Integrative analysis of single-cell and bulk RNA sequencing unveils a machine learning-based pan-cancer major histocompatibility complex-related signature for predicting immunotherapy efficacy.

Major histocompatibility complex (MHC) could serve as a potential biomarker for tumor immunotherapy, however, it is not yet known whether MHC could distinguish potential beneficiaries. Single-cell RNA sequencing datasets derived from patients with immunotherapy were collected to elucidate the association between MHC and immunotherapy response. A novel MHCsig was developed and validated using large-scale pan-cancer data, including The Cancer Genome Atlas and immunotherapy cohorts. The therapeutic value of MHCsig was further explored using 17 CRISPR/Cas9 datasets. MHC-related genes were associated with drug resistance and MHCsig was significantly and positively associated with immunotherapy response and total mutational burden. Remarkably, MHCsig significantly enriched 6% top-ranked genes, which were potential therapeutic targets. Moreover, we generated Hub-MHCsig, which was associated with survival and disease-special survival of pan-cancer, especially low-grade glioma. This result was also confirmed in cell lines and in our own clinical cohort. Later low-grade glioma-related Hub-MHCsig was established and the regulatory network was constructed. We provided conclusive clinical evidence regarding the association between MHCsig and immunotherapy response. We developed MHCsig, which could effectively predict the benefits of immunotherapy for multiple tumors. Further exploration of MHCsig revealed some potential therapeutic targets and regulatory networks.