Tissue-Specific MicroRNA Function: Integrating Target Prediction with Expression Data

MicroRNAs (miRNAs) play a critical role as post-transcriptional regulators of gene expression, influencing diverse cellular processes in both healthy and diseased states. Despite over two decades of intensive research, the comprehensive identification of miRNA targets remains a formidable challenge. Existing experimental techniques and computational tools often fall short of accurately mapping the miRNA interactome, leaving key aspects of miRNA function only partially understood. To address these limitations, we have developed microT-CNN, the latest iteration of the DIANA-microT-CDS model (Reczko M et al., Bioinformatics, 2012). MicroT-CNN utilizes an innovative multi-layer deep convolutional neural network framework to predict both canonical and non-canonical seed-based miRNA binding events, as well as previously overlooked 3′ compensatory interactions. It analyzes miRNA binding events across 3′UTRs and coding regions, enabling the comprehensive capture of the miRNA interaction spectrum. The model was trained and validated using a unique and extensive dataset of over 60,000 miRNA binding events and approximately 30,000 tissue-matched functional miRNA-gene target pairs. These data were derived from 26 cell types and incorporated findings from ~130 AGO-CLIP-seq experiments and ~70 miRNA perturbation experiments, ensuring robustness and accuracy. For assessing miRNA tissue specificity, we utilized DIANA-miTED (Kavakiotis I et al., Nucleic Acids Research, 2022), a systematic resource comprising miRNA expression profiles derived from the analysis of 15,183 raw human small RNA-Seq (sRNA-Seq) datasets from the Sequence Read Archive (SRA) and The Cancer Genome Atlas (TCGA). Gene tissue expression data were obtained from the Genotype-Tissue Expression (GTEx) project. This combined analysis was validated using TarBase(Skoufos G et al., Nucleic Acids Research, 2023), a database of experimentally supported miRNA targets, resulting in a significant enhancement in the accuracy of miRNA target annotations. Our findings highlight the potential of deep learning approaches and integrated datasets in advancing miRNA target prediction, paving the way for a deeper understanding of miRNA-mediated regulation in tissue-specific and pathological contexts. Funded by the European Union – NextGenerationEU through Greece 2.0—National Recovery and Resilience Plan, under the call ”Flagship actions in interdisciplinary scientific fields with a special focus on the productive fabric” (ID 16618), project name ”Bridging big omic, genetic and medical data for Precision Medicine implementation in Greece” (project code TAEDR-0539180). European Union – NextGenerationEU through Greece 2.0—National Recovery and Resilience Plan, (ID 16618) (project code TAEDR-0539180). This abstract was presented at the American Physiology Summit 2025 and is only available in HTML format. There is no downloadable file or PDF version. The Physiology editorial board was not involved in the peer review process.