27.12.2012 Views

l - People

l - People

l - People

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

sRMSD has a major limitation that must be kept in mind. In the event that the holo and<br />

starting structures have low structural similarity (e.g. come from different organisms),<br />

this measure can lose meaning to varying degrees. This is clearly demonstrated for<br />

Adenylate Kinase. The holo form is taken from the bovine mitochondrial intermembrane<br />

space, while the apo form is taken from the bovine mitochondrial matrix; the two differ<br />

in sequence and structure and in particular have differ substantially in domain 1 structure.<br />

This can best be appreciated by observing the morph of 2ak3(apo) to 1ak2(holo) on<br />

molmovdb.org(morph ID: 079851-18092). For the reasons mentioned the alignment on<br />

domain 1 is very poor and the generated structure with lowest sRMSD is actually very<br />

different from the holo structure. The structure that minimizes the fitness function, on the<br />

other hand, has the correct domain orientation but has a higher sRMSD. For this protein<br />

and this protein only we do not use sRMSD as a metric of success but instead make only<br />

a qualitative evaluation.<br />

The same problem is encountered for Biotin Carboxylase, where the holo and starting<br />

structures come from different organisms, though the structural differences are not as<br />

severe. For Biotin Carboxylase we do use sRMSD as our benchmark, but again use use a<br />

qualitative evaluation to argue that the generated structure that minimizes the fitness<br />

function is a better approximation of the holo structure than the generated structure that<br />

minimizes sRMSD.<br />

Training the fitness function on Glutamine binding protein (GlnBP)<br />

253

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!