Fold recognition methods

Abstract

We are still missing a basic understanding of sequence/structure/function relationships in proteins. Analogy-based prediction algorithms remain the only reliable fold prediction tools. New methods, such as threading and hybrid threading/sequence fold recognition, can often recognize even the most distant homologues and, in some cases, even unrelated proteins with similar overall structures. This knowledge pushed the envelope of analogy-based function analysis to the point that the majority of newly sequenced genomes can be tentatively assigned to already characterized protein superfamilies. However, at this evolutionary distance, fold prediction is no longer equivalent to function prediction. Instead of having the same exact function, distantly related proteins might share some functional analogy that is not obvious to the casual observer. The main challenge facing the fold recognition field is to develop tools to follow the structure prediction with function prediction and analysis.