Introduction: Immune checkpoint inhibitors (ICIs) have changed the paradigm of cancer treatment, but their effectiveness in some patients with epidermal growth factor receptor (EGFR) mutations is unsatisfactory. Therefore, it is necessary to develop a new biomarker for combined immunotherapy strategies to maximize the clinical benefits.
Methods: We collected and investigated 34 pan-cancer scRNA-Seq cohorts from The Cancer Genome Atlas (TCGA) and 10 bulk RNA-Seq cohorts utilizing multiple machine learning (ML) algorithms to identify and verify a representative EGFR-related gene signature (EGFR.Sig) as a predictive biomarker for immunotherapy response. Core genes were identified as Hub-EGFR.Sig to predict the prognosis of cancers and to understand the crosstalk between EGFR and the tumor immune microenvironment (TIME).
Results: EGFR.Sig can accurately predict the ICI response with an AUC of 0.77, demonstrating superior predictive performance compared to previously established signatures. Twelve core genes in EGFR.Sig were identified as Hub-EGFR.Sig, of which 4 immune resistance genes were previously verified in different CRISPR cohorts. Notably, the prognosis most related to Hub-EGFR.Sig was bladder cancer, which can be divided into two clusters with different responses to immunotherapy based on Hub-EGFR.Sig.
Discussion: We developed a promising pan-cancer signature based on EGFR-related genes to serve as a biomarker for immunotherapy response and survival outcome prediction. Furthermore, core genes were identified for future targeting, which will pave the way for improving the effect of immunotherapy in the context of combination immunotherapies.
Keywords: epidermal growth factor receptor (EGFR); immune checkpoint inhibitors (ICIs); immunotherapy response; pan-cancer; scRNA-seq.
Copyright © 2025 Ye, Chen, Chen, Liu, Kong, Lin, Zhu, Sun and Xu.