Improved side-chain prediction accuracy using an ab initio potential energy function and a very large rotamer library

摘要: Accurate prediction of the placement and comformations protein side chains given only backbone trace has a wide range uses in design, structure prediction, functional analysis. Prediction most often relied on discrete rotamer libraries so that rapid fitness side-chain rotamers can be assessed against some scoring function. Scoring functions are generally based experimental parameters from small-molecule studies or empirical determined structures. Here, we describe NCN algorithm for predicting chains. A predominantly first-principles approach was taken to develop potential energy function incorporating van der Waals electrostatics OPLS parameters, hydrogen bonding term. The knowledge used is frequency rotameric states PDB. library includes nearly 50,000 rotamers, extensive date. Although computational time tends longer than other algorithms, overall accuracy exceeds all algorithms literature when placing an accurate trace. Considering buried residues, 80% total residues tested, reaches 92% chi(1), 83% chi(1 + 2), RMS deviation 1 A. Additionally, show if information available restrict chi(1) one well, then this generate structures with average 1.0 heavy side-chains atoms corresponding 2) 85.0%.