TaxoMFM improves gut microbiome analysis using taxonomic hierarchy
TaxoMFM: A Taxonomy-Aware Foundation Model for the Human Gut Microbiome
Machine Learning
Summary
Gut microbiome studies often treat bacteria as unrelated groups, missing the natural family tree of microbes. The authors designed TaxoMFM, a model that understands these family connections to better learn patterns in gut bacteria data. They created a large dataset and showed that TaxoMFM predicts diseases and traits related to gut health with good accuracy, even for groups it was never trained on. The model also keeps consistent results when the bacteria database changes, making it more reliable for research and health applications.
What this means in practice
- •For biotechnology companies: Develop better diagnostic tools for gut-related diseases using richer microbial community models that include taxonomic relationships.$Commercial implications: Enables creation of more accurate microbiome-based diagnostics for diseases like colorectal cancer and diabetes by exploiting microbial family structures.
- •For bioinformatics teams: Integrate taxonomy-aware models to improve predictions of host health traits and disease status from microbiome sequencing data.
Authors
Haochen Xue, Anastasia Godneva, Yulong Li, Qi Cao, Zhaodong Wu, Kaiyu Liang, Jionglong Su, Eran Segal, Imran Razzak
Abstract
Existing gut microbiome foundation models typically use flat taxon representations, limiting their ability to exploit known taxonomic hierarchies. We introduce TaxoMFM, a microbiome foundation model that integrates taxonomic structure into self-supervised learning. TaxoMFM jointly learns abundance patterns across taxonomic ranks and relative proportions within groups through taxonomy-aware encoding and hierarchical compositional reconstruction. We further construct TaxoGut-HPP, comprising 100,806 gut microbiome profiles from 231 source cohorts and covering 40 classification and 8 regression tasks. TaxoMFM achieves competitive performance across disease, gut health, host trait, and liver ultrasound phenotype prediction. In cross-cohort zero-shot transfer, where target cohorts are excluded from both pretraining and downstream adaptation, TaxoMFM achieves AUROCs of 0.838, 0.736, and 0.722 for colorectal cancer, type 2 diabetes, and inflammatory bowel disease, respectively. Further analyses show that TaxoMFM recovers community changes across taxonomic scales, while preserving community information and producing consistent representations under database updates and species-vocabulary shifts. Our code is available at https://anonymous.4open.science/r/TaxoMFM-F16D/.