Genome mining of tailoring enzymes from biosynthetic gene clusters for synthetic biology: A case study with fungal methyltransferases

Harnessing the potential of tailoring enzymes within fungal natural product (NP) biosynthetic gene clusters (BGCs) can significantly enhance NP diversity and production efficiency via artificially constructed microbial cell factories. To achieve this, an efficient genome mining method is crucial, especially since the functions of many putative enzymes in databases are unknown. As a test case, we aimed to identify methyltransferases (MTs) that modify a polyketide substrate without a known cognate MT. 16,748 putative MTs were annotated in 101,321 fungal BGCs and grouped into orthologous families. Three methods were explored to prioritize suitable enzymes. Among these, the machine learning method proved superior, with 11 out of 15 tested MTs successfully methylating the test substrate. This demonstrates the effectiveness of machine learning to mine tailoring enzymes that modify selected compounds, aiding synthetic biology in optimizing NP biosynthesis and facilitating the production of "unnatural products" for pharmaceutical or other bioindustrial applications.