RNA Tertiary Structure Prediction Algorithm at Atomic Accuracy

Predicting the tertiary structure of macromolecules is a great challenge in computational biology. Rosetta biomolecular modeling suite provides solutions to various challenges in bioengineering. However, the de novo modeling of RNA often involves solving some well-defined small problems, which cannot be solved by the current Rosetta framework methods. these methods need to be further optimized. In this paper, a Stepwise Monte Carlo Parallelization (SWP) algorithm is proposed. The algorithm uses Monte Carlo algorithm to randomly search millions of conformations for each motif, uses the parallel mechanism to increase the breadth of conformational sampling, and improves the modeling accuracy by returning the modeling results to the shared pool for a new round of modeling. We modeled six single-stranded RNA loops extracted from RNA switches by SWP, including four motifs that could not be solved by the modeling algorithm based on knowledge mining. SWP is an ab initio modeling method, and its performance is better than the existing algorithms under Rosetta framework. Especially in improving modeling accuracy.