Normal mode analysis of macromolecular motions in a database framework: developing mode concentration as a useful classifying statistic.

摘要: We investigated protein motions using normal modes within a database framework, determining on large sample the degree to which anticipate direction of observed motion and were useful for classification. As starting point our analysis, we identified number examples flexibility from comprehensive set structural alignments proteins in PDB. Each example consisted pair that considerably different structure given their sequence similarity. On each pair, performed geometric comparisons adiabatic-mapping interpolations high-throughput pipeline, arriving at final list 3,814 putative standardized statistics each. then computed this list, linear combination best approximated motion. integrated new mode calculations Macromolecular Motions Database, through ranking interface http://molmovdb.org. Based interpolations, statistic, concentration, related mathematical concept information content, describes can be summarized by few modes. Using able determine fraction where one could actual only also concentration comparison combinations correlated it with quantities characterizing (e.g., maximum backbone displacement or mobile atoms). Finally, evaluated ability automatically classify into variety simple categories whether not they are "fragment-like"), statistics. This involved application decision trees feature selection (particular machine-learning techniques) training testing sets derived merging "list" manually classified ones.