Identifying RNA contacts from SHAPE-MaP by partial correlation analysis

TL;DR

This paper demonstrates that partial correlation analysis of SHAPE-MaP data effectively identifies RNA contacts, offering a novel approach for RNA structural prediction that outperforms naive correlation methods.

Contribution

It introduces the use of partial correlation analysis for RNA contact detection from SHAPE-MaP data, improving accuracy over naive correlation techniques.

Findings

Partial correlation analysis outperforms naive correlation in identifying RNA contacts.

The method is applicable to various high-throughput RNA structural assays.

Results suggest potential for enhanced 3D RNA structure prediction.

Abstract

In a recent paper Siegfried et al. published a new sequence-based structural RNA assay that utilizes mutational profiling to detect base pairing (MaP). Output from MaP provides information about both pairing (via reactivities) and contact (via correlations). Reactivities can be coupled to partition function folding models for structural inference, while correlations can reveal pairs of sites that may be in structural proximity. The possibility for inference of 3D contacts via MaP suggests a novel approach to structural prediction for RNA analogous to covariance structural prediction for proteins. We explore this approach and show that partial correlation analysis outperforms na\"ive correlation analysis. Our results should be applicable to a wide range of high-throughput sequencing based RNA structural assays that are under development.