Integrative Analysis of Bioinformatics and Machine Learning Algorithms Identifies a Novel Diagnostic Model Based on Costimulatory Molecule for Predicting Immune Microenvironment Status in Lung Adenocarcinoma.

Costimulatory molecules are an indispensable signal for activating immune cells. However, the features of many costimulatory molecule genes (CMGs) in lung adenocarcinoma (LUAD) are poorly understood. This study systematically explored expression patterns of CMGs in the tumor immune microenvironment (TIME) status of patients with LUAD. Their expression profiles were downloaded from The Cancer Genome Atlas and the Gene Expression Omnibus databases. Two robust TIME subtypes ("hot" and "cold") were classified by K-means clustering and estimation of stromal and immune cells in malignant tumor tissues using expression data. The "hot" subtype presented higher infiltration in activated immune cells and enrichments in the immune cell receptor signaling pathway and adaptive immune response. Three CMGs (CD80, LTB, and TNFSF8) were screened as final diagnostic markers by means of Least Absolute Shrinkage Selection Operator and Support Vector Machine-Recursive Feature Elimination algorithms. Accordingly, the diagnostic nomogram for predicting individualized TIME status showed satisfactory diagnostic accuracy in The Cancer Genome Atlas training cohort as well as GSE31210 and GSE180347 validation cohorts. Immunohistochemistry staining of 16 specimens revealed an apparently positive correlation between the expression of CMG biomarkers and pathologic response to immunotherapy. Thus, this diagnostic nomogram provided individualized predictions in TIME status of LUAD patients with good predictive accuracy, which could serve as a potential tool for identifying ideal candidates for immunotherapy.