Single-cell multiome of the human retina and deep learning nominate causal variants in complex eye diseases
Problem
Most genetic variants linked to eye diseases are noncoding, making their function and target genes hard to interpret.
Why it matters
Eye diseases like macular degeneration and glaucoma are major causes of vision loss worldwide.
What the AI did
Deep learning models, integrated with a single-cell retina atlas and enhancer connectome data, predicted noncoding SNPs with causal roles and their impact on transcription factor binding.
Results
Atlas of more than 50,000 cells; nominated pathogenic SNP-target gene interactions for multiple vision disorders.
Impact
Provides a peer-reviewed resource nominating causal variants and target genes, though causality is nominated rather than validated.
Record
| Venue | Cell Genomics |
|---|---|
| Published | 2022 |
| Work type | article |
| Authors | Sean K. Wang, Surag Nair, Rui Li, Katerina Kraft, Anusri Pampari, Aman Patel, Joyce B. Kang, Christy Luong et al. |
| Citations | 88 (OpenAlex, at import time) |
| References | 71 |
| Model or method | base-resolution deep learning models (unnamed) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1016/j.xgen.2022.100164 |
| Open access | gold |
| Keywords | retina, single-cell multiome, deep learning, GWAS, causal variants, eye disease |
Importance rating
significant — thousands to millions affected
Interprets noncoding variants for multiple vision disorders (AMD, glaucoma, diabetic retinopathy, myopia), nominating causal SNP-target gene pairs; significant but causality unvalidated and no disease-burden numbers given.