miRBiT: a rules-based single-sample serum miRNA classifier for pan-cancer detection with multi-cohort validation

Area of Science:

• Biochemistry
• Molecular Biology
• Oncology

Background:

• Liquid biopsy presents a minimally invasive method for early cancer diagnosis.
• MicroRNAs (miRNAs) are stable, non-coding RNAs with diagnostic potential due to their dysregulation in disease.
• Current miRNA-based cancer classifiers often rely on cohort comparisons, limiting clinical application.

Purpose of the Study:

• To develop a single-sample miRNA classifier for robust cancer detection.
• To enable personalized and minimally invasive cancer screening.
• To validate the classifier across diverse datasets and disease cohorts.

Main Methods:

• Development of miRNA binary trend (miRBiT), a rules-based single-sample classifier.
• Training and testing miRBiT on extensive datasets (16,190 samples) across multiple independent cohorts (9 datasets, 8 disease cohorts).
• Utilizing intra-sample miRNA expression signatures for classification.

Main Results:

• miRBiT achieved high sensitivity and specificity in classifying 'cancer' versus 'non_cancer' samples, including healthy and other diseases.
• The classifier demonstrated robust performance across independent datasets and disease cohorts.
• An interactive web application, miRBiT Explorer, was developed for visualizing serum miRNA expression in 46,349 samples.

Conclusions:

• miRNAs hold significant potential for accurate, personalized cancer classification.
• miRBiT offers a scalable solution for minimally invasive cancer screening and early detection.
• The study validates the utility of miRNAs in robust, large-scale cancer diagnostics.