Beyond seed match: Improving miRNA target prediction using PAR-CLIP data

Since miRNA plays an important role in post-transcript regulation, many computational approaches have been proposed for miRNA target prediction. Yet, the existing algorithms lack the capability to predict the true target when the perfect seed match presents in mRNA sequences and methods based on seed-match still suffer from a high false positive rate. Therefore, this paper proposes a new prediction method that exploits the data produced by the PAR-CLIP, which is a recent high throughput, high precision technology for genome-wide miRNA targets. This algorithm searches true miRNA targets among the candidates with seed-matches by using machine learning approaches. The target prediction results on top 20 expressed miRNAs in HEK293 cells of AGO1-4 proteins PAR-CLIP data show that given presence of seed pairing, the proposed method greatly outperforms the traditional miRNA target prediction algorithms and improve the precision significantly. Because biologists usually need to mutate the seed region to validation the miRNA targets, and only capable of conducting biological experiments on limited miRNA and mRNA sequences due to the time and cost, the proposed approach will make significant impact on the biology and healthcare fields.