Immune clustering reveals molecularly distinct subtypes of lung adenocarcinoma

Abstract

AbstractLung adenocarcinoma, the most prevalent type of non-small cell lung cancer, consists of two driver mutations in KRAS or EGFR. In general, these mutations are mutually exclusive, and biologically and clinically different. In this study, we attempted to find if we could separate lung adenocarcinoma tumors by their immune profile using an unsupervised machine learning method. By projecting RNA-seq data into inferred immune profiles and using unsupervised learning, we were able to divide the lung adenocarcinoma population into three subgroups, one of which appeared to contain mostly EGFR patients. We argue that EGFR mutations in each subgroup are different immunologically which implies a distinct tumor microenvironment and might relate to the relatively high resistance of EGFR-positive tumors to immune checkpoint inhibitors. However, we could not make the same claim about KRAS mutations.Simple SummaryLung adenocarcinoma, the most prevalent type of non-small cell lung cancer, is most commonly driven by mutations in KRAS or EGFR. In this study, we attempted to find if we could separate lung adenocarcinoma tumors by their immune profile using an unsupervised machine learning method. We used established tools to infer the immune profile of each tumor from its RNA-seq and using unsupervised learning, we were able to divide the lung adenocarcinoma population into three subgroups, one of which appeared to contain mostly patients with EGFR mutations. We argue that tumors with EGFR mutations in each subgroup are different immunologically which implies a distinct tumor microenvironment and might relate to the relatively high resistance of EGFR-positive tumors to immune checkpoint inhibitors. However, we could not make the same claim about KRAS mutations.