Predicting Molecular Interactions in Structural Proteomics

Introduction As the number of files in the Protein Data Bank exceeded fifty thousand (representing around ten thousand different protein domains at 95% level of sequence identity), it is becoming increasingly important to start developing the understanding of the protein function and the next level of subcellular structural organization . This, among other aspects, requires understanding of what other biological molecules or cellular structures interact with each domain, which residues are involved in this interaction (e.g. ), and what conformational changes accompany the binging. Structure-based computational approaches to these questions invariably face the issue of protein flexibility, which is further complicated by the existence of unstructured, partially structured, or conditionally structured interfaces . While the dream of predictive milli-second-scale molecular dynamics serving as a “computational microscope” persists (K.Schulten, award lecture at the ISQBP meeting in Ascona, 2008, also ) and may even be getting more tangible as the computers become faster, the ability to make reliable predictions on the basis of such trajectory is still lacking.