| advertisement: compare things at compare-stuff.com! |
If, using either sequence-only or structure-based fold recognition techniques, one or more sequences of known structure are found to be related to a novel sequence under investigation it is useful to build a model of the novel protein based upon known (parent) structure(s). This is called comparative modelling. Its four main stages are: alignment of the novel sequence with the parent(s) and other homologous sequences, copying the core or framework from the parent to the model, building the non-core (loop) regions into the model, and refinement of side-chain geometry and packing. The final stages overlap with ab initio methods, described below. In general, the quality (and usefulness) of a model depends upon the relatedness of novel sequence and parent structures, and also alignment accuracy. The reader is directed to a recent review covering this large subject[Sanchez & Sali, 1997].