A Machine Learning-Based Genome Mining Approach Reveals Unprecedented Biarylitide Diversity

Area of Science:

• Microbiology
• Biochemistry
• Bioinformatics

Background:

• Biarylitides are bacterial ribosomally synthesized and post-translationally modified peptides (RiPPs) characterized by biaryl bridges formed by cytochrome P450 enzymes.
• Their precursor peptides are encoded by small genes, making them difficult to detect with standard genome mining techniques.
• Existing methods fail to capture the full biosynthetic diversity of biarylitides.

Purpose of the Study:

• To comprehensively chart the biosynthetic space of biarylitides using a machine learning approach.
• To identify novel biarylitide biosynthetic gene clusters (BGCs) and explore variations in precursor motifs and modifying enzymes.
• To experimentally investigate and elucidate the nature of cross-links in previously unstudied biaryl formations.

Main Methods:

• Repurposing a machine learning algorithm for genome mining of RiPPs.
• Analyzing variations in precursor motifs, P450 enzymes, and additional modifying enzymes.
• Experimental investigation of biaryl formation in core peptide motifs like YWH, YVH, and YWY.

Main Results:

• Identification of 277 biarylitide BGCs, significantly expanding the known diversity.
• Characterization of biaryl cross-links in previously uninvestigated peptide motifs.
• Demonstration of machine learning's efficacy in uncovering hidden RiPP biosynthetic potential.

Conclusions:

• The study substantially broadens the known diversity of biarylitide precursors and BGCs.
• It provides a foundation for the systematic exploration of other RiPP families.
• Machine learning offers a powerful tool for discovering novel RiPPs and their biosynthetic pathways.