Faecal microbiome-based machine learning for multi-class disease diagnosis
Problem
Shared microbial signatures across diseases make accurate single-disease diagnosis from faecal microbiome data difficult, and non-invasive diagnostic approaches are needed.
Why it matters
Non-invasive screening for diseases like colorectal cancer and IBD could enable earlier diagnosis and better outcomes.
What the AI did
A machine-learning multi-class model was trained on faecal metagenomic data from 2,320 individuals (325 microbial species, nine phenotypes) to predict multiple diseases simultaneously.
Results
AUROC 0.90-0.99 (IQR 0.91-0.94) and sensitivity 0.81-0.95 on an independent test set
Impact
Demonstrates non-invasive multi-disease prediction from stool metagenomics, validated on an independent test set.
Record
| Venue | Nature Communications |
|---|---|
| Published | 2022 |
| Work type | article |
| Authors | Qi Su, Qin Liu, Raphaela Iris Lau, Jingwan Zhang, Zhilu Xu, Yun Kit Yeoh, Thomas Wai Hong Leung, Whitney Tang et al. |
| Citations | 133 (OpenAlex, at import time) |
| References | 37 |
| Model or method | not stated in source |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41467-022-34405-3 |
| Open access | gold |
| Keywords | machine learning, faecal microbiome, multi-class diagnosis, metagenomics, AUROC |
Importance rating
significant — thousands to millions affected
Abstract targets non-invasive diagnosis of major human diseases including colorectal cancer, IBD, obesity, and cardiovascular disease, with strong multi-disease performance and stated potential clinical application in diagnostics and treatment monitoring.