Fig. 5: Landscapes of immune clusters and cell types across NSCLC subtypes and cohorts.

a Immune subtyping based on cell type and pathway enrichment scores. Cell type-based clustering was performed with 205 tumor and 85 normal adjacent to the tumor (NAT) samples, and pathway-based clustering was performed using only tumor samples. The tumor-infiltrating lymphocyte (TIL) pattern, clinical histology (diagnostics, DX), multiomics subtype, tumor stage, and tissue information are described. IC, immune cluster. b (top) DEG_HTE and DEG_CTE were used to generate the UMAP plot of scRNA-seq data. A two-sided t-test was conducted to assess the statistical significance of the differences in gene expression. The color of each point represents the module score of each cell; higher scores are shown in red. UMAP information was obtained from multiple NSCLC studies²³. (bottom) The correlations of 10 cell types and immune clusters with the pattern of TILs were analyzed. The sizes and colors of the circles indicate the statistical significance and correlation coefficient of the correlations, respectively. The horizontal black dotted line indicates P = 0.05. c Hazard ratios for overall survival (OS, left) and relapse-free survival (RFS, right) related to various cell types in the cell type-based immune cluster. A hazard ratio lower than zero (blue box) indicates that the hot-tumor-enriched (HTE) status or a high cell type score was associated with prolonged survival. Error bars (gray lines) represent mean ± 95% confidence interval (CI). Red text indicates statistical significance in the survival analysis by the log-rank Mantel‒Cox test (n = 174). d Correlations of the RNA expression, protein expression, and protein activity of 10 immunomodulators with immune cluster status for our cohort as well as other lung cancer multiomics cohorts^10,13. Correlation coefficients and p-values were obtained from a generalized linear model (GLM). e Correlations between the expression or activity of immunomodulators and the status of driver mutations in our cohort and the Satpathy and Gillette cohorts. The top associations between immunomodulators and known driver genes are described. f The left boxplot shows the RNA and protein expression of SLAMF7 in samples (n = 205) with wild-type or mutant SMARCA4 (n = 205); the right boxplot shows the RNA and protein expression of SLAMF7 in HTE and cold-tumor-enriched (CTE) samples (n = 174). The two-sided t-test was performed to test the differences in expression. The box represents the 25th and 75th percentiles, the central mark denotes the median, and the whiskers extend to the most extreme points within ±1.5 × IQR. g (left) Box (top) and balloon (bottom) plots showing the mean expression of marker genes of HTE and 10 cell types across the multiomics subtypes (n = 174). The marker genes were defined as the top−300 and −30 most overexpressed genes in HTE samples and highly cell type-enriched samples, respectively. (right) The bar (top) and box (middle and bottom) plots show the mutation frequency of SMARCA4 and RNA/protein expression of SLAMF7 across multiomics subtypes (n = 174), respectively. The Kruskal‒Wallis test was performed to assess the differences among the multiomics subtypes. The box represents the 25th and 75th percentiles, the central mark denotes the median, and the whiskers extend to the most extreme points within ±1.5 × IQR.