Tumor subtype classification tool for HPV-associated head and neck cancers.

Area of Science:

• Oncology
• Genomics
• Bioinformatics

Background:

• HPV-associated Head and Neck Squamous Cell Carcinoma (HNSCC) exhibits distinct molecular subtypes: IMU (immune strong) and KRT (highly keratinized).
• These subtypes possess unique molecular characteristics, tumor microenvironments, and clinical outcomes, suggesting differential treatment responses.
• A standardized method for subtyping HPV+ HNSCC tumors is currently lacking.

Purpose of the Study:

• To develop and validate a machine learning-based classifier for reliably subtyping HPV+ HNSCC tumors into IMU and KRT categories.
• To highlight the clinical significance of these molecular subtypes in HPV+ HNSCC.
• To provide a webtool for accessible tumor subtyping.

Main Methods:

• RNA sequencing (RNA-seq) data from 67 HNSCC tumors at the University of Michigan Health were analyzed.
• A total of 229 HPV+ HNSCC RNA-seq samples from combined datasets were used to train and test the classifier.
• The classifier's performance was validated on additional cohorts, and associations between subtype and 37 clinicodemographic/molecular variables were assessed.

Main Results:

• The machine learning classifier achieved 100% accuracy on the test set.
• Validation across two independent cohorts confirmed the classifier's ability to separate tumors based on known subtype features.
• Twenty-one significant associations were identified between subtype and clinicodemographic/molecular variables, including novel findings.

Conclusions:

• A reliable classifier utilizing bulk RNA-seq data has been developed for subtyping HPV+ HNSCC into IMU and KRT molecular subtypes.
• This tool enhances the understanding of HPV+ HNSCC heterogeneity.
• The findings underscore the importance of molecular subtyping for HPV+ HNSCC management.