A Novel Gene Signature based on Immune Cell Infiltration Landscape Predicts Prognosis in Lung Adenocarcinoma Patients

The tumor microenvironment (TME) is created by the tumor and dominated by tumor-induced interactions. Long-term survival of lung adenocarcinoma (LUAD) patients is strongly influenced by immune cell infiltration in TME. The current article intends to construct a gene signature from LUAD ICI for predicting patient outcomes.

For the initial phase of the study, the TCGA-LUAD dataset was chosen as the training group for dataset selection. We found two datasets named GSE72094 and GSE68465 in the Gene Expression Omnibus (GEO) database for model validation. Unsupervised clustering was performed on the training cohort patients using the ICI profiles. We employed Kaplan-Meier estimators and univariate Cox proportional-hazard models to identify prognostic differentially expressed genes in immune cell infiltration (ICI) clusters. These prognostic genes are then used to develop a LASSO Cox model that generates a prognostic gene signature. Validation was performed using Kaplan-Meier estimation, Cox, and ROC analysis. Our signature and vital immune-relevant signatures were analyzed. Finally, we performed gene set enrichment analysis (GSEA) and immune infiltration analysis on our finding gene signature to further examine the functional mechanisms and immune cellular interactions.

Our study found a sixteen-gene signature (EREG, HPGDS, TSPAN32, ACSM5, SFTPD, SCN7A, CCR2, S100P, KLK12, MS4A1, INHA, HOXB9, CYP4B1, SPOCK1, STAP1, and ACAP1) to be prognostic based on data from the training cohort. This prognostic signature was certified by Kaplan-Meier, Cox proportional-hazards, and ROC curves. 11/15 immune-relevant signatures were related to our signature. The GSEA results indicated our gene signature strongly correlates with immune-related pathways. Based on the immune infiltration analysis findings, it can be deduced that a significant portion of the prognostic significance of the signature can be attributed to resting mast cells.

We used bioinformatics to determine a new, robust sixteen-gene signature. We also found that this signature's prognostic ability was closely related to the resting mast cell infiltration of LUAD patients.