AI classifier integrating taxonomic and functional microbial features in melanoma and NSCLC improves prediction of response to ICI
This study demonstrates the feasibility of predicting ICI response from baseline gut metagenomic profiles across heterogeneous melanoma, NSCLC and RCC cohorts. By combining taxonomic and functional features within a Leave-One-Dataset-Out framework, the approach achieved reproducible cross-cohort discrimination and outperformed previously reported results on several publicly available datasets. Together with clinical proof-of-concept studies showing that microbiome modulation can improve ICI outcomes, these findings support the rationale for developing microbiome-based therapies with specifically designed taxonomic and functional compositions.
Joël Doré, Liyang Diao, Elsa Schalck, Germain Pinganaud, Antoine Sabourin, Grégoire Noiret, Cyrielle Gasc, Lisa Derosa, Laurence Zitvogel, Anne-Claire Toffart, Dalil Hannani, Antoine Bonnefoy, Sheri Simmons, Emmanuel Prestat, AI classifier integrating taxonomic and functional microbial features in melanoma and NSCLC improves prediction of response to ICI, in Proceedings of the ESMO AI & Digital Oncology Congress 2026, November 2026.
Click here to access the poster.