DeepWAS: Multivariate genotype-phenotype associations by directly integrating regulatory information using deep learning
Problem
GWAS identify disease-associated genetic variants but do not directly link them to regulatory mechanisms, which are only inferred in post-hoc analyses.
Why it matters
Connecting variants to regulatory mechanisms helps explain the biology of diseases like multiple sclerosis and depression.
What the AI did
DeepWAS integrates deep learning predictions of variants' regulatory effects on cell type-specific chromatin features directly into a multivariate GWAS setting.
Results
Up to 40,000 regulatory SNPs associated with MS (4,888 cases/10,395 controls), MDD (1,475 cases/2,144 controls) and height (5,974 individuals); 43-61 regulatory SNPs identified (abstract truncated)
Impact
Provides a method coupling disease-associated variants to their regulatory impact on chromatin features.
Record
| Venue | PLoS Computational Biology |
|---|---|
| Published | 2020 |
| Work type | article |
| Authors | Janine Arloth, Gökçen Eraslan, Till F. M. Andlauer, Jade Martins, Stella Iurato, Brigitte Kühnel, Mélanie Waldenberger, Josef Frank et al. |
| Citations | 74 (OpenAlex, at import time) |
| References | 103 |
| Model or method | DeepWAS (deep learning-based method) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1371/journal.pcbi.1007616 |
| Open access | gold |
| Keywords | GWAS, deep learning, regulatory variants, multiple sclerosis, chromatin |
Importance rating
significant — thousands to millions affected
Couples disease-associated variants to regulatory mechanisms for MS, MDD, height; a genuinely new capability for functional genomics, with large cohorts given.