Multimodal deep learning applied to classify healthy and disease states of human microbiome
Problem
Classifying healthy versus disease states from human microbiome metagenomic data, where existing tools lacked accuracy needed for diagnostics.
Why it matters
Accurate microbiome-based classification could enable earlier, less invasive disease detection and diagnosis.
What the AI did
MDL4Microbiome combines taxonomic profiles, genome-level relative abundance, and metabolic functional features in a multimodal deep learning classifier.
Results
Classification accuracies of 0.98, 0.76, ... (abstract truncated)
Impact
Showed multimodal feature combination improves microbiome disease-state classification on tested datasets.
Record
| Venue | Scientific Reports |
|---|---|
| Published | 2022 |
| Work type | article |
| Authors | Seung Jae Lee, Mina Rho |
| Citations | 51 (OpenAlex, at import time) |
| References | 50 |
| Model or method | MDL4Microbiome (multimodal deep learning) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41598-022-04773-3 |
| Open access | gold |
| Keywords | microbiome, metagenomics, multimodal deep learning, disease classification, diagnostics |
Importance rating
significant — thousands to millions affected
Microbiome-based diagnosis of IBD, type 2 diabetes, liver cirrhosis, and colorectal cancer could enable less invasive detection of major diseases; abstract shows high accuracies but no deployment or scale figures.