Incorporating knowledge-based biases into an energy-based side-chain modeling method: application to comparative modeling of protein structure.

摘要: The performance of the self-consistent mean field theory (SCMFT) method for side-chain modeling, employing rotamer energies calculated with flexible model (FRM), is evaluated in context comparative modeling protein structure. Predictions were carried out on a test set 56 backbones varying accuracy, to allow prediction accuracy be analyzed as function backbone accuracy. A progressive decrease was observed decreased. However, even very low substantially higher than random, indicating that FRM can, part, compensate errors modeled tertiary environment. It also investigated whether introduction FRM-SCMFT knowledge-based biases, derived from backbone-dependent library, could enhance its performance. bias conformations alone did not improve probabilities improved considerably. This incorporated through two different strategies. In one (the indirect strategy), used reject unlikely rotamers priori, thus restricting by subset containing only most probable library. other direct transformed into pseudo-energies added average potential respective rotamers, thereby creating hybrid energy-based/knowledge-based energies, which prediction. For all degrees an optimal strength existed both strategies predictions more accurate pure energy-based predictions, and predictions. Hybrid knowledge-based/energy-based methods obtained compared SCWRL method, based same approximately but significantly higher.