Prediction of RNA:DNA:DNA triple helix formation using next-generation sequencing data

Abstract RNA:DNA:DNA triple helix (triplex) formation regulates gene expression, but is difficult to study experimentally in vivo. This makes accurate computational prediction of triplex formation highly important in the field of RNA research. Current predictive methods have used canonical Hoogsteen base pairing rules, which whilst biophysically valid, may not reflect the plastic nature of cell biology. Herein, we present TriplexAligner, a local alignment tool implementing probabilistic scoring matrices learned from triplex-forming sequences captured in published triplexRNA-seq and triplexDNA-seq experiments. Short, conserved sequence elements were found to be enriched at points of triplex formation. Probabilistic mapping codes between RNA and DNA sequences were learned by Expectation-Maximisation, and used as scoring matrices for local alignment. TriplexAligner predicts RNA-DNA interactions identified in all-to-all sequencing data more accurately than previously published tools, and also predicts previously studied triplex interactions with known regulatory functions.