The Next Base Editors Will Be Engineered, Not Found

By Matt Nethery, Principal Data Scientist, Computational Biology

Naturally occurring base editors have expanded the reach of genetic medicine, but wild-type enzymes were shaped by evolution rather than engineered for therapeutic applications. Consequently, unengineered deaminases often suffer from limitations like unwanted bystander edits, off-target modifications, and constrained targetable windows. Overcoming these barriers requires moving away from simply searching nature's catalogue to actively constructing purpose-built enzymes using generative artificial intelligence and machine learning pipelines. Training computational models on massive sequence datasets enables the design of de novo deaminase architectures with sequence identities distinct from known natural enzymes. When coupled with active learning optimization in mammalian cells, these computational methods yield base editors with superior target specificity, minimal bystander activity, and expanded access to previously intractable disease-relevant genomic sites.

Read the full insights to explore how computational protein design is defining the next generation of precision gene editors.