Structural Prediction of RNA Loops by Simulated Annealing

Abstract

Structural modelling of RNAs have received relatively less attention compared to protein modelling, owing partly due to the involved, long and highly orientable loop structures. RNA loop structures determined by experimental methods are often found wanting due to chemical interactions of the detection method with the structures and for this reason alone, a computational analogue of an RNA loop structure is important to devise. In this work RNA loop modelling problem is presented as an energy minimization problem subject to kinematic and volumetric constraints of the loop. The kinematic constraints are generated from redundancy characterization of RNA loops like in case of a robotic manipulator. The loop energy is estimated from established force fields (like CHARMM etc.)with necessary approximations and the resulting potential optimization problem is solved by simulated annealing methods to obtain stable loop configurations free from loop-loop infringement and loop-anchor infringement. The algorithm is applied, with a fixed and geometric cooling schedule, on several RNA loops of varying number of residues and the resulting structures are accepted as valid configurations if certain statistical criteria are met related to the position error in loop-anchor atom coordinates. For small to moderate loop size, provides results that are comparable to the structures stored in established repositories (Protein data bank, RLooM etc.). Native and optimized structures of some such loops are presented as a proof of concept for the methodology discussed