Fig. 3: Mutation landscapes of GPS genes were depicted and compared in different NSCLC cohorts.

To comprehensively depict the mutational characteristics of the 3 GPS-genes in NSCLC along with different clinicopathological features such as sex, smoking status, ethnicity, and histology, four large-scale NSCLC cohorts [OrigiMed-NSCLC (n = 1954), OncoSG-LUAD (n = 302), MSK-NSCLC (n = 1423), TCGA-NSCLC (n = 1144)] were enrolled for further investigation. a–d OncoPrint of putative driver genes including TP53, EGFR, and KRAS and 3-GPS genes was plotted to depict their mutational characteristics and variant details in each cohort. e Multivariate logistic regression analysis was performed on the 3 GPS-genes in the meta-cohort (n = 4208) merged from the four cohorts. The forest plot illustrated that all the 3 GPS-genes are significantly mutated in smokers. Particularly, STK11 is more frequently mutated in the Caucasian ethnicity. In the aspect of pathological subtype, STK11 is more frequently mutated in LUAD, while FAT1 is more frequently mutated in LUSC. f Similar distribution of TMB was observed in each cohort: (i) Significant difference in TMB level was observed among distinct GPS groups; (ii) GPS-pos LUADs always display a high level of TMB, while GPS-zero LUADs display the lowest TMB; (iii) GPS-neg LUADs, which tend to be resistant to ICB therapy, always display significantly higher TMB than GPS-zero samples in each cohort, even significantly higher than GPS-pos samples in the OncoSG cohort. g, h A total of 488 LUAD patients with detailed information of age, sex, pathological stage (pstage), TMB, GPS, and survival follow-up (PFS and OS) were extracted from the TCGA project, and multivariate Cox regression analysis demonstrated that pstage acts as the only one independent risk factor for both PFS and OS. In comparison, GPS fails to exert a significant influence on prognosis of LUAD patients who did not receive ICB therapy. ***p < 0.001.