About the Article
Article Files
[This article belongs to Volume - 70, Issue - 12]
Published on : 2025-12-16 23:19:41
Article Code: AMJ-16-12-2025-12361
Title : Gut Microbiome Signatures Predicting Immune Checkpoint Inhibitor Response in Metastatic Melanoma. A Multi-Omics Machine Learning Analysis
Author(s) : Elena K. Volkov, MD, PhD, Kenji Nakamura, PhD, Mohammed Al-Rashid, MBBS, PhD
Abstract :
Background: Response rates to immune checkpoint inhibitors (ICIs) in metastatic melanoma remain heterogeneous
(40-60%), with gut microbiome composition emerging as a key modulator of anti-tumor immunity. However,
predictive microbiome signatures lack validation across diverse cohorts and omic layers.
Objectives: To develop and validate a multi-omics model integrating gut microbiome, metabolomic, and clinical data
to predict ICI response in metastatic melanoma patients.
Methods: We analyzed 456 patients with metastatic melanoma from six academic centers (2019-2023). Shotgun
metagenomic sequencing of fecal samples was performed pre-treatment, alongside plasma metabolomics (LC-MS)
and immune profiling (CyTOF). Machine learning models (XGBoost, random forest) were trained to predict objective
response (RECIST 1.1) at 6 months. External validation was performed on an independent cohort (n=112).
Results: A 12-feature microbiome-metabolome signature achieved AUC 0.91 (95% CI: 0.88-0.94) in predicting ICI
response. Key predictors included Akkermansia muciniphila abundance (OR=2.34, p<0.001), short-chain fatty acid
production capacity, and baseline CD8+ T-cell exhaustion markers. The model stratified patients into high-response
(ORR 72.3%) and low-response (ORR 18.7%) groups (p<0.001). External validation AUC was 0.87.
Conclusions: Multi-omics integration of gut microbiome and metabolomic data robustly predicts ICI response in
melanoma, offering a non-invasive tool for treatment stratification. This framework may extend to other cancer types
where microbiome-immune crosstalk is operative.