A study published in PLOS Computational Biology describes RNACOREX, a computational tool designed to identify disease-associated RNA regulatory networks and use them to classify patient samples based on gene expression data. The software focuses on interactions between microRNAs (miRNAs) and messenger RNAs (mRNAs), which play an important role in gene regulation but are difficult to interpret in diagnostic workflows.
miRNAs influence gene expression by binding to mRNAs, often acting together rather than individually. Large databases contain thousands of predicted and experimentally validated miRNA–mRNA interactions, but many predicted interactions are false positives, and their relevance to specific diseases is often unclear. This limits the usefulness of miRNA data for disease classification and biomarker development.
RNACOREX addresses this challenge by combining prior biological knowledge with patient expression data in a single analysis pipeline. The software first filters potential miRNA–mRNA interactions using established databases, including TargetScan, DIANA-microT, miRTarBase, and TarBase. Each remaining interaction is then scored in two ways: a structural score that reflects existing biological evidence and a functional score that measures how strongly the interaction is associated with a clinical outcome using expression data.
Based on these ranked interactions, RNACOREX builds probabilistic models known as Conditional Linear Gaussian classifiers. These models analyze continuous miRNA and mRNA expression values without requiring data simplification and generate both predictions and an interpretable network of RNA interactions. This allows users to see which specific RNAs and interactions contribute to each classification result.
The authors evaluated RNACOREX using miRNA and mRNA data from 13 cancer types included in The Cancer Genome Atlas. In these analyses, patients were classified into long- and short-survival groups. Across tumor types, RNACOREX achieved classification performance similar to commonly used machine learning methods such as random forest, gradient boosting, and support vector machines. Unlike many of these approaches, RNACOREX also provides a visual and interpretable representation of the underlying regulatory network.
The software highlights RNA interactions that recur across multiple cancers as well as those that appear tissue specific. These findings may help researchers prioritize candidate biomarkers or regulatory pathways for further study.
The authors emphasize that RNACOREX is a research tool and does not establish clinical utility on its own. However, it provides a structured and transparent approach to exploring RNA regulation in disease. For laboratories working with transcriptomic data, RNACOREX offers a way to link RNA expression patterns to clinical phenotypes while retaining interpretability.
Newsletters
Receive the latest pathologist news, personalities, education, and career development – weekly to your inbox.
