Design of prime-editing guide RNAs with deep transfer learning
Problem
Designing effective prime-editing guide RNAs (pegRNAs) is difficult due to lack of accurate, broadly applicable efficiency prediction and design optimization methods.
Why it matters
Prime editing is a promising genome-editing tool, and pegRNA quality determines editing success for research and therapy.
What the AI did
OPED, an interpretable nucleotide language model using transfer learning, predicts pegRNA efficiency and optimizes pegRNA designs.
Results
pegRNAs with high OPED scores showed significantly increased editing efficiencies; OPED outperformed state-of-the-art approaches and installed ClinVar pathogenic variants in PE2, PE3/PE3b and ePE systems.
Impact
Validated on published datasets and experimentally across multiple prime-editing systems, enabling more efficient pegRNA design.
Record
| Venue | Nature Machine Intelligence |
|---|---|
| Published | 2023 |
| Work type | article |
| Authors | Feng Liu, Shuhong Huang, Jiongsong Hu, Xiaozhou Chen, Ziguo Song, Junguo Dong, Yao Liu, Xingxu Huang et al. |
| Citations | 36 (OpenAlex, at import time) |
| References | 46 |
| Model or method | OPED (nucleotide language model, transfer learning) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s42256-023-00739-w |
| Open access | hybrid |
| Keywords | prime editing, pegRNA, transfer learning, language model, ClinVar, genome editing |
Importance rating
significant — thousands to millions affected
Prime editors are called promising genome-editing tools; OPED enables efficient pegRNA design and installation of ClinVar pathogenic variants across PE systems, supporting research and potential therapy design.