Fig. 2: Proteomic subtyping of the gastric cancer cohort and its association with clinical characteristics.

a The heatmap depicts the relative abundance (Z score of FOT) of the signature proteins in four subtypes of 179 GC samples. Biological functions related to these signature proteins are denoted on the right. b The association of four proteomic subtypes with clinical characteristics (including gender, grade, Lauren’s type, primary site, HER2 status, RECIST, and tumor purity, etc.) are annotated with *P-values (two-sided Fisher’s exact test for categorical variables, and two-way ANOVA test for continuous variables). c The Kaplan–Meier curves of overall survival (OS) of each proteomic subtype (G-I, n = 28; G-II, n = 56; G-III, n = 75; and G-IV, n = 20). P-value is calculated by two-sided log rank test. d Left panel: Bubble plot showing the KEGG pathway enrichment (two-sided Fisher’s exact test) of each proteomic subtype. Right panel: Diagram illustrating the differentially expressed signatures and signaling cascades involved in G-I to G-IV. The little heatmap under each protein depicted the Z score of average protein abundance in each proteomic subtype. Red, upregulated proteins; blue, downregulated proteins. e The correlation of the extracellular matrix proteins and different drugs, including 5-FU, OXA, and DOC. The drug sensitivity (half maximal inhibitory concentration [IC50]) of gastric cancer cell lines was from the Cancer Dependency Map Project (DepMap). P-values are derived from two-sided Pearson’s correlation test. f Dose-response curves of MKN45 and MGC803 cell lines overexpressing THSD4 after 72-h treatments with DOC, OXA, and 5FU. Barplots showing the comparison of IC50 values in each group. Bars represent the mean of n = 3 independent experiments with error bars indicating SD. P-values are calculated using two-sided Student’s t test. g Boxplot showing the differential expression of ECM proteins validated by PRM assay (two-sided Wilcoxon rank-sum test, n = 10 biologically independent samples per group). Boxplots show median (central line), upper and lower quartiles (box limits), 1.5× interquartile range (whiskers). Source data are provided as a Source Data file.