Stanford University’s Gao Lab is using artificial intelligence to enhance the safety and effectiveness of cell and gene therapies. In a recent study published in Cell Systems, researchers utilized machine learning models to improve therapeutic protein design by reducing immune response risks.
Xiaojing Gao, the senior author and assistant professor at Stanford’s School of Engineering, explained, “In this paper, we raise the question: Why not design treatments that avoid immune reactions from the start?” The team focused on zinc fingers, proteins that regulate gene expression and naturally bind with human DNA. This approach offers an alternative to technologies like CRISPR, which may trigger immune responses due to their bacterial origin.
Eric Wolsberg, a PhD student and lead author of the paper, highlighted their progress in designing zinc finger DNA-binding domains with low predicted immunogenicity risk. To address challenges with unnatural junctions in engineered proteins, they employed MARIA—a machine learning algorithm developed by co-authors Binbin Chen and Ash Alizadeh—to predict immunogenicity and guide safer designs.
The team further refined zinc fingers using ESM-IF1, a protein language model trained on natural sequences. Gao noted that this method allowed for “smart, targeted changes” rather than random mutations. They ensured new designs passed both functionality and immunogenicity tests before proceeding.
Comparative tests showed AI-enhanced proteins improved human gene production significantly more than original versions. Gao stated that they have advanced zinc finger engineering while maintaining function and reducing immunogenicity. Future efforts aim to develop an end-to-end algorithm for designing human-compatible gene therapies.
This research received funding from several sources including the National Institutes of Health and various Stanford grants. Other contributors included Josh Tycko, Michael Bassik, Lacramioara Bintu among others associated with Stanford’s institutes.
For more information about Stanford science stories or inquiries contact Taylor Kubota at University Communications via tkubota@stanford.edu.
