ABSTRACT

There are three basic approaches to protein structure prediction: homology modeling, threading, and ab initio folding (although the differences among the approaches is becoming less pronounced). In homology modeling, the target sequence has a clear evolutionary relationship to another (template) protein whose structure has already been solved. As of CASP4 [26], the best way to proceed is to obtain the best sequence alignment to the template backbone, rebuild the side chains of the protein and fill in the gaps in the alignment, typically in the loops between secondary structural elements [27]. It remains to be established in CASP5 if progress has been made in improving the alignment on subsequent modeling.