Exploring the RNA folding energy landscape using scalable distributed cyberinfrastructure

Abstract

The increasing significance of RNAs in transcriptional or post-transcriptional gene regulation processes has generated considerable interest towards the prediction of RNA folding and its sensitivity to environmental factors. We use Boltzmann-weighted sampling to generate RNA secondary structures, which are used to characterize the energy landscape, via the distributions of energies and base-pair distances. Depending upon the length of an RNA, the number of sequences investigated, and the sample size of generated structures --- generating and analyzing sufficient samples can be computationally challenging. We introduce and develop a lightweight and extensible runtime environment that is effective across a range of RNA sizes and other parameters, as well as over a range of infrastructure -- from traditional HPC grids to clouds, without requiring any changes at the application or user level. The Adaptive Distributed Application Management System (ADAMS) is built upon an extensbile and interoperable pilot-job and supports the concurrent execution of a broad range of task sizes across a range of infrastructure. We use ADAMS to investigate the folding energy landscape for two RNA systems of different sizes: a set of S-adenosyl methionine (SAM) binding RNA sequences known as SAM-I riboswitches and the S gene of the Bovine Corona Virus (BCoV) RNA genome that comprises 4092 nucleotides. Results of the energy and base-pair distance distributions suggest different energy landscapes, implying different folding dynamics. With obtained results, we demonstrated the possibility of utilizing this protocol to explore microscopic origins for reported sequence-dependent variation of binding affinity and gene expression in the two RNA systems.