Document Type

Poster

Publication Date

4-17-2026

Abstract

Oral squamous cell carcinoma (OSCC) is diagnosed at late stages in the majority of cases despite its physically accessible location, resulting in five-year survival rates that drop from 90% at Stage I to below 50% at Stage III or IV. Current screening relies on visual examination and invasive biopsy, neither of which can detect the molecular changes that precede visible tumor formation. MicroRNAs (miRNAs) in saliva reflect tumor biology before lesions appear, but the signal spans over 2,000 species simultaneously, making traditional statistical methods insufficient for reliable diagnostic pattern extraction.

This study applies a Random Forest classifier to miRNA expression data from GEO dataset GSE124566, which contains paired tumor and matched normal tongue tissue samples from 10 patients (n = 20). After filtering for human miRNA species, 2,006 features were retained. The model was evaluated using leave-one-out cross-validation, achieving 100% accuracy, 100% sensitivity, 100% specificity, and an AUC of 1.000. Feature importance analysis identified ten high-weight miRNAs, led by hsa-miR-513c-5p (strongly downregulated in tumor tissue), hsa-miR-22-3p, hsa-miR-21-3p, and hsa-miR-424-5p. Eight of the ten top features were upregulated in cancer; the downregulation of miR-513c-5p and miR-23b-3p is consistent with tumor-suppressive function. The appearance of both miR-21 strands (3p and 5p) in the top ten suggests oncogenic overproduction of miR-21 beyond normal strand asymmetry.

These findings provide a proof-of-concept framework for a non-invasive, saliva-based OSCC screening tool. Validation in larger salivary datasets is required before clinical application, but the results demonstrate that machine learning can extract a diagnostic biomarker from transcriptomic data that would otherwise be too complex for traditional analysis.

Included in

Education Commons

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.