Using Machine Learning to Improve Gene Therapy Design

Stanford University researchers have developed a machine learning-based approach to engineer safer and more effective proteins for use in cell and gene therapies. The work centers on using zinc fingers – tiny, naturally occurring human proteins involved in gene regulation.

Xiaojing Gao, senior author of the paper and assistant professor of chemical engineering at Stanford explains: “In this paper, we raise the question: Why not design treatments that avoid immune reactions from the start? With advances in computational tools, we’re now trying to predict which changes to a protein could trigger an immune response, and only move forward with designs that are less likely to be rejected by the body.”

The team wanted to redesign zinc fingers so that they could target specific genomic sites associated with disease, without setting off the immune system. To do this, they used three machine learning tools.

First, an algorithm helped identify zinc finger combinations that could bind to chosen DNA sequences. However, linking these proteins together to extend their reach introduced unnatural junctions that could be potential red flags for immune cells. To address this, the team used MARIA, a model originally developed to identify immunogenic regions for cancer vaccine design, in reverse: to predict and avoid sequences likely to trigger immune reactions.

Finally, a protein language model called ESM-IF1 suggested small, targeted genetic tweaks to enhance protein functionality while preserving low immunogenicity.

The AI-optimized zinc fingers were tested in lab experiments, showing up to six-fold improvements in gene activation compared to unmodified proteins. Gao describes the work as taking “the engineering of zinc fingers to a hitherto unvisited place.”

Eric Wolsberg, lead author of the paper, added: “The most significant part of our work is our progress in designing zinc finger DNA-binding domains that can target any genomic site we choose while maintaining a low predicted risk of triggering immune responses.”