Predicting microRNA target genes using pan-cancer correlation patterns.

The interaction relationship between miRNAs and genes is important as miRNAs play a crucial role in regulating gene expression. In the literature, several databases have been constructed to curate known miRNA target genes, which are valuable resources but likely only represent a small fraction of all miRNA-gene interactions. In this study, we constructed machine learning models to predict miRNA target genes that have not been previously reported. Using the miRNA and gene expression data from TCGA, we performed a correlation analysis between all miRNAs and all genes across multiple cancer types. The correlations served as features to describe each miRNA-gene pair. Using the existing databases of curated miRNA targets, we labeled the miRNA-gene pairs, and trained machine learning models to predict novel miRNA-gene interactions. For the miRNA-gene pairs that were consistently predicted across the models, we called them significant miRNA-gene pairs. Using held-out miRNA target databases and a literature survey, we validated 5.5% of the predicted significant miRNA-gene pairs. The remaining predicted miRNA-gene pairs could serve as hypotheses for experimental validation. Additionally, we explored several additional datasets that provided gene expression data before and after a specific miRNA perturbation and observed consistency between the correlation direction of predicted miRNA-gene pairs and their regulatory patterns. Together, this analysis revealed a novel framework for uncovering previously unidentified miRNA-gene relationships, enhancing the collective comprehension of miRNA-mediated gene regulation.