Blueprint for an algorithm for conformational transformation tolerant search through a database of protein structures

摘要

The range of protein domain motion which can be modeled using current computational techniques falls short of describing many physiologically relevant cases. Structural homology modeling is an obvious alternative, but it may involve a nontrivial search for non-identical structures related by a conformational transformation, which are at the same time of low homology in the primary sequence. How rich is PDB in structures related in this way? The available algorithms still fall somewhat short of speeds desirable for the large scale study. We present a pilot study to investigate the potential usefulness of an algorithm specifically designed do find pairs of structures related through a "flexible" match. The study enables us to point out that the available methods sometimes overlook interesting structural matches, and to lay out the minimal requirements the algorithm should satisfy to be functional on a structural database of currently realistic size.