Protocol to use protein language models predicting and following experimental validation of function-enhancing variants of thymine-N-glycosylase

Protein language models (PLMs) are machine learning tools trained to predict masked amino acids within protein sequences, offering opportunities to enhance protein function without prior knowledge of their specific roles. Here, we present a protocol for optimizing thymine-DNA-glycosylase (TDG) using PLMs. We describe steps for "zero-shot" enzyme optimization, construction of plasmids, double plasmid transfection, and high-throughput sequencing and data analysis. This protocol holds promise for streamlining the engineering of gene editing tools, delivering improved activity while minimizing the experimental workload. For complete details on the use and execution of this protocol, please refer to He et al.