Mutational signature-based biomarker to predict the response of immune checkpoint inhibitors therapy in cancers

Patients have a limited response rate to immune checkpoint inhibitors (ICIs) therapy. Although several biomarkers have been proposed, their ability to accurately predict the response to ICIs therapy remains unsatisfactory. In addition, mutational signatures were validated to be associated with ICIs therapy. Therefore, we developed a mutational signature-based biomarker (MS-bio) to predict the response to ICIs therapy. Based on differentially mutated genes, we extracted six mutational signatures (single-base substitution (SBS)-A, SBS-B, SBS-C, SBS-D, double-base substitution (DBS)-A, and DBS-B) as MS-bio, and constructed a random forest (RF) model to predict the response. Internal and external validations consistently demonstrated the excellent predictive capability of MS-bio, with an accuracy reaching up to 0.82. Moreover, MS-bio exhibited superior performance compared to existing biomarkers. To further validate the accuracy of MS-bio, we explored its performance in The Cancer Genome Atlas (TCGA) cohort and found that the predicted responders were immunologically "hot". Finally, we found that SBS-C had the highest importance in prediction and was related to T cell differentiation. Overall, here we introduced MS-bio as a novel biomarker for accurately predicting the response to ICIs therapy, thereby contributing to the advancement of precision medicine.