Improving microRNA target prediction by modeling with unambiguously identified microRNA-target pairs from CLIP-ligation studies

Motivation: MicroRNAs (miRNAs) are small non-coding RNAs that are extensively involved in many physiological and disease processes. One major challenge in miRNA studies is the identification of genes targeted by miRNAs. Currently, most researchers rely on computational programs to initially identify target candidates for subsequent validation. Although considerable progress has been made in recent years for computational target prediction, there is still significant room for algorithmic improvement. Results: Here, we present an improved target prediction algorithm, which was developed by modeling high-throughput profiling data from recent CLIPL (crosslinking and immunoprecipitation followed by RNA ligation) sequencing studies. In these CLIPL-seq studies, the RNA sequences in each miRNA-target pair were covalently linked and unambiguously determined experimentally. By analyzing the CLIPL data, many known and novel features relevant to target recognition were identified and then used to build a computational model for target prediction. Comparative analysis showed that the new algorithm had improved performance over existing algorithms when applied to independent experimental data.