Immunotherapy has gradually become an important therapy option for lung cancer patients.
The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases were responsible for all the public data.
In our study, we firstly identified 22 characteristic genes of NSCLC immunotherapy response using the machine learning algorithm. Molecule subtyping was then conducted and two patient subtypes were identified Cluster1 and Cluster2. Results showed that Cluster1 patients had a lower TIDE score and were more sensitive to immunotherapy in both TCGA and combined GEO cohorts. Biological enrichment analysis showed that pathways of epithelial-mesenchymal transition (EMT), apical junction, KRAS signaling, myogenesis, G2M checkpoint, E2F targets, WNT/β-catenin signaling, hedgehog signaling, hypoxia were activated in Cluster2 patients. Genomic instability between Cluster1 and Cluster2 patients was not significantly different. Interestingly, we found that female patients were more adaptable to immunotherapy. Biological enrichment revealed that compared with female patients, pathways of MYC target, G2M checkpoints, mTORC1 signaling, MYC target, E2F target, KRAS signaling, oxidative phosphorylation, mitotic spindle and P53 pathway were activated. Meanwhile, monocytes might have a potential role in affecting NSCLC immunotherapy and underlying mechanism has been explored. Finally, we found that SEC14L3 and APCDD1L were the underlying targets affecting immunotherapy, as well as patients survival.
These results can provide direction and guidance for future research focused on NSCLC immunotherapy.