Application of Protein Language Models to low-N Engineering of metalloproteinase inhibitors

Targeting metalloproteinases (MPs) has a great potential in developing novel therapeutics as these proteases are known to be responsible in several diseases. Directed evolution was previously used to engineer binding affinity and selectivity of natural inhibitors of MPs, tissue inhibitors of metalloproteinases (TIMPs). However, directed evolution is a time-consuming and laborious process. Machine learning approaches can facilitate directed evolution by narrowing the search for finding optimal mutations. This work investigates the application of state-of-theart pre-trained protein language models in screening MMP inhibitor variants. Features extracted from several protein language models for a low-N experimental library of MMP variants are visualized and compared with regards to their performance in screening strong inhibitors.