Active Learning-Guided Optimization Of Large Gene Insertion Effectors In Mammalian Cells​

By Gavin Ellis, Senior Scientist, Translational Biology

Optimizing large gene insertion effectors in mammalian cells requires navigating complex fitness landscapes where traditional, unguided screens often fall short. By combining predictive machine learning models with iterative active learning rounds, developers can systematically enhance the potency of novel large serine recombinases. Testing single-amino-acid variants establishes baseline activity data, yielding early potency gains over wildtype proteins. Evaluating pairwise double-mutant combinations captures critical epistatic interactions, unlocking higher-order mutations and further boosting insertion efficiency. This closed-loop design framework bypasses the need for known protein structures, providing an efficient path to refine targeted gene integration effectors for advanced therapeutic applications.

Explore the poster insights to learn how active learning-guided protein engineering accelerates effector optimization for complex genomic medicines.