Back

An AI-designed adenine base editor

Yuan, Y.; Chen, Y.; Liu, R.; Que, G.; Yuan, Y.; Li, G.

2024-04-29 bioengineering
10.1101/2024.04.28.591233 bioRxiv
Show abstract

Adenine base editors (ABEs) allow the efficient programmable conversion of adenine to guanine without causing DNA double strand breaks. Previous ABEs were generated by multiple rounds of directed evolution or derived by rational design based on the evolved ones. Although powerful, these methods search the local space for ABEs optimizations. Artificial intelligence (AI) based methods have the ability to efficiently explore much larger protein space for protein design. But currently there is no AI-designed ABE with wet experimental validation. Here, we demonstrate the first successful AI-designed ABE, which is named ABE10. ABE10 includes an AI-designed adenine deaminase enzyme fused with SpCas9n. The sequence identity between AI-designed enzyme and other publicly accessible variants is as low as 65.3%. ABE10 shows improved editing efficiency compared to current state-of-the-art ABE8 at multiple human genome sites tested. ABE10 also shows low off-target editing rate and reduced cytosine bystander effect. Our work demonstrates new direction for optimization of gene editing tools.

Matching journals

The top 10 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.