l - People

!
!
Hinge prediction by combining sequence features
As the GOR-like method did not work well, we sought to measure the predictive power
of the various sequence features studied above. The HI scores we have reported provide
an intuitive means of weighing the relative predictive value of each sequence feature.
We show how to combine the HI scores for several features in order to make a more
powerful predictor, which we call HingeSeq. We define this predictor as follows:
" p(a j h) p(ak h) p(al h) %
HS(i) = log $
'
10$
# p(a j )p(ak )p(al ) '
&
= HIam(no)acid (i) + HIs*condary)structure (i) + HIactive)site (i)
Equation 4
For simplicity, statistical independence of the various features was assumed in creating
this definition. Here the i‘s correspond to individual amino acids in the protein
sequence. For each i, j designates one of the 20 amino acid types, k designates the
secondary structural classification, and l designates active site versus non-active site
classification.
Thus
HI am"no#acid (i) is assigned according to residue type by looking up the corresponding
value in Table 2.1. Similarly,
!
HI s"condary#structure(i)is obtained according to secondary
structure type from Table 2.3. Following Table 2.2 approximately, we assign
74