Integrative machine learning-driven prioritization of ceRNA networks in adrenocortical carcinoma

Adrenocortical carcinoma (ACC) is a rare and highly aggressive endocrine malignancy in urgent need of robust biomarkers and novel therapeutic targets. In this study, a machine learning (ML)-driven framework is introduced to systematically model microRNA-messenger RNAmiRNA-mRNA regulatory associations and reconstruct context-specific ceRNA networks in ACC, leveraging harmonized RNA-Seq and miRNA-Seq data from The Cancer Genome Atlas-ACC and Genotype-Tissue Expression (2025). Multiple regression models were benchmarked, with random forest achieving superior predictive performance (R² = 0.9467 for normal and 0.9044 for the tumor datasets), enabling accurate identification of high-confidence miRNA-mRNA interactions. Integration of ML predictions with established reference datasets (TargetScan and miRTarBase) was demonstrated to have strong biological validity, while subsequent network analyses revealed extensive topological rewiring and a loss of regulatory hub connectivity in tumor tissue versus normal adrenal tissue. Moreover, a set of new, high-confidence miRNA-mRNA interactions in ACC was identified, implicating new candidates in oncogenic signaling and extracellular matrix remodeling. Together, a robust resource is provided for prioritizing functional interactions and regulatory hubs for future experimental validation, and the power of integrative ML approaches is underscored in advancing systems-level understanding and biomarker discovery in rare cancers such as ACC.