Deep learning-based enhancement of epigenomics data with AtacWorks
Problem
Low cell count, low-coverage or noisy ATAC-seq data limits detection of active regulatory regions in the genome.
Why it matters
Chromatin accessibility measurements constrain biological discovery when samples are small or noisy.
What the AI did
AtacWorks, a deep learning toolkit, denoises sequencing coverage and identifies regulatory peaks at base-pair resolution from low-quality ATAC-seq data.
Results
Results on par with conventional methods using ~10 times as many cells; peaks detected at base-pair resolution; generalizes to unseen cell types and platforms.
Impact
Enables single-cell experiments with far fewer cells and supports new biological discoveries, per the abstract.
Record
| Venue | Nature Communications |
|---|---|
| Published | 2021 |
| Work type | article |
| Authors | Avantika Lal, Zachary Chiang, Nikolai Yakovenko, Fabiana M. Duarte, Johnny Israeli, Jason Daniel Buenrostro |
| Citations | 66 (OpenAlex, at import time) |
| References | 46 |
| Model or method | AtacWorks (deep learning toolkit) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41467-021-21765-5 |
| Open access | gold |
| Keywords | ATAC-seq, deep learning, chromatin accessibility, denoising, single-cell |
Importance rating
significant — thousands to millions affected
Abstract states ~10x fewer cells needed for single-cell ATAC-seq results and new discoveries in rare stem-cell subpopulations; broad benefit to genomics research capability.