Predicting the sequence specificities of DNA- and RNA-binding proteins by deep learning
Problem
Predicting which DNA and RNA sequences are bound by proteins is needed to model gene regulation and identify causal disease variants.
Why it matters
Binding specificities underpin understanding of regulatory processes and finding disease-causing mutations.
What the AI did
A deep learning approach (DeepBind) was trained on experimental data to predict sequence specificities, packaged as an automatic software tool handling millions of sequences per experiment.
Results
Deep learning outperformed other state-of-the-art methods across diverse datasets and metrics, including training on in vitro and testing on in vivo data.
Impact
Highly cited (2869 citations) peer-reviewed tool with released software for mapping protein-binding specificities.
Record
| Venue | Nature Biotechnology |
|---|---|
| Published | 2015 |
| Work type | article |
| Authors | Babak Alipanahi, Andrew Delong, Matthew T. Weirauch, Brendan J. Frey |
| Citations | 2,869 (OpenAlex, at import time) |
| References | 49 |
| Model or method | DeepBind (deep learning) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/nbt.3300 |
| Open access | hybrid |
| Keywords | deep learning, DeepBind, DNA-binding, RNA-binding, sequence specificity, mutation map |
Importance rating
significant — thousands to millions affected
Abstract states sequence specificities are essential for modeling regulatory processes and identifying causal disease variants—foundational genomics capability; no scale stated.