Accuracy of side-chain prediction upon near-native protein backbones generated by ab initio folding methods

摘要: The ab initio folding problem can be divided into two sequential tasks of approximately equal computational complex- ity: the generation native-like backbone folds and positioning side chains upon these backbones. prediction side-chain confor- mation in this context is challenging, because at best only near-native global fold protein known. To test effect displace- ments backbones on for generated initio, sets (I 4ACa RMS error) four small proteins were by methods. steric environment surrounding each residue was probed placing native conformation decoys, followed torsion-space optimi- zation to remove clashes a rigid back- bone. We observe that average 40% x1 angles displaced 40° or more, effec- tively setting limits accuracy side- chain modeling under conditions. Three different algorithms subsequently used conformation. three meth- ods remarkably similar: 49% 51% predicted correctly overall (33% 36% x112 angles). Interestingly, when inter-side-chain interactions disre- garded, mean increased. A consen- sus approach described, which conformations are defined based most frequently x given method set bones. find consensus modeling, de facto includes flexibility, im- proves prediction: proved 51-54% (36-42% x112). Implica- tions structure discussed. Pro- teins 33:204-217, 1998. r 1998 Wiley-Liss, Inc.