Human disease prediction from microbiome data by multiple feature fusion and deep learning
Problem
Predicting human diseases from microbiome (metagenomic) data suffers information loss and severe overfitting because datasets are high-dimensional with few samples.
Why it matters
Reliable microbiome-based disease detection could enable earlier, non-invasive diagnosis and treatment.
What the AI did
MetaDR, a machine learning framework, fuses multiple feature types with deep learning to predict human diseases from metagenomic data while limiting overfitting.
Results
Competitive prediction performance with reduced running time; no quantitative metrics stated in source
Impact
Evidence supports a competitive prediction framework, but no concrete benchmark numbers are given.
Record
| Venue | iScience |
|---|---|
| Published | 2022 |
| Work type | article |
| Authors | Xingjian Chen, Zifan Zhu, Weitong Zhang, Yuchen Wang, Fuzhou Wang, Jianyi Yang, Ka‐Chun Wong |
| Citations | 30 (OpenAlex, at import time) |
| References | 51 |
| Model or method | MetaDR (deep learning-based framework) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1016/j.isci.2022.104081 |
| Open access | gold |
| Keywords | microbiome, deep learning, disease prediction, metagenomics, feature fusion |
Importance rating
real but narrower advance
Abstract says human disease prediction from microbiome data has broad implications in metagenomics; framework is competitive but no performance numbers, diseases, or population scale stated.