ContactSeek gene editing accuracy improves with AlphaFold3 modeling
Researchers used AlphaFold3 to redesign base editors, cutting unwanted DNA edits by about 83% in one test reported in Nature.
By Lucas Ferreira · Science & Environment Writer
3 min read
ContactSeek gene editing accuracy improved in tests after researchers used an AlphaFold3-based system to identify small changes that made DNA editors less prone to unintended edits. The work, reported in Nature, matters because gene-editing tools can miss their intended target and alter other DNA or RNA sequences.
Haowei Meng and colleagues described the platform as a way to redesign genome editors by studying how the editing protein, its guide RNA and DNA targets are likely to interact. Hoi Yee Chu and Alan S.L. Wong of the University of Hong Kong discussed the study’s significance in a related News and Views article in the same journal.
Gene editing lets scientists insert, delete, alter or replace DNA bases in living organisms. It is used in research on disease-causing mutations and crop improvement, but tools such as CRISPR can sometimes act at sites other than the intended place in the genome.
What is ContactSeek?
ContactSeek is an AI-guided framework that uses AlphaFold3 to predict contact probabilities between parts of a gene-editing protein and the molecules it works with. In plain terms, it looks for likely points of close interaction that may explain why an editor binds correctly in some cases and incorrectly in others.
The Nature study says the system compares predicted contacts at intended DNA targets with contacts at off-target sites. That comparison can point researchers to parts of the editor that help drive unwanted edits, giving them candidates for small redesigns.
The approach differs from methods that focus on predicted three-dimensional structures of the editor and DNA. The researchers wrote that contact probability was a more sensitive signal than predicted structure for detecting differences in molecular interactions.
How much did the redesigned editor reduce off-target edits?
The team tested ContactSeek on an adenine base editor, a gene-editing tool used to change single DNA letters. After the AI analysis suggested small modifications, the redesigned editor reduced genome-wide off-target editing by about 83% for one of the guide RNAs tested, according to the study.
The researchers also applied the platform to a Cas12a-based cytosine base editor and reported improved accuracy there as well. Base editors are designed to make specific single-letter DNA changes without cutting both strands of the DNA helix in the way associated with some CRISPR systems.
The findings do not mean the system is ready for medical use. The researchers said more testing is needed across different cell types and genome targets before clinical applications can be considered.
Even with that limit, the authors said they expect ContactSeek could be used across a broad set of DNA- and RNA-editing tools. The study frames AI contact modeling as a practical way to find modest protein changes that may make gene editors more selective.
This story draws on original reporting from Phys.org.