On heuristic bias in fragment-assembly methods for protein structure prediction

Abstract

We discuss the issue of heuristic bias in fragment-assembly methods for protein structure prediction. We explain the importance of this issue, which has been paid insufficient attention by evolutionary computation researchers engaging with the structural biology community. We proceed by describing preliminary data that illustrates the significant (and expectable) impact that fragment library composition has on search performance, and discuss the challenges this poses for the development of improved fragment libraries.