Skip to main content
. 2026 Feb 9;14(3):e02151-25. doi: 10.1128/spectrum.02151-25

Fig 4.

Integrated analysis of immune cell profiles and microbial signatures in stomach cancer correlating with ICB response. Network connections between microbes and immune genes support combined predictive models that outperform individual domain approaches.

A comprehensive integration of immune cell profiles and microbial signatures uncovers potential predictors of ICB responsiveness in STAD. (a and b) Distribution of ICB response across INC and MC subtypes in STAD. (c) Boxplot illustrates the differences in ICB response scores among INC subtypes. ns indicates no significant difference, * indicates P < 0.05, ** indicates P < 0.01, *** indicates P < 0.001 and **** indicates P < 0.0001. (d) Sankey diagram depicting the distributional relationships among INC subtypes, MC subtypes, and ICB response status R vs. NR. (e) UpSet plot showing the overlap of gene features selected across cross-validation folds, highlighting the consistency and variation in immune-related gene selection. (f and g) Heatmap and network diagram illustrating the relationships between the top 10 microbial genera (based on relative abundance) and the top 11 immune-related genes selected by intersecting features across cross-validation folds. (h) Bar plots depicting the AUC values of gene-only, microbe-only, and combined models across six machine learning algorithms: logistic regression, decision tree, naïve Bayes, random forest, SVM, and XGBoost.