Discovery and protein language model-guided design of hyperactive transposases

The diversity and biochemical potential of the PiggyBac transposase gene insertion system remains largely unexplored. Using a eukaryotic transposon mining pipeline, we expand the explored diversity by two orders of magnitude and experimentally validate a subset of highly divergent PiggyBac sequences. Fine-tuning a protein language model to further expand PiggyBac sequence space discovers transposases with improved activity and that are compatible with T cell engineering and Cas9-directed transposase-assisted integration.