Differential BUM-HMM: a robust statistical modelling approach for detecting RNA flexibility changes in high-throughput structure probing data

Abstract

Combining RNA structure probing with high-throughput sequencing technologies has greatly enhanced our ability to analyze RNA structure at transcriptome scale. However, the high level of noise and variability encountered in these data call for the development of computational tools that robustly extract RNA structural information. Here we present diffBUM-HMM, a noise-aware model that enables accurate detection of RNA flexibility and conformational changes from high-throughput RNA structure-probing data. DiffBUM-HMM is compatible with a wide variety of high-throughput RNA structure probing data, taking into consideration biological variation, sequence coverage and sequence representation biases. We demonstrate that, compared to the existing approaches, diffBUM-HMM displays higher sensitivity while calling virtually no false positives. DiffBUM-HMM analysis of ex vivo and in vivo Xist SHAPE-MaP data detected many more RNA structural differences, involving mostly single-stranded nucleotides located at or near protein-binding sites. Collectively, our analyses demonstrate the value of diffBUM-HMM for quantitatively detecting RNA structural changes and reinforce the notion that RNA structure probing is a very powerful tool for identifying protein-binding sites.