In recent years, mRNA-based vaccines with promising safety and functional characteristics have gained significant momentum in cancer immunotherapy. However, stable immunological molecular subtypes of lung adenocarcinoma (LUAD) and novel tumor antigens for LUAD mRNA vaccine development remain elusive. Therefore, a novel approach is urgently needed to identify suitable LUAD subtypes and potential tumor antigens.

The Cancer Genome Atlas (TCGA), the Genotype Tissue Expression (GTEx), and Gene Expression Omnibus (GEO) databases were utilized to retrieve gene expression data. The LUAD Immunological Multi-Omics Classification (LIMOC) system was developed using seven machine learning (ML) algorithms by performing integrative immunogenomic analysis of single-cell and bulk tissue transcriptome profiling. Subsequently, a panel of approaches was applied to identify novel tumor antigens.

First, the LIMOC system was construct to identify three subtypes: LIMOC1, LIMOC2, and LIMOC3. Second, we identified CHIT1, LILRA4, and MEP1A as novel tumor antigens in LUAD; these genes were up-regulated, amplified, and mutated, and showed a positive association with APC infiltration, making them promising candidates for designing mRNA vaccines. Notably, the LIMOC2 subtype had the worst prognosis and could benefit most from mRNA immunization. Furthermore, we performed a comprehensive in silico screening of approximately 2,000 compounds and identified Sorafenib and Azacitidine as potential subtype-specific therapeutic agents.

Overall, our study established a robust LIMOC system and identified CHIT1, LILRA4, and MEP1A as promising tumor antigen candidates for development of anti-LUAD mRNA vaccines, particularly for the LIMOC2 subtype.