Risk stratification and prognosis prediction based on inflammation-related gene signature in lung squamous carcinoma

Inflammation is known to have an intricate relationship with tumorigenesis and tumor progression while it is also closely related to tumor immune microenvironment. Whereas the role of inflammation-related genes (IRGs) in lung squamous carcinoma (LUSC) is barely understood. Herein, we recognized IRGs associated with overall survival (OS), built an IRGs signature for risk stratification and explored the impact of IRGs on immune infiltration landscape of LUSC patients.

The IRG signature was a predictor for patients with LUSC and might serve as a potential indicator of the efficacy of immunotherapy. The nomogram based on the IRG signature showed a relatively good predictive performance in survival.

The RNA-sequencing and clinicopathological data of LUSC patients were downloaded from The Cancer Genome Atlas (TCGA) and the Gene Expression Omnibus (GEO) database, which were defined as training and validation cohorts. Cox regression and least absolute shrinkage and selection operator analyses were performed to build an IRG signature. CIBERSORT, microenvironment cell populations-counter and tumor immune dysfunction and rejection (TIDE) algorithm were used to perform immune infiltration analysis.

A two-IRG signature consisting of KLF6 and SGMS2 was identified according to the training set, which could categorize patients into two different risk groups with distinct OS. Patients in the low-risk group had more anti-tumor immune cells infiltrated while patient with high-risk had lower TIDE score and higher levels of immune checkpoint molecules expressed. The IRG signature was further identified as an independent prognostic factor of OS. Subsequently, a prognostic nomogram including IRG signature, age, and cancer stage was constructed for predicting individualized OS, whose concordance index values were 0.610 (95% CI: 0.568-0.651) in the training set and 0.652 (95% CI: 0.580-0.724) in validation set. Time-dependent receiver operator characteristic curves revealed that the nomogram had higher prediction accuracy compared with the traditional tumor stage alone.