Optimized CRISPR guide RNA design for two high-fidelity Cas9 variants by deep learning
Problem
High-fidelity Cas9 variants are valuable for genome editing but their use is hampered by lack of knowledge governing guide RNA activity.
Why it matters
Reliable guide RNA design determines whether these highly specific genome-editing tools can be widely applied.
What the AI did
A genome-scale screen measured indel rates for over 50,000 gRNAs per nuclease; models built on 1,031 features showed an RNN combined with biological features best predicted activity, and an online design tool DeepHF was released.
Results
Model outperformed other models and popular gRNA design tools; indel rates for >50,000 gRNAs per nuclease covering ~20,000 genes.
Impact
Provides a freely accessible web tool and activity data for designing guides for three Cas9 nucleases.
Record
| Venue | Nature Communications |
|---|---|
| Published | 2019 |
| Work type | article |
| Authors | Daqi Wang, Chengdong Zhang, Bei Wang, Bin Li, Qiang Wang, Dong Liu, Hongyan Wang, Yan Heng Zhou et al. |
| Citations | 330 (OpenAlex, at import time) |
| References | 71 |
| Model or method | RNN (DeepHF tool) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41467-019-12281-8 |
| Open access | gold |
| Keywords | CRISPR, Cas9, guide RNA, deep learning, RNN, genome editing |
Importance rating
significant — thousands to millions affected
Reliable gRNA design determines whether high-specificity Cas9 editing tools can be widely applied; broad enabling capability for genome editing but no scale stated.