Protein structure prediction provides comparable performance to crystallographic structures in docking-based virtual screening.

ABSTRACT: Structure based virtual screening has largely been limited to protein targets for which either an experimental structure is available or a strongly homologous template exists so that a high-resolution model can be constructed. The performance of state of the art protein structure predictions in virtual screening in systems where only weakly homologous templates are available is largely untested. Using the challenging DUD database of structural decoys, we show here that even using templates with only weak sequence homology (<30% sequence identity) structural models can be constructed by I-TASSER which achieve comparable enrichment rates to using the experimental bound crystal structure in the majority of the cases studied. For 65% of the targets, the I-TASSER models, which are constructed e

SUBMITTER: Du H 

PROVIDER: S-EPMC4431978 | biostudies-literature | 2015 Jan

REPOSITORIES: biostudies-literature