ProDomino is a new machine learning tool that predicts where two proteins can be fused without disrupting their function—an important step in designing biotech tools like base editors and protein switches. Protein fusions are common in both nature and lab applications but can be difficult to create due to folding issues. ProDomino was trained on a large dataset of natural domain insertions and achieved high accuracy (AUROC 0.84) when benchmarked on the bacterial protein AraC. The tool was used to design light- and drug-responsive protein fusions, including a Cas12a variant and a light-sensitive antibiotic resistance switch. ProDomino significantly sped up protein engineering, reducing design time from months to weeks.
More Reading
Rational engineering of allosteric protein switches by in silico prediction of domain insertion sites
https://doi.org/10.1101/2024.12.04.626757