Fig. 1: An ML-guided, cell-free enzyme engineering platform.
From: Accelerated enzyme engineering by machine-learning guided cell-free expression
Schematic shows how a design-build-test-learn workflow is applied to rapidly map sequence-function landscapes. Putative residues directing enzyme catalysis are rationally selected based on structural insights, evolutionary trends, and computational tools (e.g., ROSETTA71, EVmutation47, PROSS55) (design). Site saturation mutagenesis and cell-free gene expression are carried out in less than 24 h to generate sequence-defined libraries (build). The libraries can then be screened for desirable protein fitness metrics (test). Information from the test phase, including failures, is used to identify functionally important amino acid residues that feedback on iterative designs, as well as fit ML models (learn).
