| Protein structure alignment provides the ability to detect homology between proteins at a greater evolutionary distance than sequence similarity. In order to identify remote homologies, the modular parts, or domains, of a protein need to be compared. Computationally, domain boundaries are generally determined either by sequence conservation or by structural compactness. Although in most cases the boundaries identified by these very different approaches are very similar, comparisons of databases of the different domain types reveal instances when disagreement occurs. We have investigated how often domain disagreement occurs and the effect of differing domain boundaries on structure alignment results. Our comparisons indicate that domain disagreement occurs on 16% of protein chains in the non-identical subset of MMDB. Although in most instances sequence domain boundaries perform no better than structure domains when used as queries in structure alignments, there are interesting cases when structure alignment performance is markedly improved. We present several examples of sequence boundaries showing significant improvement in structure alignment results.; The generation of protein test sets for computational studies such as molecule docking is a labor intensive and time consuming task. We have created a set of methods for the computational generation of protein sets based on protein structures, sequence domain information, and small molecule chemical properties. Not only are the methods robust, agreeing with data assigned by manual curation, we have been able to identify ancient contact sites in domains in the Conserved Domain Database (CDD). |
