ABSTRACT
Despite the importance of microRNAs (miRNAs) in regulating a broad variety of biological processes, accurately predicting the transcripts they repress remains a challenge. Recent research suggests improved miRNA target prediction using a biochemical model combined with empirically-derived affinity predictions across 12mer sequences. Here, we translate this approach into a generally applicable, flexible and user-friendly tool (scanMiR). By compressing and handling miRNA 12mer affinity predictions into lightweight models, scanMiR can efficiently scan for both canonical and non-canonical binding sites on transcripts and custom sequences (including circRNAs and lncRNAs). Aggregation of binding sites into predicted transcript repression using a generalized biochemical model correlates better with experimental data than the most accurate alternative publicly available predictions. Moreover, a flexible 3’-supplementary alignment enables scanMiR to highlight and visualize unconventional modes of miRNA target mRNA interactions, such as bindings leading to target-directed miRNA degradation (TDMD) and slicing. By specifically scanning for these unconventional binding sites in brain-derived expression data, we provide a first systematic overview of potential TDMD and slicing sites on brain-specific lncRNAs as well as circRNAs. Finally, in addition to the main bioconductor package implementing these functions, we provide a user-friendly web application enabling the scanning of sequences, the visualization of predicted bindings, and the browsing of predicted target repression.
Competing Interest Statement
The authors have declared no competing interest.