RRDB: A comprehensive and non-redundant benchmark for RNA-RNA docking and scoring

Abstract

Motivation With the discovery of more and more noncoding RNAs and their versatile functions, RNA-RNA interactions have received increased attention. Therefore, determination of their complex structures is valuable to understand the molecular mechanism of the interactions. Given the high cost of experimental methods, computational approaches like molecular docking have played an important role in the determination of complex structures, in which a benchmark is critical for the development of docking algorithms. Results Meeting the need, we have developed the first comprehensive and nonredundant RNA-RNA docking benchmark (RRDB). The diverse dataset of 123 targets consists of 78 unbound-unbound and 45 bound-unbound (or unbound-bound) test cases. The dataset was classified into three groups according to the interface conformational changes between bound and unbound structures: 47 ′ easy′, 38′ medium′ and 38 ′ difficult′ targets. A docking test with the benchmark using ZDOCK 2.1 demonstrated the challenging nature of the RNA-RNA docking problem and the important value of the present benchmark. The bound and unbound cases of the benchmark will be beneficial for the development and optimization of docking and scoring algorithms for RNA-RNA interactions.