rsRNASP1: A distance- and dihedral-dependent statistical potential for RNA 3D structure evaluation

Abstract

Knowledge-based statistical potentials have been shown to be rather important for RNA three-dimensional (3D) structure prediction and evaluation. Recently, several statistical potentials/scoring functions have been developed for RNA 3D structure evaluation, but their performances are still at an unsatisfactory level for the data sets from various 3D structure prediction methods. In this work, we developed an all-atom distance- and torsion-angle-dependent statistical potential with residue separation for RNA 3D structure evaluation, named as rsRNASP1, by considering torsion angles for the backbone, sugar ring, and base to involve local structure features. The extensive examinations against available RNA test data sets show that rsRNASP1 has an overall higher performance than existing top statistical potentials/scoring functions in identifying native/near-native structures and ranking the decoy structures. Especially, rsRNASP1 shows an apparently improved performance on a new data set from the CASP15 competition.