Multi-omics and Machine Learning-driven CD8+ T Cell Heterogeneity Score for Prognosis

Abstract

The heterogeneity of head and neck squamous cell carcinoma (HNSCC) poses a significant challenge to treatment, underscoring the urgent need for more precise and personalized therapeutic approaches. CD8+ T cells, integral components of the tumor immune microenvironment, have emerged as key targets for immunotherapy. Our research has established a correlation between a decrease in CD8+ T cell score and a poor clinical prognosis, highlighting the prognostic value of this biomarker. By analyzing the gene expression related to CD8+ T cells, we have differentiated HNSCC into cold and hot tumor subtypes, uncovering disparities in clinical prognosis and responses to immunotherapy. Utilizing eight machine learning methods, we identified the key gene OLR1. Single-cell analysis of HNSCC tissues and peripheral blood, along with spatial transcriptome analysis, revealed that OLR1 predominantly functions in macrophages, modulating the immune microenvironment of HNSCC. The expression level of OLR1 may serve as a predictive marker for immunotherapy responses. Moreover, drug sensitivity analysis and molecular docking studies have indicated that simvastatin and pazopanib are potential inhibitors of OLR1. These findings suggest that simvastatin and pazopanib could open up innovative potential therapeutic avenues for individuals with HNSCC.

My Contribution

Provided ML support, implementing methods including LASSO regression to identify key prognostic genes from multi‑omics data and supply features for building the CD8+ T cell heterogeneity score (CD8THS).

Access paper here

Recommended citation: He D, Yang Z, Zhang T, Luo Y, Peng L, Yan J, Qiu T, Zhang J, Qin L, Liu Z, Zhang X, Lin L, Sun M. Multi-omics and machine learning-driven CD8+ T cell heterogeneity score for head and neck squamous cell carcinoma. Molecular Therapy Nucleic Acids. 2025;36(1):102413.
Download Paper