminent.

tified through the use of animals and tissue cultures,
and so their relevance to human disease is
largely speculative.) In other words, there is no possibility
of failure in target validation, because identified
targets are ipso facto validated. (This is, of
course, no guarantee of their drugability—their
responsiveness to small-molecule intervention.)
It would not be possible to overstate the value of in
vivo human validation. Most of what passes for
target validation today is largely conjectural in
relation to the disease in question.
—Diabetes researcher,
Harvard Medical School
Second, the frequency of the causal polymorphisms
is known at the outset. If a study identifies multiple
genes associated with a particular disease, it will
also reveal their relative culpability. Consider the
example of Alzheimer’s disease, a heritable but
genetically complex disorder. On the one hand,
there are variants in three genes—PS1, PS2, and
APP—that are very rare but very potent: if a person
has any of them, he or she is almost certain to
develop Alzheimer’s. On the other hand, there is
the ApoE4 polymorphism of the ApoE gene, which
has a more modest effect on disease susceptibility
but is much more common in the population at
large, and among patients with Alzheimer’s.
Information of this kind can be useful for predicting
a drug’s potential marketability: although it
might be equally feasible to develop a drug that
influences the rarer variants, a drug targeting
ApoE4 might expect broader effectiveness, and
thus a larger market, and so might take precedence
in further research. (The rarer variants may still be
worth pursuing, using pathway analysis.)
Finally, the nature of the relevant polymorphisms is
known—different disease-inducing mechanisms
among variant forms, for instance. Such information
may help to streamline clinical trials if used by
efficacy-based pharmacogenetics to identify “nonresponders”—patients
who lack the crucial DNA
alteration, and hence are unlikely to experience the
intended effect of a candidate drug—and exclude
them from the trials. (In modeling the potential of
disease genetics, we have included this effect of
efficacy-based pharmacogenetics. See the section
on pharmacogenetics below for further details.)
Depending on the approach taken, cost savings per
drug could be as great as $420 million, with the
potential time savings ranging from 0.7 to about 1.6
years (producing an added $290 million of value
per drug). (See Exhibit 6.) Of the cost savings, the
vast majority would be yielded by the improvements
in success rates: $390 million, consisting of $110
million in validation and $280 million in the clinic.
EXHIBIT 6
DISEASE GENETICS OFFERS GREAT SAVINGS POTENTIAL
Cost to drug
Pre-genomics
Candidate
gene study
Genome-wide scan
Genome-wide scan
plus pathway analysis
Time to drug
Pre-genomics
Candidate
gene study
Genome-wide scan
Genome-wide scan
plus pathway analysis
ID Biology
Target ID Target Validation
0
0
200
5
Chemistry
400
Screening Optimization
10
460
485
455
600
800
13.1
15
14
14.7
Development
16.5
Preclinical Clinical
SOURCES: Industry interviews; scientific literature; public financial data;
BCG analysis.
880
1,000
$M
20
Years
29

30
Efficiency improvements in target discovery account
for the remaining savings.
The Uncertainty
For these vast savings to materialize, two requirements
will have to be met. First, disease genetics
must prove scientifically feasible for the relevant
common diseases. Second, not only must studies in
humans work; in addition, the targets they identify
must be drugable; failing that, identifying the disease
genes is pointless, and all the effort that has
gone into finding them will be wasted. (See sidebar,
“Drug-Resistant?—Are Disease Genes Drugable
Targets?”)
Feasibility—the Limitations of Technology
Fundamental technological concerns still hover over
disease genetics. Can it actually be done? The results
so far have been very modest. The bonanza of
clearly documented disease-susceptibility genes for
common multigenic diseases has yet to materialize.
Candidate gene studies, for instance, are by definition
limiting: they focus on a subset of genes defined
by a prior hypothesis, and therefore risk excluding
some crucial culprits. And genome-wide linkage
studies, although highly successful in addressing
single-gene diseases, have proved disappointing for
DRUG-RESISTANT?—ARE DISEASE GENES DRUGABLE TARGETS?
The skeptics pose an awkward question: Will disease-related
genes ever prove to be drugable in significant
numbers? The record so far is hardly encouraging.
Some disease genes, such as CFTR in cystic
fibrosis, were identified long ago, yet have failed to
generate therapeutics. The infrequency of success
stories, such as Ceredase—a drug for type I
Gaucher’s disease that was essentially a creation of
disease genetics—only highlights the general trend
of failure.
These long-identified disease genes tend to be for
single-gene disorders, however. And such disorders
are by their nature difficult to cure. They are binary
phenomena: the gene is broken, you get the disease.
the more common multigenic kind of disease: a disease-related
gene might be accurately pinpointed in
affected families (such as the BRCA1 breast cancer
susceptibility gene), only for it then to show very low
prevalence outside the families used in identifying
it. True, these two approaches might become more
tractable now, in the wake of the sequencing of the
human genome and the development of comprehensive
SNP maps (catalogs of the characteristics
and locations of SNPs in the genome).
As for genome-wide association studies, considered
by many experts to be the most promising of all,
they have only recently became practicable: all the
requisite tools (a full genome sequence with a SNP
map to match, genotyping technologies, and so on)
appear to be in place. But the approach remains virtually
untested, owing to the still exorbitant cost of
genotyping. The preferable form of genome-wide
association studies would clearly be the indirect
kind—still covering the entire genome, but genotyping
far fewer SNPs—yet even here the current
cost is a prohibitive $400 million or so for each disease
investigated. Within five years, however, genotyping
costs are expected to fall to $20 million or
less, and the essential proof-of-concept tests can
then take place more routinely. (See sidebar,
Finding a small-molecule therapeutic to repair a completely
defective protein is an extremely difficult challenge.
(Indeed, Ceredase is a protein therapeutic.)
Most disorders, by contrast, are attributable not to a
single gene but to multiple genes, and perhaps other
factors too. This means that the system as a whole
can still function, just hampered to a greater or
lesser degree. Such cases benefit from patching up,
so drugs can be beneficial without actually constituting
a cure. There is little reason to doubt that
such palliative drugs will soon emerge in abundance,
as disease genetics becomes ever faster at
identifying some of the genes implicated in multigenic
disorders.

GENOTYPING—HOPES RISE AS PRICES FALL
For companies wishing to pursue disease genetics,
one of the major stumbling blocks has been the prohibitive
cost of genotyping. The current average
genotyping cost in the industry is about 50 cents per
SNP. At that price, even narrow candidate-gene studies
could cost as much as $15 million; a direct
genome-wide association study could cost upwards
of $5 billion.
Matters are about to change, however. Costs are
declining, and are expected to continue to fall dramatically—as
much as 50-fold—over the next few
years, thanks to competition and customer demand
on the one hand, and improved automation and
“Genotyping—Hopes Rise as Prices Fall.”) And
once that happens, the two major questions outstanding
can be resolved: the number of patients
needed to attain statistically significant results, and
the utility of SNP maps.
Regarding patient populations, first: the more
patients, of course, the easier it is to discern a true
difference above randomly occurring fluctuations
(noise)—and the more expensive. Size your sample
too sparingly, and you risk emerging empty-handed;
too generously, and you overspend. Striking the
right balance is tricky. It depends on how common
the sought-for SNPs are, and that is very difficult to
estimate. It also depends on the strength of the
association between the disease and the suspect
polymorphism. For common multigenic diseases,
susceptibility depends on a specific combination of
several genetic changes, and is influenced by environmental
factors as well, which weakens the association.
The weaker the association, the larger the
sample needed to detect the influence of a specific
gene. Perhaps the issue will dissolve before being
resolved. If prices plummet as expected, and if
databases become as comprehensive as hoped, sufficient
sample sizes will become easily affordable.
SNP maps can help in the quest to identify a
disease-susceptibility gene, but only if two condi-
miniaturization (which allows companies to reduce
their consumption of expensive reagents) on the
other.
Our calculations are, accordingly, based on a cost of
one cent per SNP genotype—a likely price across the
industry, according to expert consensus, within the
next five years. (Some companies must already be
benefiting from genotyping costs considerably lower
than the current industry average.) That said, the
cost of conducting disease genetics studies will
remain far from negligible, and companies will need
to continue to take it into account when assessing
risk.
tions prevail. First, for any association studies
(whether narrow or broad) to work, the genetic variants
associated with the disease need to be fairly
common—prevalent in more than 1 percent of the
population at large—and that is far from guaranteed.
Then, these polymorphisms need to be either
recognizable (that is, they must produce discernible
changes in a protein) or at least detectable by an
indirect measure (called linkage disequilibrium;
that is, the presence of a particular SNP cluster in
individuals with a given disease). Assuming reasonable
costs, indirect genome-wide association studies—of
all the approaches, the one most likely to
prevail—would, according to our model, result in a
total savings of $395 million in cost per drug on average,
with 0.7 years of time saved to market (producing
an additional $260 million of value per drug).
Of course, for that to happen, you need to do more
than find disease genes—you have to turn them to
good effect by ultimately producing a drug. And
here disease genetics presents a further challenge:
given the long odds involved in pharmaceutical
R&D, will the number of targets yielded by disease
genetics be sufficient?
Practicability—the Limitations of Human Studies
The problem is that human studies—identifying
polymorphic genes in humans—will tend to pro-
31

32
duce targets high in quality but low in quantity, perhaps
just one to five for an average disease. That
would produce at best a 25–30 percent chance of
yielding a drug, given that chemistry and development
remain far from fail-safe. For disease genetics
to live up to its promise, it will need to improve
those odds considerably. And to do that, it will have
to call on a supplementary technique: pathway
analysis.
Disease-susceptibility genes, if identified through
disease genetics, serve not only as targets themselves,
but also as guides to additional targets. The
genes form part of broader disease pathways, and
these pathways contain other molecules that may
serve as targets (perhaps 10 to 15 targets per pathway,
according to experts). These new targets are
identified by pathway analysis, often taking the form
of the study of “simple” experimental systems, such
as those of Drosophila melanogaster or C. elegans (a.k.a.
fruitflies and worms). When studied in the laboratory,
these systems disclose the genetic components
of a given pathway. (Other approaches to pathway
analysis include expression profiling of tissue samples
and studies of protein-protein interaction.)
Implementing pathway analysis will result in a lower
cost per drug on average: it improves efficiency.
Costs are reduced because pathway studies are relatively
inexpensive and fewer human-derived targets
are required—pathway analysis expands the pool of
potential targets tenfold or more. As targets, they
prove to be of high quality, moreover (in keeping
with the disease-susceptibility gene that inspired
their discovery), achieving good success rates in
clinical trials. For although they themselves are not
yet validated in humans or clearly implicated in the
disease, they participate in a pathway that is.
So pathway analysis should give human studies the
requisite boost, with enough targets emerging to
yield a drug more often than not. On the down
side, there are the time and cost of additional animal
research and the loss of some advantages in
clinical trials. Adding pathway analysis, via genetic
studies of Drosophila melanogaster, to indirect
genome-wide association studies would result in a
total savings of $425 million for each drug on average,
regardless of the original human study
approach, though it would add nearly two years of
additional work (producing an additional $255 million
of value per drug).
The first one to do genetic studies takes a huge risk.
If it works, you’ll see everyone running to join the
crowd.
—R&D executive,
leading pharmaceutical company
Implementing Disease Genetics
The savings promised by disease genetics are enormous,
and companies cannot ignore them. But they
cannot ignore the risks either, and will need to exercise
rigorous selectivity and discipline when it comes
to pursuing specific disease genetics studies—which
approaches to adopt, for instance, and which diseases
to explore. Placing bets in this way is going to
be nerve-racking enough. But companies face
another difficult set of choices as well, in the various
operational issues that need to be addressed.
Placing Bets
Some diseases have clear appeal as objects of
genetic research: asthma, Alzheimer’s disease, and
diabetes, for example, being complex diseases that
afflict large populations and obviously contain heritable
factors. They have already become competitive
areas of study. Although opinion is still divided
over the likely impact of disease genetics, substantial
bets are being placed by various companies—
emerging biotechs and established pharmaceutical
companies alike. A few claim they are already seeing
benefits from their investments: following its
alliance with Roche, deCODE genetics claims to
have succeeded in identifying a gene contributing
to cerebrovascular disease; GlaxoSmithKline has
announced finding genes associated with migraine,
Type II diabetes, and psoriasis; and Genset, a
French biotech company, is reported to have identified
genes implicated in prostate cancer and
schizophrenia.
But all companies embarked on, or intent on, pursuing
disease genetics must acknowledge that it is

still a scientifically risky endeavor. As they construct
their customized portfolio of bets, they will have to
keep reviewing not just their internal capabilities but
also the degree of risk they are prepared to accept.
Putting Disease Genetics into Operation
If companies do opt to participate in disease genetics,
they may still choose to maintain some distance
by outsourcing the activity or licensing in the
results. Those that decide to launch a major disease
genetics program in-house will confront a number
of significant challenges as they put the program
into operation.
Take the problem of obtaining the required samples,
for instance. From whom should samples be
collected? And how are samples to be stored? The
use of human tissue raises ethical considerations as
well. What constitutes consent? How can privacy be
protected? Who “owns” the tissue material? And
who should profit from it? (Several companies,
such as Genomics Collaborative and the not-forprofit
First Genetic Trust, are emerging to address
these conundrums.)
And, of course, there are major organizational
questions. What are the implications for human
resources and labor relations? For big pharmaceutical
companies, new capabilities will be demanded,
and new skills will have to be acquired—statistical
geneticists, for instance, and experts on pathway
genetic studies. And other capabilities may suddenly
be less in demand—perhaps even obsolete.
We will discuss implementation issues such as these
in more detail in the final chapter of this report.
Pharmacogenetics
Just as some genetic variations among individuals
may influence their susceptibility to diseases, so
others may influence their responsiveness to drug
treatments for those diseases. It is the goal of pharmacogenetics
to seek out and characterize some of
these latter variations.
The savings that pharmaceutical companies might
hope to harvest are considerable, though nothing
like as high as those that disease genetics stands to
achieve in ideal circumstances.
The Potential
The impact of pharmacogenetics on R&D productivity
will derive from the increased flexibility it
introduces into clinical development. Currently,
drug-development policy is dominated by a binary
scenario in its later stages: either shepherd a compound
through its clinical trials and out to market,
or abandon it as unmarketable if it stumbles in the
trials. Pharmacogenetics provides a more nuanced
scenario, with an expanded range of possible outcomes,
by allowing the exclusion of patients genetically
predisposed to respond poorly to the drug.
Two particular benefits emerge: “standard” clinical
trials can now be streamlined; and “failing” compounds
can now be salvaged. (See Exhibit 7.)
The streamlining of trials would apply only to compounds
destined to proceed successfully through
the clinical trial process anyway. That path can now
be made smoother. The trials can be designed more
subtly. They can be smaller and quicker than
before, now that it is possible to preselect promising
patients—that is, patients whose genetic
makeup is likely to maximize the drug’s efficacy
and minimize its side effects. So, patients lacking
the drug-susceptible variation of the target gene
EXHIBIT 7
PHARMACOGENETICS EXPANDS DEVELOPMENT CHOICES
Proportion of patients showing
poor or no response
High
Low
SOURCE: BCG analysis.
Current options
Abandon drug
before market
Continue clinical
trials to market
Options available with
pharmacogenetics
Continue trials safely by
excluding at-risk patients
Optimize clinical trials,
making them
smaller and shorter
33

34
would be excluded from the trial, in order to show
high efficacy levels for the subset of patients who
would eventually use the drug. Also excluded would
be patients having a specific genetic variation associated
with side effects.
To see the streamlining effect of such exclusions,
consider the case of Herceptin, a treatment for
advanced breast cancer. It is effective only in a subset
of patients—the 25–30 percent whose tumors
overexpress the HER2/neu oncogene. It is this
gene that serves as the drug target. By screening for
HER2/neu expression, Genentech was able to
exclude nonresponders—some two-thirds of the
subjects originally tested—early in the clinical trial.
Without this prescreening, Genentech would have
needed nine times as many patients in phase III to
achieve significant results. The cost of such a trial
would have made Herceptin economically unviable.
Turning to the second benefit, for failing compounds
pharmacogenetics lowers the hurdle by easing
the conditions for market viability. Consider
specifically those candidate drugs that reveal serious
side effects in a significant proportion of the subjects
(such as the 7 percent of Caucasians with low
levels of CYP2D6, an enzyme that helps metabolize
some 25 percent of all drugs). Traditionally, any
such drug would be perceived as too risky to market,
and would be abandoned in preclinical studies.
Today, however, pharmacogenetics makes it possible
to identify the at-risk patients, so the drug would not
be disqualified right away, and could go on to prove
marketable—patients would just need to be tested
for vulnerability before being given prescriptions.
These are the potential benefits to R&D, the primary
focus of this report. Even greater potential,
some observers believe, may lie in market advantages.
Three such advantages are possible: price
premium, share shift, and new patients. In other
words, if a perception emerges that the pharmacogenetics-assisted
drug is distinctly less risky or distinctly
more efficacious for the (now restricted) target
patient population, payers may tolerate a higher
price for the drug, physicians may favor it when
offering new patients a prescription, and patients
who have shunned previous medications (owing to
side effects, typically) may now choose to try it.
Although it seems reasonable for pharmaceutical
companies to expect some market upside from
more efficacious, better tolerated therapies, it
remains to be seen to what extent they will be able
to reap these market rewards.
Putting figures on the cost savings pharmacogenetics
benefits might achieve, our model estimates an
average of $335 million in the cost to drug—if
pharmacogenetics were to work every time it were
applied. But pharmacogenetics won’t work every
time. Given the set of cases where it is applied and
succeeds, the expected savings would average about
$80 million, as discussed below. And of course, corresponding
to the potential market upside, there is
the counterpart scenario—potential destruction of
value in the market. Why the uncertainty? (See
Exhibit 8.)
EXHIBIT 8
PHARMACOGENETICS’ POTENTIAL IS CONTINGENT
Cost to drug
Pre-genomics
Pharmacogenetics:
the promise 1
Pharmacogenetics:
expected savings 2
ID Biology
Target ID Target Validation
0
200
Chemistry
400
Screening Optimization
600
545
800
800
Development
Preclinical Clinical
SOURCES: Technical literature; industry interviews; publicly available information;
BCG analysis.
1Savings per drug assuming pharmacogenetics can be applied across the
R&D pipeline.
2 Average savings across R&D pipeline, given scientific and market limitations.
880
1,000
$M

The Uncertainty
The sizable potential savings are mirrored by sizable
risks—market and regulatory risks this time, as
well as scientific and technical risks. In some circumstances,
market dynamics might be so unfavorable
that companies would be well advised to step
back and forgo the potential savings altogether.
Once again, given the range of possible outcomes,
the economic impact could ultimately be a negative
one—far from resulting in savings, applying pharmacogenetics
could result in a net loss for R&D. To
assess pharmacogenetics realistically, companies
need to ask two questions: How feasible is it? And
how desirable is it?
Feasibility—the Technological Limitations
Pharmacogenetics will not apply to all drugs. It will
apply only where differing drug response is due
entirely to genetic variation, and where that relationship
can be elucidated.
It is fairly rare for both of these conditions to prevail.
For one thing, biology-based variation in drug
response can be due in part to environmental factors:
grapefruit juice, for example, is known to
modify the effect of certain drugs in certain individuals,
sometimes raising the uptake to dangerous
levels. For another, drug-response variation will
often be the work of multiple genes, acting severally
or jointly, and that compounds the statistical and
technological difficulties of the search.
The business guys hear about this stuff, and are
like, “Great! Make it happen!” We’re left scratching
our heads, looking like poor sports, because a lot of
it just isn’t possible.
—Senior scientist,
major biotech company
Even if the two conditions are fulfilled, a further
challenge lies in wait—to find the relevant genes in
time to streamline the trial. For pharmacogenetics
to effect this streamlining, you need to be able to
screen out nonresponders. And that means finding
those variants, or identifying the nonresponder
genotype before designing any streamlined trial.
Now, developing a robust pharmacogenetic test
would generally require more than 1,000 patients.
A phase I trial is far too small for that purpose, so
no streamlining would be possible for a phase II
trial, despite the excitement surrounding that
prospect. Streamlining could be possible in time for
phase III trials, but even that is far from assured.
(See Exhibit 9.) (The other kind of screening test,
for side effects, is less problematic, since the associated
metabolic variations are often determined in
preclinical trials. But that kind of test does not help
to streamline clinical trials, which have to be sized
to test for efficacy rather than for side effects.)
It is hard to see how these phase II trials will be
used for pharmacogenetics, because most of the
variants are expected to be relatively infrequent.
Certainly, 30 percent prevalence [which falls
within standard clinical trial sizes] would be relatively
rare.
—Geneticist,
The Whitehead Institute
EXHIBIT 9
STATISTICAL REQUIREMENTS LIMIT PHARMACOGENETICS’
POTENTIAL TO STREAMLINE TRIALS
Number of patients required to develop a test
5,000
4,000
3,000
2,000
1,000
0
10
Typical phase II trial size
Typical phase I trial size
20
30
Prevalence >30% needed to
develop a test prior to phase III trials
40
50
60
Frequency of responder genotype
SOURCES: Technical literature; industry interviews; publicly available information;
BCG analysis.
NOTE: Graph analyzes a scenario from Nature Biotechnology, vol. 18, May
2000. Scenario uses efficacy pharmacogenetics to identify the ApoE
responder genotype, a predictor of efficacy of Cognex (tacrine) for
Alzheimer’s. Strength of effect, a different variable, is not considered here.
35

36
Given these technological limitations, we estimate
that less than 15 percent of drugs will be amenable
to the application of pharmacogenetics.
Desirability—Market Economics
As already mentioned, there are circumstances in
which a company might have an incentive to shun
pharmacogenetics entirely. After all, by excluding
patients from trials, you are in effect giving the
drug a restricted label when trying to market it.
Gauging the likely effect of a restricted label
involves some complex analysis. For a start, you
need to consider two distinct groups of patients:
those who take a prescription for the full course of
treatment (which could last many years, or even the
remaining lifetime for those suffering from chronic
diseases), and those who embark on a prescription
but then discontinue it for reasons of inefficacy or
side effects.
The pharmacogenetics test would shrink these two
potential patient groups in different ways. From the
former, it would eliminate the “placebo responders.”
From the latter, it would eliminate some of
the nonresponders and negative responders.
Market fragmentation has happened in many
industries—the marketing group can’t put their
heads in the sand. We have to figure out what to do
about pharmacogenetics.
—Genetics director,
leading biotech company
Since pharmacogenetics seems to be chipping away
at a drug’s market base, why pursue it in the first
place? The answer may lie, in part, in competitive
dynamics and game theory: companies may have to
embrace pharmacogenetics because their competitors
are doing so. Merck & Co., for example,
according to a recent Wall Street Journal article, is
busy developing capabilities to reproduce pharmacogenetic
analyses conducted by its competitors, if
only to disprove any claims that a rival drug might
be superior to its own.
But the compensatory advantages can be more positive,
too—the potential for market upside, once
again: price premium, share shift, and new patients.
What a company has to judge, before adopting
pharmacogenetics for any drug in development, is
the likely breakeven point—the point at which a
price premium or increased market share begins to
offset the volume loss. Our model assumes a modest
market premium of 20 percent, and calculates the
breakeven point in various scenarios, based on four
different approaches to pharmacogenetics. (See
sidebar, “Pharmacogenetics—Four Applications,”
and Exhibit 10.)
Efficacy-based pharmacogenetics can reduce trial costs
considerably. But the market dynamics could then
cast a cloud over that economic picture. If the
restricted label, by disqualifying placebo responders
and some nonresponders and negative responders,
translates into an overall revenue loss of just 2
percent, that cancels out the savings achieved in the
clinical trials.
EXHIBIT 10
PHARMACOGENETICS’ VALUE DEPENDS ON MARKET
DYNAMICS
Patients lacking good response (%) 1
100
80
60
40
20
0
Conduct normal trials
Abandon drug
100
SOURCES: Industry interviews; BCG analysis.
Pharmacogenetics can optimize clinical trials
200
Pharmacogenetics
trials make drug viable
300
Revenue increase required from market premium (%)
1Example based on a scenario in Nature Biotechnology, vol. 18, May 2000
of ApoE4 efficacy in tacrine response; assumes response rate of 41 percent
among patients with the SNP versus 20 percent among those without it;
also assumes 50 percent of nonresponders discontinue.

PHARMACOGENETICS—FOUR APPLICATIONS
To evaluate pharmacogenetics properly, companies
need to take an especially close look at the market
dynamics. These dynamics vary according to how
pharmacogenetics is used. We have identified four
such applications (each associated with a different
category of patient responding to a given drug). The
first three are used to exclude patients from trials;
the fourth is used to expand the potential market for
the drug.
First, efficacy prediction identifies patients who will
show no real or significant response to the drug—
perhaps because they metabolize the drug in an
unusual way, or have an unusual form or combination
of susceptibility genes. A typical drug produces
this negligible response in about a third of patients,
but sometimes the proportion is far higher. For
example, Cognex (tacrine), the first drug for
Alzheimer’s, is inefficacious in more than 50 percent
of patients. The varying response is associated with
differing versions of the ApoE gene, and is therefore
readily predictable by a pharmacogenetic test.
Second, common-side-effect prediction identifies
patients likely to experience familiar side effects, as
a result of metabolic difficulties caused by wellknown
enzymes. A test can screen out negative
responders—“slow acetylators,” for instance. The
acetylation polymorphism in the NAT2 gene is one of
the commonest genetic variations in drug metabolism;
it has the effect of reducing the enzyme’s lifespan
and thus reducing the effective amount of the
enzyme in cells at any one time. This polymorphism
is present in more than 50 percent of Caucasians,
who are thus at greater risk of drug toxicity.
Knowledge of this polymorphism could save a drug
in clinical trials that would otherwise be abandoned.
Third, very-rare-side-effect prediction identifies patients
at risk for unconventional side effects, but
comes into play only after the drug is on the market.
Unlike most of the common side effects, which are
associated with metabolic pathways and usually
emerge in preclinical studies, these rare side effects
tend to be provoked by nonmetabolic genes, and to
be overlooked at first. They cannot easily be predicted,
since there are too many possible sources
(modifications of the target or of the disease pathway,
or unrelated pathways), and they may occur too
rarely to show up in clinical trials.
A case in point is Lotronex, a drug for irritable bowel
syndrome, now withdrawn from the market. Only
after its market launch, and 450,000 prescriptions,
did its severe side effect (bowel impaction) become
apparent. About one in 6,500 patients was
affected—a frequency far too rare for a standard
clinical trial to detect beforehand. (A typical trial
involves about 5,000 patients: for this side effect to
have been manifest in a statistically significant way,
a trial of nearly 100,000 patients would have been
needed.) Pharmacogenetics could in certain cases
come to the rescue of such compromised drugs, by
belatedly devising a screening test.
Finally, market expansion identifies patients who are
currently unsuited to the drug but potentially responsive
to it. Since fine-tuning of dosages or formulation
can often reduce side effects and occasionally
improve efficacy, pharmacogenetics could reassess
and upgrade many of the supposedly ineligible
patients. The market for the drug might expand considerably
as a result.
Take the case of cyclophosphamide, a chemotherapy
drug, which works only when metabolized by the
enzymes CYP3A4 and CYP3A5. Some patients
appear underresponsive to it: a genetic variation
suppresses the activity of the enzymes, thereby
decreasing the amount of active drug in the bloodstream.
The best course is not to discontinue the
drug, but to compensate by taking a higher dosage.
A pharmacogenetic test could identify the appropriate
patients prior to treatment, and their consumption
of the drug, instead of declining to zero, would
actually increase.
37

38
Side-effect-based pharmacogenetics for common side effects
can save a candidate drug that would otherwise fail.
This form of pharmacogenetics does not streamline
trials; in fact, it imposes a moderate increase in costs,
since more patients have to be recruited initially for
the vetting process. It requires a smaller upside to
break even than efficacy-based pharmacogenetics,
however, since its powers of exclusion apply only to
the second type of patient (those who would ordinarily
try the drug and then discontinue it). They do
not apply to the first group (those who would take a
full course of the drug), since placebo responders
do not suffer from side effects.
Side-effect-based pharmacogenetics for very rare side effects
is the type that would be applied for a drug already
on the market. Once the number of adverse events
(instances of severe side effects) reaches a critical
mass, the drug’s reputation suffers, and its continued
marketability is jeopardized. (In severe cases,
regulatory agencies require the drug to be removed
from the market.) To salvage it would involve implementing
screening tests for all potential patients—
a kind of postmarket surveillance. This type of
pharmacogenetics would increase costs fairly
steeply, yet it could still make economic sense if the
drug were saved.
The economics hinge on a paradox: the fewer the
adverse events, the harder it might be to save the
drug. To identify the culprit genetic marker for use
in the screening test, you need a certain minimum
number of patients who have experienced the side
effect. That could be as low as 20 (assuming you
achieved a 100 percent association with a single
SNP marker), and that would carry the modest
price tag of $100,000 (assuming the expected genotyping
cost of one cent per SNP). But the required
number could be 2,000 (assuming you achieved
only a 10 percent association), and the likelihood is
that such a number of side-effect sufferers would
simply never emerge.
It’s a crime that a very small percentage of patients
can sometimes eliminate an otherwise highly beneficial
drug from the market. Pharmacogenetics benefits
everyone here.
—Research executive,
leading pharmaceutical company
Finally, market-expansion pharmacogenetics for the
most part has highly favorable economics. Since its
effect is to expand rather than contract the market,
all it needs to ensure is that the expansion be large
enough to cover the cost of the required trial.
The prospects hinge to some extent on the incidence
of the genetic variant. Consider two drugs,
one producing side effects in poor CYP2D19
metabolizers (including 20 percent of Asians) and
the other in poor CYP2D6 metabolizers (including
7 percent of Caucasians); and suppose that dosage
adjustments could resolve the side effects. It might
turn out that the former case warrants the investment
and the latter does not, given the difference
in their potential market expansions.
All in all, these limitations reduce the expected savings
that pharmacogenetics would bestow on an
average drug to about $75 million. But of course
there is no such thing as a true “average” drug. In
some cases, pharmacogenetics could yield potential
savings as high as $335 million and potentially capture
additional upside through price premiums or
an increase in market share. In other cases, it might
save nothing, or even destroy value in the market.
The key to success is to be selective.
Implementing Pharmacogenetics
Although the savings attainable through pharmacogenetics
appear less dramatic than those attainable
through disease genetics, they are in the right circumstances
quite substantial. And the total value
added would be enormous if the hoped-for market
advantages were realized.

Realizing this value is a matter not only of market
dynamics, but also of various more speculative factors:
how acceptable pharmacogenetics will prove
to payers, patients, physicians, and regulatory agencies;
how readily physicians and patients will
embrace the screening tests to generate share shift;
and so on. So the different types of pharmacogenetics
will probably come into effect at different
times. Detection of rare side effects will probably be
introduced first, as pharmaceutical companies are
highly motivated to save drugs from failure. Efficacy
pharmacogenetics will probably progress on a
slower timetable, owing to concerns about market
fragmentation. It might even take FDA action to
turn efficacy testing into a routine procedure.
When contemplating their pharmacogenetics policy,
companies will need to scrupulously analyze
specific drugs and markets. Deciding shrewdly just
where and when to apply pharmacogenetics, for
instance, will mean assessing market dynamics earlier
than ever before in the clinical trials phase. And
that in turn will demand new decision-making
processes and communication channels, including
stronger ties between research and development,
and between R&D and commercial activities. It is
on operational and organizational issues of this
kind that the spotlight will fall in the next chapter
of this report.
A Final Word
If the new genetics can realize its full potential, the
economics of pharmaceutical R&D will undergo a
metamorphosis. Efficiency will improve handsomely
and success rates will surge. The sums saved
could exceed a half billion dollars per drug, more
than halving the current cost.
That prospect is far from assured. There are
enough risks and uncertainties to temper excite-
ment. The range of possible outcomes is wide, and
companies will have to examine minutely and apply
selectively the various genetics opportunities.
Contrast genomics technology: the productivity
improvements promised by its implementation may
be more modest, but they are clearly achievable,
despite the operational challenges. With genetics,
the operational challenges are formidable too, but
they are compounded by less distinct and possibly
more intractable challenges: technological limitations,
scientific unknowns, and (in the case of pharmacogenetics)
the vagaries of the marketplace.
So, companies determined to acquire and exploit
genetic information need to know what they are letting
themselves in for. They need to consider how
applicable genetics is to their current research
strategy. They need to spell out the level of risk they
are prepared to take on, and then plan how to manage
that risk. In short, they need a genetics strategy.
In the case of disease genetics, risk management
would best begin by contemplating the sheer magnitude
of the undertaking. Companies will be
prompted to ask themselves questions such as these:
How feasible is it for us to establish an extensive
disease genetics program in-house? On which diseases
should our program focus? Are there opportunities
to share the risk, perhaps by joining a “precompetitive”
industry consortium, along the lines
of the SNP Consortium? Or, should we adopt a waitand-see
stance, and then hope to license in the
fruits of others’ labor?
In the case of pharmacogenetics, risk management
begins by reevaluating the pipeline. On that basis,
companies will try to determine the drugs to which
pharmacogenetics applications would add most
value. Companies will also want to study intently the
market context and competitor landscape, thinking
through potential competitor moves and counter-
39

40
moves, along with the relative scientific feasibility
for each drug or therapeutic class. Such assessments
will need to be revised continuously, as different
drugs present themselves for consideration
and perhaps suggest different approaches, and as
the market and the regulatory environment continue
to evolve.
A genetics strategy would encompass all of these
issues, and would optimize any potential synergies
among genetics approaches. If a company decides
to implement both disease genetics and pharmacogenetics,
it will need to decide how to integrate and
harmonize them. Which diseases, for example,
might be amenable to disease genetics on the one
hand, and be likely to provide a market premium
on the other? If genetic redefinition of diseases
makes it possible to develop suites of drugs and
thereby address several smaller markets, how can
research best collaborate with marketing to maximize
the impact? The answers to these questions
will generate still more questions: Do we have the
requisite skills and capabilities to pursue the strategy?
Do we have the right alliances in place, or the
right alliance strategy?
Genetics is a risky endeavor. Companies cannot
avoid the risk—but they cannot lightly ignore the
potential jackpot either. They need to be selective
and smart in deciding how and where to place their
bets. With such vast winnings at stake, it seems
appropriate that the odds should be fairly long.

Chapter 3: Managerial Challenges
Preface: Looking Back and Looking Forward
The Story So Far
The genomics revolution is poised to sweep aside
the old economics of pharmaceutical R&D. The
biotechnology and pharmaceutical industries—and
perhaps health care delivery in general—are on the
brink of transformation, and companies that
embrace the revolution in the right way stand to
reap enormous benefits. Developing a new drug
should become considerably less unpredictable and
much less expensive. Companies will record
improvements both in efficiency and in success
rates all along the value chain, and the average cost
and time needed to bring a new drug to market will
fall correspondingly. (See sidebar, “Potential
Savings—From Theoretical to Practical.”)
But this benign prospect is clouded by some warnings:
great rewards will require comparably great
efforts; a new paradigm in R&D economics may
necessitate paradigm shifts in R&D management;
above all, the great promise is offset by great risks—
though, as in any revolution, the risks of standing
aside may be greater than those of getting involved.
Ensuring Your Future
All biopharmaceutical companies are, or should be,
actively deciding how best to engage in the revolution.
Making such decisions is no easy matter. The
familiar bearings are no longer there, since the
competitive and regulatory landscapes have
changed so much—and continue to change—in response
to the promise that genomics offers. Compa-
nies have been rushing to claim intellectual property
rights (in the so-called IP land grab), now that
the sequencing of the human genome has been
completed. Statutes and court decisions regulating
those IP rights keep emerging and modifying the
picture. (See sidebar, “Intellectual Property—Lost
and Found.”) And the corporate map is being redrawn:
the major mergers of recent years have created
industry superpowers, and the pace of acquisitions
and alliances is set to quicken, if anything.
(See sidebar, “Industry Changes.”)
With so much change occurring, there are bound
to be winners and losers. Although the decisions
will be unfamiliar and difficult, success will in the
end be determined by traditional criteria. The winners
will be those who make optimal strategic
choices and then implement them in an optimal
way. The two components of the winning combination
will differ from company to company, according
to each company’s size, aspirations, financial
power, capabilities, and so on. In this final chapter
of our report, we identify the strategic and operational
issues and examine the various options that
different companies might exercise.
To begin with the strategic issues, then—specifically,
the challenge of defining a strategy in the
genomics era.
Strategy—Searching for Genomic
Competitive Advantage
Before genomics, biopharmaceutical companies
used two basic tools—chemistry and molecular biol-
41

42
POTENTIAL SAVINGS—FROM THEORETICAL TO PRACTICAL
In the first two chapters of this report, we assessed
the potential savings for the two waves of the
genomics revolution: first, the substantial savings
attainable through genomics technologies; second, the
greater but far less certain savings attainable through
genetics approaches, notably disease genetics and
pharmacogenetics. Those assessments show the high
end of the achievable range, and they view the two
waves singly rather than jointly; that is, they indicate
discrete and best-case scenarios, which will be difficult
for companies to realize in practice and impossible
to combine.
A more integrated assessment needs to average out
the achievable range—to take account of worst-case
scenarios too. And it has to analyze the various likely
combinations of approaches from the two waves,
rather than treating genomics and genetics
approaches in isolation.
According to the combination selected, the R&D
value chain as a whole will assume a particular new
form and favor a particular subset of potential drugs.
That is a crucial consideration for a company engaged
in building a portfolio of technologies: the more com-
DRUG R&D VALUE CHAIN ACTIVITIES
Biology
Target ID Target Optimization Validation
Genomic target discovery
Chemical genomics target discovery
Genetic target discovery with pathway analysis
Chemistry
Screening Optimization
In silico
Traditional
binations of approaches, the greater the company’s
versatility in pursuing different drug subsets.
Drawing once again on our economic model, we
have examined the impact of each feasible
genomics-based approach to target discovery, augmented
by pharmacogenetics and genomics technology
whenever possible. And we have estimated the
realizable value in each case: first, by calculating the
cost, time, and added value for each possible combination
of approaches, and then—adjusting for the
percentage of targets each approach is able to
process—by calculating a weighted average per
drug. The result is three broad scenarios:
• A genomics-based approach: industrialized target
identification, supplemented where applicable by
downstream genomics technologies (in silico
chemistry, in silico toxicology, in vitro ADME, surrogate
markers, and pharmacogenetics)
• Chemical genomics: industrialized target identification,
followed by chemistry and traditional validation
conducted in parallel, and supplemented
where possible by the downstream technologies
just listed
30%
70%
Development
Preclinical Clinical
Genomics:
In silico/in vitro tests, surrogate markers
Pharmacogenetics
Genomics and pharmacogenetics
Traditional
73%

• Disease genetics supported by pathway analysis,
combined where possible with the various downstream
technologies
The chart on page 42 shows the three scenarios,
and the weighting applied to calculate the average
cost per drug. Here, by way of example, are realistic
savings estimates generated by our economic model
for these three approaches.
First, a genomics-based approach in a traditional
value chain structure. This provides an ideal foundation
for a portfolio perspective on drug discovery. It
applies to both known and unknown target classes,
and offers impressive potential savings of $200 million
and 1.5 years per drug.
Next, chemical genomics is probably the most competitive
approach for targets where there is an established
high-throughput chemical screening assay.
The time savings are particularly important—nearly
3.5 years—boosting drug revenues by means of
intellectual property rights and first-to-market
advantages. When combined with other genomics
approaches, chemical genomics offers potential savings
of about $100 million per drug. The approach
lends itself particularly well to certain targets, such
as GPCRs, so a company electing not to pursue
chemical genomics would be at a disadvantage if it
retained such targets on its wish list.
Finally, disease genetics supported by pathway
analysis is, theoretically, the most direct route to
understanding human disease. This approach is
applicable to known and unknown target classes and
to targets overlooked in animal-based studies. And
on the face of it, it is the most attractive approach
financially, with cost savings of $400 million. There
is an extended time to drug, however, amounting to
about one year, though that drawback would often
be offset by the early securing of intellectual property
rights. Unfortunately, this inviting approach is
still not affordable, and in fact its scientific feasibility
remains unproven. (See the graphs below for a
summary of expected savings.)
REALIZABLE RESULTS BASED ON DISCOVERY APPROACH
Weighted average of approaches across value chain
Cost to drug
Pre-genomics
Genomics-based
approach
Chemical genomicsbased
approach
Genetics-based
approach 1
Time to drug
Pre-genomics
Genomics-based
approach
Chemical genomicsbased
approach
Genetics-based
approach 1
0
0
200
5
400
10
475
600
11.3
675
800
13.3
15
780
14.7
880
Cost ($M)
15.8
1,000
20
Time (years)
1With pathway analysis, 50% through genetic simple systems, 50% through
genomics expression profiling. Assumes resolution of scientific and technological
questions.
43

44
ogy—to discover new drugs. Broadly speaking, the
drugs that emerged were much indebted to
serendipity. Research strategy consisted mainly of
choosing which therapeutic areas to investigate, and
discovery efforts focused on individual drug targets.
Development provided even fewer strategic choices:
a promising compound emerging from chemistry
would be tested on animals and humans in large
and inefficient trials (inefficient because there was
no means of identifying in advance likely responders
or nonresponders). With the rise of genomics,
there have come new technologies, new approaches,
new information, and new ways of thinking about
INTELLECTUAL PROPERTY—LOST AND FOUND
Gene patent applications are flourishing: in 2000
alone, more than 20,000 were submitted to the
United States Patent and Trademark Office. Despite
the large number of applications, two central questions
have yet to be answered. What exactly can be
patented? And what rights does a patent actually
confer on its holder?
For a gene or gene fragment (an expressed sequence
tag, or EST) to secure a patent, its “utility” has to be
established. In January 2001, the United States
Patent and Trademark Office issued Utility Examination
Guidelines to clarify the standard used. (That
in itself was encouraging to those in favor of gene
patenting, reinforcing the view that genetic material
can indeed be patented.) Following on the interim
guidelines released in 1999, the new guidelines
advise patent seekers to provide at least one “specific,
substantial, and credible” use for the gene or
gene fragment in question. This restatement effectively
fixes the height of the hurdle for applicants
and disqualifies undersubstantiated applications
from the outset. Some uncertainty remains, however:
whether it is necessary when presenting evidence
of utility to understand the actual biological
function of the genetic material, and whether gene
research and development. These have brought
with them a new opportunity, or imperative, to turn
research to competitive advantage.
So companies now have weighty strategic issues to
address. At the corporate level, the question is how
much to invest, given the current environment. For
R&D leadership, the question tends to be where to
focus those investments—in what therapy areas, on
what target classes, and so on—as well as which
technologies to adopt and how to adopt them (inhouse
or externally, for example), and how to mitigate
the associated risks.
fragments, as distinct from full-length genes, are eligible
for a patent.
If getting a patent approved seems daunting, all the
more so is enforcing it, or sheltering confidently in
its protective embrace. The strength of protection
afforded by a gene patent is still a developing legal
issue. One recent court decision, though, clearly
marks a setback for patent holders—Festo Corp. v
Shoketsu Kinzoku Kogyo Kabushiki, decided by the
federal circuit court in November 2000. It appears
to weaken many patents by precluding a broad interpretation
of most patent claims; it does so by virtually
excluding the “doctrine of equivalents.” This is a
doctrine that generally allows extension of a patent’s
claim beyond its literal language, so that a would-be
infringer who makes trivial changes to the patented
product is not thereby exempt from the patent’s constraints.
According to the Festo ruling, if the language
of the legal claim diverges from that of the
patent itself, the doctrine no longer applies. Many
patents now look very narrow and vulnerable, and
companies will have to plan their patent submissions
even more carefully in the future. Or hope for
a change of fortune: the U.S. Supreme Court is
expected to review Festo in the 2001–2002 term.

The Starting Position
Although these same broad questions will apply
equally to all companies, there can be no standard
answers. The actual options available to any company
will depend on its starting position.
Company Size
A key constraint on a company’s strategic options is
size. The largest pharmaceutical companies boast
capabilities and finances on a scale that allows full
participation in the new technologies, even when
the risk is high. Not that this exempts them from
having to make choices. In fact, since scale gives
them so many options, they arguably carry a greater
burden of strategic decision making. How to select
from such an embarrassment of riches? In addition,
they face the challenge of managing complexity. If
they are not selective enough, and embrace too
many options, the operational problems could
prove overwhelming.
The narrower capabilities and lesser scale of smallto-medium-sized
pharmaceutical companies and the
larger biotech companies could represent either a
severe drawback or a distinct advantage. On the one
hand, there are reduced opportunities and even the
prospect of being locked out by the big pharmaceutical
firms: with disease genetics, for instance, a company
with insufficient scale to build an in-house
capability would risk forfeiting potentially lucrative
intellectual property rights. On the other hand,
since lesser scale often means lesser complexity,
these modest-sized companies can compete more
flexibly, changing their tactics quickly in response to
technological advances or competitor moves.
To see how scale can affect a company’s options,
consider the differing ways in which large and midsize
companies approach the target land grab. The
larger companies have been able to take very
aggressive approaches—scaling up or pursuing big
deals to secure intellectual property rights to targets.
The smaller companies, lacking in resources,
have been unable to follow suit, but some of them
have compensated by choosing very focused strategies,
concentrating on their special competencies
and imposing higher quality standards.
Building the Fact Base
Apart from company size, the two most important
facets of a company’s starting point are the beliefs
and hypotheses held by its leadership team (roughly,
its corporate culture) and its current R&D capabilities.
Companies need to scrutinize both.
It is crucial to understand and shape the beliefs and
hypotheses of leaders throughout the organization,
especially since, with genomics and genetics, the
contributions and effects are cross-functional—that
is, the managers or sections that contribute most
are not necessarily those that benefit most. All
those affected need to articulate their perceptions
of the value and applicability of genomics and
genetics to the company. Once tested, these perceptions
should be given considerable weight when
it comes to defining company strategy.
An equally thorough assessment needs to be made
of the company’s relevant R&D capabilities—its
technologies, skills, specific knowledge of diseases
and disease mechanisms, and so on. Ideally, this will
include an audit of current R&D productivity at
every step in the value chain, identifying bottlenecks
and other constraints. The more accurate
and detailed the assessment, the more effectively
the company can address the strategic questions as
they pertain to its specific situation.
Corporate Decisions:
How Much to Invest and Where
As suggested above, even the largest pharmaceutical
companies will have to make choices. Consider
some of the huge deals of recent times: the $500
million deal between Bayer and Millennium for targets,
the $800 million deal between Novartis and
Vertex for in silico chemistry, the $500 million deal
between Roche and deCODE for disease genes.
Note that these deals concern discrete steps of the
value chain: in each case, it appears likely that the
companies concerned were acting on an explicit
preference—an established strategic preference.
After all, given the magnitude of these deals, it
seems unlikely that any one company would have
placed all three bets. More to the point, such large
deals, although essentially R&D ventures, are not
45

46
INDUSTRY CHANGES
The genomics landscape features many small startups
amid the larger genomics companies and the
genomics divisions of big pharmaceutical corporations.
But that landscape is changing. The number
of deals—of genomics companies combining with
each other or being taken over by big pharmaceutical
firms—has been growing steadily. What is driving
this tendency toward consolidation?
The Pressure to Extend Scope
Increasingly, genomics companies are aspiring to
become full-fledged drug companies. No specialized
company, it seems, has yet succeeded in building a
truly stable competitive position as a drug-industry
supplier; intellectual-property statutes do not
appear to be enough to guarantee long-term protection;
and the chances of proprietary advantage are
being nullified by the trend toward public-private
partnerships or consortia, underwritten by big pharmaceutical
companies.
Wall Street appears to place a far higher value on
integrated drug producers than on pure technology
companies (if only because the drug sector has traditionally
enjoyed such high profits and such high
regard among investors). According to a recent USB
Warburg study, the average integrated drug company
has been able to raise $870 million, as against
a mere $330 million for the average technology
company. (The study noted a further interesting
divergence among technology companies themselves:
biology companies—those focused on target
identification and validation—raised $480 million
on average, whereas companies in the chemistry
area—those focused on screening and lead optimization—raised
on average only $170 million.)
In keeping with this expansionist aspiration, most of
the recent deals have consisted of acquisitions of
downstream drug-development capabilities. Witness
LION’s acquisition of Trega (for $35 million),
Celera’s acquisition of AxyS (for $173 million), and
Lexicon’s acquisition of Coelacanth (for $32 million).
The Pressure to Achieve Scale
As sections of the value chain have become industrialized,
the value of scale in R&D has gained
prominence. And for genomics platform companies
and pharmaceutical companies alike, it may appear
quicker and neater to achieve scale through a merger
than through painstaking in-house upscaling. (Of
course, pharmaceutical companies might have other
reasons to acquire genomics companies: to jumpstart
their genomics efforts, for instance, or to
acquire otherwise rare capabilities.)
Sure enough, most of the recent mergers and acquisitions
have clearly been initiated for the sake of
increasing scale: Sequenom’s acquisition of Gemini
Genomics (for $238 million), for example, or
Sangamo’s acquisition of Gendaq (for $40 million).
These scale deals have been primarily in target identification
and validation rather than in chemistry—a
reflection of the urgency of the land grab.
The Pressure to Spend
Whatever the inducements to merge, there is a traditional
impediment—lack of wherewithal. The spirit
is willing but the purse is weak. That is certainly not
a constraint, however, on some of the large
genomics companies at the moment. In mid-2001,
three top companies—Human Genome Sciences,
Celera, and Millennium—boasted over $4 billion in
cash between them, representing about 25 percent
of their combined market capitalizations. Idle money
cries out to be spent—probably, for these companies,
on diversification more than on scaling.
Market expectations, a sense of urgency, an abundance
of funds: all signs pointing to continued consolidation
in the near term.

R&D decisions alone. Almost certainly, the decisions
were thrashed out at the corporate level.
For more modest-sized companies, strategic choices
often go beyond matters of preference or emphasis.
The question might be whether to concentrate all
their efforts on some value chain steps and forgo
others altogether. Certainly it no longer makes
sense for even midsized pharmaceutical companies
to compete in target identification. And at the
smaller end of the scale, companies with less than
$400 million in R&D, say, may find themselves asking
even more radical questions: Can we afford
research at all? Should we not focus exclusively on
licensing instead? Again, it is at the corporate level,
rather than within R&D alone, that such questions
will eventually be settled.
It is not just through major partnerships and
investment decisions, however, that the corporate
level is impinging on R&D strategy. More and more,
specific R&D activities are having ramifications
beyond R&D itself, and invoking corporate-level
participation. Pharmacogenetics, for instance,
often touches on corporate strategy as much as on
R&D strategy. Should the company continue to pursue
a promising compound, say, when the risk of
market fragmentation might outweigh the positive
market effects? Should the company attempt to resurrect
candidate drugs previously killed because of
rare side effects? And so on.
R&D Leadership Decisions:
Where and How to Compete
With genomics and genetics now part of the landscape,
R&D decision making has become more
complex. The options are far more numerous:
there are more ways of gaining access to capabilities,
more technologies to choose among, and even
new dimensions in which to compete. R&D executives
must select a combination of options that not
only dovetail with the company’s starting position
and aspirations but can also be integrated smoothly
with one another.
Choosing a Research Focus
The dimensions of competition include:
• Disease states. Some disease states have become
more tractable, thanks to genomics approaches,
and any company continuing to investigate them
will have to deploy genomics if it is to remain
competitive. Just which therapeutic areas or disease
states are most amenable to genomics is
determined by several factors: the degree to
which the disease is genetic in nature, the current
understanding of disease processes at a molecular
or genetic level, and so on.
• Target class. Some genomics approaches are at
odds with traditional therapeutic-area borders,
and favor a broader deployment—around target
class—rather than the old focus on disease state.
(The targets within a class are usually similar in
structure and biochemical function.)
• Therapeutic modalities. Small-molecule drugs
still dominate the market, but they no longer
monopolize it. Some new therapeutic modalities
have already established a foothold—injectible
protein therapeutics, for instance, based on
secreted factors and antibodies. Others remain
very much in the experimental stage—gene therapy
and anti-sense techology, for example—though
adventurous companies are pursuing them
undaunted (as exemplified by Lilly’s recent $200
million deal with Isis to gain access to anti-sense
capabilities).
These dimensions are interconnected, of course,
and even interdependent. Take Novartis’s interest
in oncology, for example—a broad disease state.
Given that interest, it made sense for the company
to focus on kinases, a key target class in oncology.
Kinases constitute one of the few target classes that
are amenable to a particular genomics approach, in
silico drug design. Novartis has duly set about augmenting
its expertise with the appropriate
genomics technology, forming an alliance with
Vertex to that end.
47

48
Selecting Technologies
According to the research focus adopted by the
company, certain technologies will press their
claims immediately. An oncology program, for
instance, would certainly argue for the incorporation
of a transcription profiling approach, as more
and more cancers are being redefined at the level
of RNA expression. But each claim would have to be
assessed by reference to the company’s aspirations
and current capabilities. How comfortably would a
candidate technology fit in with the company’s risk
profile or existing skills mix, for example?
In addition, companies will need to consider the
current stage of development of genomics technologies.
When is the best time to buy into the
favored technology? As noted throughout this
report, although some genomics approaches are
practicable today—in the early steps of the value
chain, notably—others remain speculative:
genome-wide association studies, for instance. A
company’s risk profile will determine whether it
wishes to be on the “bleeding edge” or to be a technology
follower. Either way, the company will want
to chart the evolution of genomics technologies
and approaches, and adjust its own strategy accordingly.
A technology scouting function is indispensable,
now more than ever.
Whether the technology is proven or unproven,
companies will need to decide not just whether and
when to invest, but also how—how to keep a sharp
focus and mitigate the risks involved. The options
vary from company to company, again according to
company size. With disease genetics, say, a large
pharmaceutical company that chose to pursue the
technology in-house would face the question of how
to apply it—to which therapeutic areas, for example.
A smaller company, by contrast, unable to build
a program in-house, and obliged to take a different
approach, would face such questions as what kind
of joint ventures to pursue and what focus to apply.
Deciding How to Acquire or
Gain Access to Capabilities
In general, there are several ways to attain a desired
capability, but in some cases the options are limited.
When the item is a proprietary database or tool, for
instance, the company will have to license it in (or
pay a provider for service) rather than buy it outright;
or when a company views its own information
as too confidential to outsource, it will be forced to
implement the related technology in-house. In
many cases, though, a company will face the choice
between building in-house capabilities and outsourcing.
The in-house option, to justify itself,
would have to confer some significant strategic or
cost advantage. A company could have a cost advantage
if it had developed a proprietary method, for
example, or if it could boast greater scale or experience
in a given approach.
Some though not all of the new technologies show
clear scale benefits, thanks to industrialized processes
and informatics. (Among the most obliging
technologies in this regard are expression profiling,
traditional HTS and µHTS, and exploitation of
informatics-based analysis. The least obliging are
medicinal chemistry and animal models, and somewhere
in between are compound synthesis and
management, proteomic expression analysis, structural
biology, and in silico chemistry.) Unfortunately,
building scale in-house could be disproportionately
costly for small-to-midsized pharmaceutical
companies, even for the most scale-friendly
technologies. These companies are unlikely to realize
cost advantages; they risk spreading their technology
dollars too thin. The wiser option would be
partnering or licensing.
If a company decides to develop a given technology
in-house, it should review that decision regularly.
What is today a strategically advantageous capability
may be commoditized tomorrow. The perception of
sequencing, for instance, seems to be shifting, from
a need-to-have technology to something that can
readily be outsourced.
If a company decides to outsource a given technology,
it will have to decide further on a prospective
partner or partners. It might even opt to join forces
with competitors. A model partnership of this kind
has been the SNP Consortium. A group of pharmaceutical
companies, helped by various academic

institutions, banded together to identify 300,000
SNPs (in the end, the total was about one million)
and put them into the public domain. This joint
effort had two very beneficial effects for its participants.
First, it enabled the companies to concentrate
more on their core interest, finding drugs;
second, it forestalled the efforts of genomics companies,
which would have sought to patent and
extract rents from these SNPs. Other candidates for
“coopetition” of this kind include protein structure
modeling and broad-scale sample collection for disease
association studies.
CASE STUDY: BRINGING RESEARCH INTO FOCUS
A midsized pharmaceutical company initiated an
R&D strategy overhaul. With the broad goal of increasing
productivity, the effort was directed at reducing
the number of and concentrating the areas of
investigation and boosting the company’s access to
relevant technologies.
The CEO and R&D director commissioned a review
of the company’s research and development capabilities.
A cross-functional project team then set about
defining those capabilities precisely at all steps of
the value chain, rating their quality, aligning them
with the diseases and markets of interest, and identifying
gaps and synergies.
With specific therapeutic areas in mind, the company
next turned to optimizing its genomics technology
portfolio. The project team embarked on a threestage
assessment of the various strategic options
and their corresponding technologies.
First, having audited the company’s capabilities and
research interests, the team identified relevant
industry trends, and from these generated a list of
the strategic options. (Genetics was listed, for example,
as a major target-discovery trend in a favored
research area—a therapeutic area with many heritable
diseases.) The team then compiled a list of
Putting the Strategy into Operation
Defining a genomics strategy is a good start, but
even the most brilliant strategy is futile if it remains
defined on paper only. The point is to put it into
operation. Putting a strategy into operation consists
essentially of making changes and managing them
effectively. In the case of genomics and genetics,
the changes that need to be made are profound,
affecting all aspects of the R&D organization and,
by extension, the corporation as a whole—core
processes, organizational structure, job descrip-
matching technologies, whether currently owned or
not—those that would enhance the options’ chance
of success. (Against the genetics option, for
instance, were listed such matches as SNP maps
and genotyping technologies.) Finally, the team evaluated
each technology’s likely impact on productivity,
using a sophisticated productivity model the
company had established.
What emerged was a ranking of various strategic
options and their required technologies—in effect,
the basis for a new, integrated technology strategy
and the blueprint for an optimized technology portfolio.
Company executives were now in a position to
ponder access arrangements—whether upgrading,
licensing, or partnering.
More broadly, the result of this effort has been a dramatic
focusing and aligning of research activities,
directed by a well-articulated strategy. The R&D
managers have been able to commit to specific productivity
metrics and time frames. They have been
able to agree on a distinct research focus: three key
therapeutic areas and limited modalities. And they
now have a specific technology strategy, with a clear
plan for acquiring the necessary components and an
investment program to make it happen.
49

50
tions, interfaces, and so on. The necessary work can
be divided into three broad areas:
•Rebalancing the value chain
• Establishing the new organization and its
governance
• Managing organizational change
Rebalancing the Value Chain
The old ways of conducting R&D are often unsuited
to the new era. As the first chapter showed, the traditional
R&D value chain no longer works. For one
thing, its smooth flow quickly becomes disrupted by
a series of bottlenecks, induced by the different
productivity gains at different phases. For another,
its sequence is unsustainable, since the new technologies
dance to a different schedule: much of the
chemistry phase might now take place simultaneously
with target validation, for example. To reinstate
a smooth flow (while enjoying the new, much
accelerated rate of throughput), R&D needs to ease
the bottlenecks and adjust to a reconfigured value
chain. And that means redistributing resources
and, more importantly, redesigning processes, as
well as keeping the new value chain in balance.
Restoring Balance: Reallocation versus Redesign
At first sight, the bottleneck problem would seem
fairly simple to resolve: scale up downstream steps
to meet the increased demand. But how feasible
would that be? The number of targets identified
could increase sixfold or more. To scale up to meet
that increase, a company accustomed to spending
$1 billion on all of R&D would now have to spend
more than $1.5 billion on target validation alone.
Another simple approach suggests itself: adjust
resources along the value chain in order to bring
the uneven phases back into balance, shifting funds
from more efficient phases (notably target discovery)
to less efficient downstream phases (such as
preclinical). But such reallocation of resources is,
on its own, an overcautious measure, and will not
have a really dramatic impact on R&D economics. It
neglects, or even distracts from, the central opportunity
that genomics offers: the opportunity to
“raise the game” by changing fundamentally the
way R&D is conducted.
This transformation of R&D will derive above all
from the bold reconfiguring of processes, for the
sake of both physical process flow and information
flow. For the former, new technology platforms
need to be integrated and optimized, both within
value chain steps and across the value chain. (In
some cases, this may require a discipline and a
rearrangement comparable to the moving assembly
line introduced by Henry Ford in 1913.) As for
information flow, the tremendous amount of data
generated by the new technologies remains worthless
unless translated into functional information
and supplied punctually—that is, in time to influence
the decisions being made. (See sidebar,
“Establishing a Unified Informatics Structure.”)
The extent of the redesign, and the particular
shape the new flows take, depend very much on the
company’s strategy choices. Processes that are newly
industrialized, but that still follow a traditional
R&D sequence, need to be systematized. In some
cases, however, the traditional R&D value chain will
need to be disrupted. To integrate chemical genomics
and genetics, for instance, would necessitate a
major restructuring of the value chain. Chemical
genomics introduces a new parallelism, as target
validation and chemistry activities are conducted
simultaneously; the two processes now interact
rather than just interface. And genetics introduces
feedback loops, where late-phase findings (such as
genetic information from the clinic) feed back into
earlier steps of the value chain (such as diseasegenetics-based
target discovery).
In anticipation of any process redesign, individual
function heads should be pondering the contingencies:
how and when genomics might affect
them, and what actions to take when it does. As one
bottleneck is relieved, another is created: When will
the bottleneck reach their step in the chain, and
what will its impact be? What new technologies and

ESTABLISHING A UNIFIED INFORMATICS INFRASTRUCTURE
For a company to extract full value from the new
technologies and the copious data that will emerge
from them—that is, to transform data into knowledge—it
will have to devise a comprehensive informatics
vision and architecture. The vision will articulate
the role of informatics as a potential source of
competitive advantage. The architecture will have
three essential components
First, there must be an optimized information flow
across the newly industrialized research and development
value chain. Standards will need to be
established to ensure that data are formatted, organized,
and defined consistently. Hardware and
applications will have to be linked and networked
appropriately so that information can move where it
needs to, feeding subsequent steps in the process.
Second, a centralized knowledge management system
is required, to capture and store the data, integrate
it with external data, and make it available
throughout the company. Finally, powerful analytical
tools will be required, to mine and make sense of the
data—sophisticated algorithms, visualization tools,
and so on.
To develop and integrate these components, most
companies will have to invest heavily. The costs may
look particularly high in relation to traditional costs,
but that is partly because the industry has generally
underinvested in IT. (One large biotech company, following
its informatics upgrade, reports a threefold
increase in its annual IT budget.) To focus the investment
accurately, a coherent plan is once again
essential. A critical decision is whether to develop
the capabilities in-house, outsource to solutions
providers, or purchase and integrate informatics
packages. The choice or choices made will depend
on such factors as available internal expertise, the
amount of integration with legacy information sys-
tems required, and the availability of reliable integration
vendors, package suppliers, or solutions
providers.
That last factor may prove particularly difficult to
assess. Who would provide the most reliable and
suitable assistance? Today, no single provider of
application software provides all the functionality
needed. The informatics industry is crowded with
small start-ups offering niche products; the cumulative
market capitalization of all publicly listed bioinformatics
companies scarcely amounts to one-eighth
of GlaxoSmithKline’s annual R&D budget. And while
larger IT solutions providers are gearing up to serve
the burgeoning needs of this market, they are still in
the process of developing internal life-sciences capabilities.
Given that each prospective solutions
provider is likely to try to make its offering the centerpiece
of the company's informatics architecture,
and given that multiple solutions will need to be
knitted together, companies would be wise to solicit
independent, unbiased advice before deciding on
particular vendors.
In addition to managing all this informatics complexity,
companies will have to deal with a further
challenge if they are to implement the new information
regime successfully. They will have to find a way
to resolve the human-resources and organizational
issues that are bound to arise. Talented, experienced
informatics personnel are difficult to come by: how
to find and keep the right people and how to fit them
into the organizational structure are questions that
companies will need to address more actively and
imaginatively than ever. So too the question of how
to change processes and behaviors generally,
throughout the organization, to ensure fullest use of
the new informatics tools—a question examined in
some detail elsewhere in this chapter.
51

52
CASE STUDY: REDISCOVERING DISCOVERY RESEARCH
In a global pharmaceutical company, the discovery
division was slowly undergoing a change of character,
from a pre-genomics one of small, independent
efforts to an up-to-date one of highly defined and
sequenced processes. The company was anxious to
speed up and optimize this inevitable transition. It
had recently created several centers of excellence in
research, each containing a particular combination
of technologies, resources, and expertise, but these
new groupings remained in need of improved
process flow, both within and between them.
An “industrializing” approach was proposed: why
not treat each center of excellence as a factory, and
in that way rethink or refashion the discovery program
as a whole?
A factory-based structure imposes a strict discipline.
Typically, factories have clear, measurable objectives,
with defined inputs and outputs, and specified
resources and roles. Efficiency is closely regulated:
internal processes and interfaces are optimized for
approaches will be available, and how effectively
will they relieve the bottleneck? The head of development,
for example, should already be contemplating
the inevitable increase in demand for clinical
trial capacity and weighing the various options,
such as pharmacogenetics, for meeting it.
Retaining Balance:
Capacity Planning and Management (CPM)
After the initial jolt of genomics, supply and demand
should get back into alignment, thanks to the combined
forces of resource reallocation and process
redesign. But this restored balance is a precarious
one, and needs careful and regular maintenance.
That is where capacity planning and management,
or CPM, can play an invaluable role. By enabling an
organization to keep supply and demand aligned,
CPM also enables it to make rational plans, linked to
capacities and resources, and thereby to manage
projects with optimal efficiency.
scale, quality, and productivity; external interactions
are monitored regularly for compatibility and cost
effectiveness.
In keeping with this ethos, the company set about
defining processes within each potential factory as
clearly as possible. The project team set targets for
inputs, outputs, and quality standards; it identified
activities that could be completed inside the factory,
as distinct from those supplied as support from outside;
and it itemized links between factories themselves,
between factory and nonfactory research,
and between research units and units outside research
and development.
The result has been a subtly redesigned discovery
division. Processes and functions are now clearly
assigned to specific factories, expectations and
achievements are more transparent than before, and
the interactions throughout discovery research are
now easily tracked from factory to factory, with the
map being constantly refined.
Though well established in high-profile corporations
such as General Electric, Hewlett-Packard,
and Cisco, CPM is conspicuously rare in biotech
and pharmaceutical companies. For the genomics
revolution to realize anything like its full productivity
potential, efficient CPM will be immensely beneficial
if not imperative.
Establishing the New Organization and its
Governance
Implementing the process changes just mentioned
will entail a thorough review of a company’s existing
hierarchies and procedures. For the process
changes to yield optimal value, changes also need
to be made in traditional decision-making methods
and in organizational structures.
New Linkages and Interfaces
To begin with organizational changes. With the
value chain so much altered in appearance, and

processes now so different from before, many
disparities and stresses will inevitably develop in
an unaltered managerial system. The old structures
will creak and strain under the unfamiliar
new pressures. To restore congruence, a company
may need to undertake some bold organizational
reshaping—shifting or removing divisional
borders, reassigning personnel, redistributing
areas of responsibility, and so on—not just within
R&D, but also within the company as a whole and
even beyond, in the alliances the company might
enter into.
The R&D Department. Incorporating the requisite
new capabilities, it goes without saying, represents a
formidable organizational challenge: not only do
they have to mesh with existing capabilities, they
need to coordinate with one another as well. To
implement in silico drug design, for example, it
would be almost essential to provide an informatics
interface between structural biology and chemistry
data. Meanwhile, a comparable reorganization of
personnel has to be undertaken. Biologists and
chemists, for example, can no longer proceed in
isolation, but must now work alongside each
CASE STUDY: MANAGING CAPACITY—EMPOWERMENT THROUGH CPM
A large pharmaceutical company was facing a capacity
crisis. Both development and staffing levels were
under pressure, mainly as a result of productivity
improvements in basic research and competition for
scientific talent. As a key part of the remedy, the company
undertook worldwide implementation of capacity
planning and management—a considerable challenge
for such a complex organization, where demand
was uncertain and resources were not fungible.
Development of CPM had four main components:
• Quantifying capacity and demand. Appropriate
units of capacity and demand were defined for
each function. In clinical departments, for
instance, the typical unit of capacity was defined
as a team of monitors, coordinators, and support
personnel, and the unit of demand was a study.
• Business processes. To exploit CPM fully and foster
cooperation among departments and project
teams, various new business processes were initiated—most
importantly, the tracking and interpretation
of demand and capacity information, and
the consequent adjustment of timelines and
resource allocation. Appropriate linkages needed
to be made to related functions such as facilities
planning and human resources.
• Change management. Since CPM tends to affect
deeply the way an organization operates—publi-
cizing the relative productivity and workload of
different departments, for example—some managers
react more negatively than others. The company
took steps, both before and during the
implementation of CPM, to ease the transition.
The message was constantly reinforced—that the
changed regimen was beneficial, essential, and
permanent.
• IT support. With CPM quickly generating a wealth
of information, some centralized and some requiring
broad dissemination, the company recognized
that its CPM initiative needed extra IT support. It
identified suitable vendors with the requisite flexibility
and pharmaceutical experience.
The CPM endeavor has been widely hailed. No
longer is the question “Do we have the capacity to
do these projects?” met with silence. Nowadays, the
CPM team can provide a detailed, graphical depiction
of capacity and demand in each department and
overall, and an analysis of the capacity impact of
each project.
Looking ahead, the company expects CPM to contribute
to enhanced revenues, by speeding time to
market and increasing the number of indications per
compound. It also expects to use CPM to improve
portfolio management, and to save costs through
more rational investment in hiring and facilities.
53

54
other—often literally—on collaborative projects or
in formal discovery partnerships. And genetics
requires far closer collaboration between basic
research and development than ever before.
One excellent example of rethinking traditional
organizational structure and boundaries is
GlaxoSmithKline. Alert to the impact of scale, the
company has on the one hand consolidated functions
where scale and coordination provide a clear
advantage, and on the other, engaged in decentralizing
where size and complexity could prove a drawback.
Specifically, prompted by the scale benefits,
the company decided to organize centrally both the
front end and the back end of R&D (that is, target
discovery and full development). For the steps in
between, conversely, where the company’s enormous
scale would risk encumbering innovation, it
has established smaller, more autonomous centers
of excellence (based on different therapeutic
areas), which attempt to simulate the feel of smaller
biotech companies.
The Entire Corporation. So, enhanced control of data
and increased cross-functionality of personnel are
set to change the structure and tone of the R&D
department. But their sphere of operation is
broader than that. As with the strategic issues discussed
earlier, the company as a whole is implicated.
New lines of communication, and possibly
new chains of command, will need to be extended
between R&D and other units. In particular, the
relationship between R&D and marketing will be
fundamentally transformed: with R&D facing
greater choice and placing bigger bets earlier than
ever, commercial input will be crucial. And pharmacogenetics
will require new ways of thinking
about markets, competitors, and customers. (Pharmacogenetics
may also inspire new linkages
between pharmaceutical and diagnostic units for
corporations that have both).
Coordinating the commercialization process
between R&D and marketing has always been a delicate
balancing act. Most biopharmaceutical companies
have established product development project
teams to drive the process. These cross-
functional teams are charged with developing product
strategy and coordinating the various functions
as products progress from R&D into the market.
The job has now become even more complex and
tricky, owing to larger global efforts, greater information
flow, more specialized functions, and
increased liaison with global strategic marketing
(especially when companies consider the options
for applying pharmacogenetics to molecules in
development).
Beyond the Corporation. Finally, new partnership
models need to be considered. Although traditionally
organized partnerships are still appropriate in
many cases, new and more flexible forms of alliance
will sometimes be required, notably when it comes
to collaborating with academic or not-for-profit
institutions and to joining horizontal networks or
consortia.
R&D Governance
One potential source of gain in R&D is improved
decision making. Consider again the example at the
start of the value chain—the glut of identified targets
and the need to decide which ones should proceed
to the next phase. Genomics technologies
have created this quandary, but they have also provided
the means for solving it. Using new genomic
methods of “aptitude-testing,” decision makers can
confidently preselect the most promising targets
and forward them downstream.
Even decisions unrelated to genomics technologies
stand to improve, since the new genomics regimen
fosters a culture of rigorous selection criteria. In
fact, one of the most important, though perhaps
least noted, benefits of genomics is the way it
encourages a thorough rethinking of decisionmaking
processes. New kinds of data now present
themselves for interpretation, and they enter the
calculations earlier and in greater abundance than
the old kinds did. And R&D decision makers have
to take into account a new set of factors too, beyond
the confines of R&D, in order to maximize
value—factors such as marketing and IP implications,
for example.

CASE STUDY: REDESIGNING R&D GOVERNANCE
A large drug company recently completed an intensive
three-month project to redesign discovery governance
and is already reaping the benefits. The
company had always placed a high value on the
quality of its scientists and their entrepreneurial
drive. Now, however, it was growing increasingly dissatisfied
with its existing system of allocating
resources: the decision-making procedures were
proving very troublesome to navigate, decision makers
were difficult to identify, communication was
poor, and the decisions themselves often seemed
politically motivated rather than guided by scientific
and commercial promise.
In redesigning the decision-making procedures, the
company began with a thorough review of its current
governance process, both as espoused and as practiced
(the two were remarkably distinct in certain
instances). Various root causes of undesirable outcomes
were identified: these included perverse
incentives (that is, incentives encouraging behavior
Managing Organizational Change
With the advent of genomics, R&D personnel suddenly
find themselves in alien territory. As the scientific
methods change, the old instinctive
approaches and behaviors need to change as well.
Among the greatest challenges facing R&D executives
is managing the human side of change.
How the Scientist’s Job is Changing
R&D science is shifting from an arena of experimentation
to one increasingly concerned with theoretical
biology. The challenge is now less how to
get the data than what to do with the data collected.
Scientists who formerly could do their jobs virtually
on their own—conduct their own experiments, and
generate and analyze the data themselves—now
find they need to collaborate with others who have
more specialized technological skills, in areas such
as informatics, robotics, or microfabrication.
Indeed, the scientists of the pre-genomics era are
destined to evolve into two kinds of successors:
at odds with company strategy); unclear criteria,
which project champions were disinclined to clarify,
let alone follow; and inadequate allocation of decision
rights (that is, too vague a definition of who was
entitled to make which decisions), which often
meant that no decision was made at all.
From the lessons learned, a new governance process
was devised. Not just devised, but activated: by
modifying incentives, the company ensured that
practice was now properly aligned with espousal.
The new process is working well: R&D managers
navigate it easily, and decisions are being made and
communicated clearly and consistently. It allows scientists
more time to focus on their projects, and it
gives those projects appropriate funding and management
involvement. It has accordingly won the
confidence of those affected by it, and can claim a
considerable contribution to the marked improvement
in productivity that has followed its adoption.
those who interpret the data and devise plans for
exploiting it, and those who continue to develop
and optimize the technologies required for generating
the data. (Companies should be sure to recognize
and reward the latter group for its contributions,
and not relegate it to second-class status.)
All scientists will need to become comfortable with
new ways of working together—more sharing or collectivist
now, less conducive to solitary initiative.
The scientists of the future will still take responsibility
for their own work, but perhaps will no longer
take the credit for it: that will be ascribed to team
effort.
Managing the Transition
Changing from bench-based to information-based
work in this way, and from favoring fairly independent
endeavors to promoting a more collaborative
ethos, is bound to be awkward or even painful for
most of those involved, scientists and managers alike.
55

56
The formidable operational and organizational
changes will entail cultural changes too: in fact, the
new processes and structures may prove far less difficult
to establish than new habits and attitudes.
Consider informatics. It is not enough simply to
introduce powerful new IT tools within traditional
silos—within chemistry, for example, where in silico
approaches would boost the efficiency of screening
and optimization. To achieve their full impact,
these IT tools need to be deployed across functions:
to bring biologists and chemists together, to incorporate
data from the clinic into discovery, and so
on. And that will require not just new software, or
even new managerial positions, but new ways of
thinking and of relating to colleagues.
Some idea of what lies in store can be gleaned from
the history of another transformational technology—CAD/CAM
for airplane design. Like
genomics, it promised to transform a costly and
labor-intensive R&D process into a highly automated
and efficient one. After languishing in niche
applications in the 1970s and ’80s, it finally proved
its worth in the 1990s, when Boeing used it in
designing the first “paperless” airplane, the Boeing
777. To exploit the technology fully, the company
had to break down departmental barriers and
encourage collaboration across the full range of
functions. Jobs and job responsibilities had to
change. Cherished traditions were called into question.
The company held quarterly meetings at
which employees could ask questions and voice
their concerns. The transformation was a struggle,
but ultimately a great success: Boeing continues to
push the envelope in “in silico” airplane design.
When pharmaceutical companies convert to genomics,
they will have to temper the discomforts of
transition in their turn. And that means engaging
the emotional and behavioral issues—the human
issues—as deeply as the operational ones. 4 Attentive
management of the human issues, which has played
such a prominent role in so many industries in the
throes of reform, is going to be particularly crucial
when it comes to the massive institutional changes
demanded by the genomics revolution.
A Final Word
To stake a claim in the changing biopharmaceutical
landscape, let alone feature prominently within it, a
company will have to make itself radically amenable
to change. Defining a strategy is certainly a step in
that direction, and initiating that strategy is certainly
a gesture of commitment. But wholehearted
commitment is evidenced not by initiating the strategy
but rather by maintaining it—that is, monitoring
the new structures and procedures constantly,
responding to shifts in external and internal circumstances,
and introducing further changes
repeatedly, aggressive or defensive, as new opportunities
or new challenges arise, though always in line
with the controlling wisdom of the strategy itself.
If the unfamiliar outer landscape provokes feelings
of unease, so too will a company’s inner landscape,
once all the requisite operational and organizational
changes are in place. In particular, the increase
in cross-functional activity may be disorienting
for some executives of the old school. Many of
the ancient landmarks, tidy borders, and familiar
categories will no longer be there to give them their
bearings. Short of attempting a counterrevolution
or withdrawing into obscurity, they will need to
familiarize themselves with the new terrain fairly
promptly—and accept it affirmatively, not grudgingly.
Changes in attitude will perhaps prove the
most difficult changes of all to bring about, and a
company’s prosperity could be in jeopardy if they
fail to take effect.
4. For a fuller discussion of the emotional aspects of change, read The Change Monster, by Jeanie Daniel Duck, published by Crown in 2001.

Conclusion
The international pharmaceutical industry is pressing
ahead in an unexpectedly difficult environment.
Drug companies face unfamiliar frustrations.
On one side, pricing policies are coming increasingly
under threat (witness the recent moves in various
U.S. states to restrict access to costlier drugs
for Medicaid patients). From the other side, the
pressure of expectation increases too, with financial
analysts continuing to count on triumphant product
launches and enormous growth. In such an
environment, corporate well-being, or even survival,
depends on boosting productivity.
It is against this background that the genomics revolution
is unfolding. In their quest for improved
productivity, companies should welcome the new
technologies and approaches. Genomics promises
prodigious benefits: it will unlock storehouses of
information about the workings of human disease,
and greatly refine—perhaps even personalize—
health care. More to the point, it promises to transform
how pharmaceutical research is conducted.
The paradigm will shift from small-scale and
serendipitous to global, industrialized, and systematic;
and from methodical and compartmentalized
to fluid and cross-functional. The impact on R&D
economics is likely to be tremendous: in the best
case, productivity could as much as double.
Looking beyond R&D, genomics and genetics also
promise to transform the way pharmaceutical companies
conduct their business in the coming years.
If genetics realizes its potential, for example, treatments
will become more sophisticated, markets may
fragment, and the shape and value of marketing
and sales organizations will change dramatically.
The entire system of health care delivery, already in
flux, will complete its metamorphosis.
The offer that genomics and genetics are holding
out is really an offer that companies cannot refuse.
Companies that fail to accept the offer adequately
will find themselves not simply uncompetitive but
possibly right out of contention. There is nowhere
to hide, and certainly no safety in inaction. To shun
the promise of pharmacogenomics out of a fear of
market fragmentation, for instance, is not to avert
the fragmentation but simply to cede the market to
one’s rivals.
Embracing the revolution appropriately will require
both boldness and finesse: managers will have to
make major strategic decisions, and to implement
them will have to radically reconfigure operations.
The decisions take careful analysis to get right, and
the operational hurdles need nimble negotiation to
surmount. It all adds up to a formidable but by no
means impossible task. And for companies that do
it well, the rewards will be handsome.
The opportunities are unprecedented. So are the
challenges. The shrewd company will be one that
remains responsive to both, as it tries to keep its
head and to prosper in these revolutionary times.
57

Methodology
The many diverse studies on drug development
have reached diverse conclusions. They have put
the current price tag at anywhere between $350 million
and $800 million per drug (all underestimates,
in our view: our own calculation is $880 million).
Not surprisingly, when it comes to the likely impact
of genomics and genetics on the economics of drug
development, opinions diverge again.
For our study, we conducted an extensive program
of discussions in an effort to compile accurate figures
for all the main activities in the R&D process,
both pre- and post-genomics. The result is a robust
bottom-up model of R&D, based on the time, cost,
and likely success rate for each step of the value
chain.
Our model goes beyond existing models of R&D in
three important ways:
• It is the product of primary research. Other estimates
have tended to build on the findings of earlier
work; our model also draws on more than 100
discussions at nearly 50 companies and academic
institutions.
• It analyzes the discovery phase more closely than
has previously been possible. Earlier models typically
assigned a conjectural figure to represent
the sunk cost of discovery, but the art of discovery
is becoming industrialized, and we have duly
been able to model its activities more scientifically.
A more detailed understanding of the economics
of discovery has resulted, and that in turn
has allowed us to more accurately quantify the
impact of genomics on R&D. (See the chart on
page 60 for technologies modeled.)
• It is activity-based and flexible. The numbers
cited in this report represent an average drug,
unless noted otherwise, but in our research we
ranged far wider than that, and assessed each step
of the value chain under a range of circumstances.
So the model allows for scenario building
and sensitivity analysis, as well as enables us to
tailor inputs to match the unique circumstances
of individual companies.
As already mentioned, all numbers cited in the text
are for an average drug. In any individual case, cost
and time will vary according to factors such as therapeutic
area and target class.
All numbers cited in the text are for a relevant
drug, that is, one to which the technology under
discussion could be applied. Various technologies
may not apply to all targets or drugs. Where specific
limitations are likely to be a significant factor, that
is pointed out.
When we discuss the “value added” to a drug, we
are referring to its net present value: the current
59

60
value of expected profits, discounted by a representative
hurdle rate, less the cost of developing the
drug. For the average drug, we assume peak annual
sales of $500 million and 11 years to patent expiration.
Also, our numbers reflect the fact that R&D
Post-genomics Pre-genomics
GENOMICS TECHNOLOGY ASSUMPTIONS
Biology
Target ID Target Validation
Target identification
• Limited numbers of genes
• Molecular biology and biochemistry
techniques
Target validation
• Cell and tissue studies
• Mouse knockouts
Target identification
• Large numbers of genes
• Industrialized techniques
(e.g., gene chip expression)
• Bioinformatics
(e.g., database searches for homologies)
Target validation
• Cell and tissue studies
• Mouse knockouts
SOURCES: BCG analysis; industry interviews; scientific literature.
dollars saved are pretax dollars; these “saved” dollars
are subject to taxation, which explains in part
why the expressed NPV calculations can be lower
than the R&D savings.
Chemistry
Screening Optimization
Screening
• Parallel synthesis for library design
• Assay development for high-throughput
screening (HTS)
• HTS
Chemical optimization
• Bench synthesis
• Parallel synthesis
Screening
• Structural biology (target structure)
• SAR profiling of library
• Assay development for LTS1 • Virtual screening and LTS1 Chemical optimization
• In silico-supported bench synthesis
• In silico early ADME/tox
1 LTS = LOw-throughput screening; generally more information-rich, but less standardized, assays that cannot be used in HTS.
Development
Pre-clinical Clinical
Preclinical (ADME/tox)
• Animal testing
Clinical
• Patient trials
Preclinical (ADME/tox)
• Animal testing
• In silico ADME/tox
• In vitro toxicology
• Surrogate markers
Clinical
• Patient trials
• Surrogate markers

The Boston Consulting Group publishes other reports
that may be of interest to senior health care executives.
Recent examples include:
The Pharmaceutical Industry into Its Second Century:
From Serendipity to Strategy
A report by The Boston Consulting Group, January 1999
Ensuring Cost-Effective Access to Innovative Pharmaceuticals:
Do Market Interventions Work?
A report by The Boston Consulting Group and Warner-Lambert,
April 1999
Patients, Physicians, and the Internet: Myth, Reality, and Implications
A report by The Boston Consulting Group, January 2001
For a complete list of BCG publications and information about
how to obtain copies, please visit our Web site at www.bcg.com.
Vital Signs: The Impact of E-Health on Patients and Physicians
A report by The Boston Consulting Group, February 2001
Vital Signs Update: The E-Health Patient Paradox
A BCG Focus by The Boston Consulting Group, May 2001
Vital Signs Update: Doctors Say E-Health Delivers
A BCG Focus by The Boston Consulting Group, September 2001
In addition, BCG’s Health Care practice publishes Opportunities for Action
in Health Care, essays on topical issues for senior executives.

Amsterdam
Athens
Atlanta
Auckland
Bangkok
Berlin
Boston
Brussels
Budapest
Buenos Aires
Chicago
Cologne
Copenhagen
Dallas
Düsseldorf
Frankfurt
Hamburg
Helsinki
BCG
www.bcg.com
Hong Kong
Istanbul
Jakarta
Kuala Lumpur
Lisbon
London
Los Angeles
Madrid
Melbourne
Mexico City
Milan
Monterrey
Moscow
Mumbai
Munich
New Delhi
New York
Oslo
Paris
San Francisco
São Paulo
Seoul
Shanghai
Singapore
Stockholm
Stuttgart
Sydney
Tokyo
Toronto
Vienna
Warsaw
Washington
Zürich