Factors Influencing the Ability of Knowledge-based Potentials to Identify Native Sequence-Structure Matches

摘要: Several types of potentials are derived from a dataset known protein structures by computing statistical relations between amino acid sequence and different descriptions the conformation. These formulate in ways backbone dihedral angle preferences, pairwise distance-dependent interactions residues, solvation effects based on accessible surface area calculations. Parameters affecting characteristics performance critically assessed monitoring recognition native fold strict screening test, where each is threaded through repertoire motifs, generated all corresponding structures. Sequence gaps not allowed, to avoid additional approximations. Results show that residue interaction computed distances average side-chain centroids perform significantly better this test than those considering inter-C alpha or beta distances. Combining structural also beneficial. The some these fact so good they recognize correct for tested proteins, including subunits be unstable absence quaternary interactions. Most strikingly, representing preferences as many 68 chains out total 74, even though consider solely local along chain, which, being same considered secondary structure prediction methods, well incapable determining full three-dimensional fold. This leads us question ability procedures screen limited act stringent potentials. We concede, however, useful fast tests, capable revealing gross shortcomings potentials, possible biases towards due, example, memory.