Marine microbes are prolific yet underexplored producers of bioactive compounds and the source of new anti-cancer and antimicrobial drugs. A key challenge is mining microbial chemistry at scale to identify and prioritize novel compounds. New ‘omics techniques have created unprecedented amounts of data on the chemistry and genomes of microbes. However, this data remains difficult to interpret. Current computational methods to link biosynthetic gene clusters (BGCs) to their products largely operate in low-data regimes and cannot reliably identify metabolites from genomes. Here, we will use cutting-edge machine learning techniques to mine multi-omics data.  This project will lay the foundations towards building a genomic predictor for chemical potential using protein language models and novel BGC representations and using transformer-based models for metabolomics data and multi-modal contrastive learning to directly link BGCs with chemical features. Our goal is to create new tools to explore and mine ‘omics data to expedite the discovery of new microbial chemistry and therapeutics. This project will be in collaboration with Prof. Ben Sanchez-Lengeling. 

 

The SUDS Scholar will analyze genomics and metabolomics data from a collection of 150 marine bacteria. The goal is to create a pipeline to make ‘omics data AI-ready, though data organization; setting up data schema and data manuals; and performing exploratory data analysis (EDA).