New Base Editing Tools Developed Using AI-Predicted Protein Structure Clustering

A group led by Gao Caixia from the Institute of Genetics and Developmental Biology of the Chinese Academy of Sciences used AI-assisted methods to uncover unique deaminase proteins through structural prediction and classification. This approach was developed to help discover and create desired genetic traits in plants but also has applications in phylogenetics, metagenomics, protein engineering and evolution, and genome editing.

The research, published in Cell, sheds light on the discovery of novel proteins and the application of engineered enzymes. Traditionally, efforts to identify new proteins have relied on amino acid sequences, which lack a robust link between protein structure and function.

Base editing, a precise genome editing technology, has the potential to revolutionize molecular crop breeding by introducing desired traits into elite germplasm. Several deaminases have expanded the capabilities of cytosine base editing. However, existing sequence-based methods for identifying proteins as base editors have limitations in editing specific DNA sequences or species.

In this study, the researchers employed AlphaFold2, an AI-based protein structure prediction tool, to predict the structures of proteins within the deaminase family. By clustering and analyzing deaminases based on structural similarities, they discovered five new clusters of deaminases with cytidine deamination activity for DNA base editing.

Through further classification and protein engineering efforts, they reclassified a group of cytidine deaminases known as SCP1.201, previously believed to act on double-stranded DNA, and demonstrated their primary deamination activity on single-stranded DNA. Through subsequent protein profiling and engineering, they developed a collection of new DNA base editors with remarkable features. These deaminases exhibit higher efficiency, lower off-target editing events, editing at different preferred sequence motifs, and a significantly smaller size.

The researchers emphasized that the suite of base editors developed in this study enables tailored applications for therapeutic or agricultural breeding purposes. They achieved the smallest single-strand specific cytidine deaminase, allowing for the packaging of an efficient cytosine base editor in a single adeno-associated virus.