Leveraging Disease Association Degree for High-Accuracy MicroRNA Target Prediction

AO_SCPLOWBSTRACTC_SCPLOWMicroRNAs (miRNAs) regulate gene expression by binding to mRNAs, inhibiting translation, or promoting mRNA degradation. Accurate identification of functional miRNA-target interactions (MTIs), typically validated by methods like western blot or reporter assay, remains challenging due to the scarcity of experimental data compared to the vast number of sequence-based predictions. This study pioneers a novel approach focusing solely on the disease association degree between miRNAs and their target genes. We propose that this single feature is sufficient for distinguishing experimentally validated functional MTIs from sequence-based predicted MTIs in a binary classification task. To quantify miRNA-gene disease association, we fine-tuned Sentence-BERT to generate disease description embeddings and compute their semantic similarity. Remarkably, using only disease association features, miRTarDS achieved an F1 score of 0.88 on the task of distinguishing functional from predicted MTIs in the external validation set. The approach also exhibits generalizability across different gene-disease association databases. This study demonstrates disease association as a powerful, independent dimension for prioritizing high-confidence functional MTIs.