# Two inflammation-related genes model could predict risk in prognosis of patients with lung adenocarcinoma

## TL;DR

This study identifies two inflammation-related genes that can predict the prognosis of lung adenocarcinoma patients, offering a new tool for personalized treatment.

## Contribution

A novel two-gene model (MS4A1 and MS4A2) is developed to predict prognosis in lung adenocarcinoma patients.

## Key findings

- Two LUAD subtypes were identified with distinct clinical outcomes and immune infiltration profiles.
- A two-mRNA model (MS4A1 and MS4A2) effectively separates high-risk and low-risk LUAD patients with strong predictive power.
- The model's predictive accuracy was validated in three independent datasets.

## Abstract

In lung adenocarcinoma (LUAD), there remains a dearth of efficacious diagnostic studies including some inflammation-related genes to identify the LUAD subgroups with different clinical outcomes.

First, two molecular subgroups were identified with mRNA expression profiling from The Cancer Genome Atlas (TCGA) by K-means algorithm. Gene set enrichment analysis (GSEA), immune infiltration, and Gene set variation analysis (GSVA) were applied to explore the biological functions between these two subtypes. Then, univariate and multivariate Cox regression analyses were selected to evaluate the independence of these subtypes in LUAD. Next, lasso regression was applied to identify the high-precision mRNAs to predict the subtype with favorable prognosis. Finally, a two-mRNA model was constructed using the method of multivariate Cox regression, and the effectiveness of the model was validated in a training set (n = 310) and three independent validation sets (n = 1.

Comprehensive genomic analysis was conducted of 310 LUAD samples and identified two subtypes associated with molecular classification and clinical prognosis: immune-enriched and non-immune-enriched subgroup. Then, a new model was developed based on two mRNAs (MS4A1 and MS4A2) in TCGA dataset and divided these LUAD patients into high-risk and low-risk subgroup with significantly different prognosis (HR = 1.644 (95% CI 1.153–2.342); p < 0.01), which was independence of the other clinical factors (p < 0.05). In addition, this new model had similar predictive effects in another three independent validation sets (HR > 1.445, p < 0.01).

We constructed a robust model for predicting the risk of LUAD patients and evaluated the clinical outcomes independently with strong predictive power. This model stands as a reliable guide for implementing personalized treatment strategy.

The online version contains supplementary material available at 10.1007/s12094-025-03861-w.