BaRDIC: robust peak calling for RNA-DNA interaction data

Abstract

ABSTRACTChromatin-associated non-coding RNAs play important roles in various cellular processes by targeting genomic loci. Two types of genome-wide NGS experiments exist to detect such targets: “one-to-all”, which focuses on targets of a single RNA, and “all-to-all”, which captures targets of all RNAs in a sample. As with many NGS experiments, they are prone to biases and noise, so it becomes essential to detect “peaks” – specific interactions of an RNA with genomic targets. Here we present BaRDIC – Binomial RNA-DNA Interaction Caller – a tailored method to detect peaks in both types of RNA-DNA interaction data. BaRDIC is the first tool to simultaneously take into account the two most prominent biases in the data: chromatin heterogeneity and distance-dependent decay of interaction frequency. Since RNAs differ in their interaction preferences, BaRDIC adapts peak sizes according to the abundances and contact patterns of individual RNAs. These features enable BaRDIC to make more robust predictions than currently applied peak-calling algorithms and better handle the characteristic sparsity of all-to-all data. BaRDIC package is freely available athttps://github.com/dmitrymyl/BaRDIC.