# Identification of a prognostic signature based on immunogenic adverse event-related genes to guide therapy for non-small cell lung cancer

## TL;DR

This study identifies a gene signature linked to immune-related adverse events in non-small cell lung cancer, which helps predict treatment response and survival outcomes.

## Contribution

A novel eight-gene prognostic signature is developed to predict immunotherapy response and survival in NSCLC patients based on immune-related adverse event-related genes.

## Key findings

- The eight-gene signature showed strong predictive accuracy across multiple cohorts with AUC values ranging from 0.758 to 0.807.
- Patients with immune-related adverse events had significantly better clinical outcomes, including higher response rates and longer survival.
- Three key genes (POU2AF1, ANKRD44, CRTAM) showed elevated protein expression in patients experiencing immune-related adverse events.

## Abstract

Immune checkpoint inhibitors (ICIs) improve outcomes in non−small cell lung cancer (NSCLC), yet reliable predictive biomarkers are still lacking. Given that immune−related adverse events (irAEs) often correlate with better ICI efficacy, this study aimed to develop and validate an irAE−related gene signature for risk stratification and treatment–response prediction in NSCLC.

Transcriptomic and clinical data were obtained from the Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases, designated as training and validation cohorts, respectively. Least absolute shrinkage and selection operator (LASSO) regression and Cox proportional hazards models were applied to identify eight irAE-associated genes for constructing a risk score (RS). Multivariate analyses evaluated differences in overall survival (OS), immunotherapy response, irAE incidence, immune escape potential, and tumor microenvironment (TME) profiles between high- and low-risk groups. Prognostic performance was validated using Kaplan–Meier curves, Cox regression, and receiver operating characteristic (ROC) analysis. A nomogram integrating RS with clinical factors was developed to improve prediction stability.

Via gene set variation analysis (GSVA) enrichment of 30 known immunologic gene sets in TCGA-LC, we stratified samples by immune score (cutoff = 0.23), identified 1,057 differentially expressed genes (DEGs) between groups, intersected with cancer-normal DEGs, and selected 132 irAE-related DEGs via Pearson correlation—genes functionally tied to T-cell activation and cytokine-mediated signaling pathways. LASSO–Cox regression derived an eight-gene prognostic signature demonstrating robust cross-cohort validation [area under the curve (AUC): TCGA = 0.770, GSE50081 = 0.767, GSE37745 = 0.758] and predictive accuracy for irAEs in GSE186143 (AUC = 0.807). In a validation NSCLC cohort (n = 41), irAE occurrence correlated with superior clinical outcomes: objective response rate (87.50% vs. 44.00%), disease control rate (93.75% vs. 76.00%), median progression-free survival (20.0 vs. 9.9 months), and overall survival (21.4 vs. 12.1 months). Three key risk genes (POU2AF1, ANKRD44, CRTAM) showed significantly elevated protein expression levels and positive rates in irAE patients. High RS cohorts exhibited immunosuppressive TME characteristics and increased immune escape potential. A clinically relevant nomogram integrating RS with clinical factors (age, stage, and tumor recurrence) improved prognostic stability.

Our irAE-associated gene signature robustly stratifies NSCLC patients for immunotherapy response and survival. Integrating RS with clinical parameters provides a practical tool to balance efficacy and safety.