How Can AI-Directed Protein Engineering Improve Genomic Medicines?

By Gavin Ellis, Senior Scientist, Translational Biology

Optimizing proteins for advanced gene editing requires navigating an immense biological landscape where natural enzymes often fall short of therapeutic needs. Combining protein language models, active learning, and automated mammalian profiling offers a systematic solution to engineer custom effectors with enhanced potency and precision. Training machine learning models on high-quality, mammalian cell-derived activity data allows computational platforms to navigate complex protein fitness landscapes efficiently, bypassing the limitations of traditional bacterial screening methods. Rapid design-build-test cycles systematically identify beneficial single and combination mutations, generating novel enzymes—such as large serine recombinases—that outperform wild-type variants found in nature.

Integrating artificial intelligence with automated wet-lab validation turns protein optimization into a fast, repeatable framework, providing drug developers with tailored tools to expand the reach of cell and gene therapies.