WebServer

Micro RNAs (miRNAs) are short noncoding RNAs that regulate gene expression at the post-transcriptional level by targeting messenger RNA 3’ UTR regions. The intricacies of miRNA targeting pose a challenge to current understanding. Despite the availability of many computational tools for miRNA target prediction, these tools still require hand-crafted features that are fed to classification engines. In this study, we proposed a deep learning-based model, termed Mimosa, for predicting miRNA targets at the gene level. We trained Mimosa using the pairs of miRNAs & candidate target sites (CTS) datasets and evaluated it on pairs of miRNAs & mRNA 3’UTRs datasets. We incorporated seed match patterns, including both canonical and non-canonical seed matches, into the Transformer encoder model to enable in-depth feature extraction. This innovative approach eliminates the need for a separate CTS selection step. The performance benchmarking test results demonstrated the superiority of Mimosa compared with state-of-the-art approaches.