Gene co-expression modules integrated with immunoscore predicts survival of non-small cell lung cancer

This study aimed to deconvolve the levels of infiltrating immune cells in non-small cell lung cancer (NSCLC) and to identify specific gene co-expression modules associated with prognosis of NSCLC.

CIBERSORT algorithm was employed to infer the relative abundance of 22 immune cell subtypes in 1751 NSCLC subjects. The patterns of immune infiltration were identified for NSCLC with different clinical and genomic features and were used to construct an immunoscore by LASSO regression associated with NSCLC survival. Weighted gene co-expression network analysis (WGCNA) was employed to identify specific modules related to immunoscore and NSCLC survival. An integrated prognostic model was constructed with immunoscore combined with the available clinical variables and the selected gene modules to predict the prognosis of NSCLC.

We found distinct immune infiltration patterns for NSCLC with different genotype. EGFR-mutant NSCLC was characterized by enriched resting memory CD4+ T cell. An immunoscore was established based on the infiltration abundance of 17 selected immune cell subtypes. Patients with a low immunoscore had a prolonged survival and higher abundance of CD4+ T cell, resting dendritic cells and resting mast cells. The WGCNA analysis identified the gene modules significantly associated with immunoscore and the prognosis of NSCLC. The immunoscore was further incorporated with clinical parameters and selected gene modules to fit a predictive model which stratified patients into subgroups with significantly different survival.

The distinct immune profiles are associated with differential overall survival of NSCLC and the integrated model can robustly predict the prognosis of NSCLC.