Bacteria in the human gut have the ability to activate, inactivate, and reactivate drugs with both intended and unintended effects. For example, the drug digoxin is reduced to the inactive metabolite dihydrodigoxin by the gut Actinobacterium E. lenta, and patients colonized with high levels of drug metabolizing strains may have limited response to the drug. Understanding the complete space of drugs that are metabolized by the human gut microbiome is critical for predicting bacteria-drug relationships and their effects on individual patient response. Discovery and validation of drug metabolism via bacterial enzymes has yielded >50 drugs after nearly a century of experimental research. However, there are limited computational tools for screening drugs for potential metabolism by the gut microbiome. We developed a pipeline for comparing and characterizing chemical transformations using continuous vector representations of molecular structure learned using unsupervised representation learning. We applied this pipeline to chemical reaction data from MetaCyc to characterize the utility of vector representations for chemical reaction transformations. After clustering molecular and reaction vectors, we performed enrichment analyses and queries to characterize the space. We detected enriched enzyme names, Gene Ontology terms, and Enzyme Consortium (EC) classes within reaction clusters. In addition, we queried reactions against drug-metabolite transformations known to be metabolized by the human gut microbiome. The top results for these known drug transformations contained similar substructure modifications to the original drug pair. This work enables high throughput screening of drugs and their resulting metabolites against chemical reactions common to gut bacteria.

Background Enzymatic and chemical reactions are key for understanding biological processes in cells. Curated databases of chemical reactions exist but these databases struggle to keep up with the exponential growth of the biomedical literature. Conventional text mining pipelines provide tools to automatically extract entities and relationships from the scientific literature, and partially replace expert curation, but such machine learning frameworks often require a large amount of labeled training data and thus lack scalability for both larger document corpora and new relationship types. Results We developed an application of Snorkel, a weakly supervised learning framework, for extracting chemical reaction relationships from biomedical literature abstracts. For this work, we defined a chemical reaction relationship as the transformation of chemical A to chemical B. We built and evaluated our system on small annotated sets of chemical reaction relationships from two corpora: curated bacteria-related abstracts from the MetaCyc database (MetaCyc_Corpus) and a more general set of abstracts annotated with MeSH (Medical Subject Headings) term Bacteria (Bacteria_Corpus; a superset of MetaCyc_Corpus). For the MetaCyc_Corpus, we obtained 84% precision and 41% recall (55% F1 score). Extending to the more general Bacteria_Corpus decreased precision to 62% with only a four-point drop in recall to 37% (46% F1 score). Overall, the Bacteria_Corpus contained two orders of magnitude more candidate chemical reaction relationships (nine million candidates vs 68,0000 candidates) and had a larger class imbalance (2.5% positives vs 5% positives) as compared to the MetaCyc_Corpus. In total, we extracted 6871 chemical reaction relationships from nine million candidates in the Bacteria_Corpus. Conclusions With this work, we built a database of chemical reaction relationships from almost 900,000 scientific abstracts without a large training set of labeled annotations. Further, we showed the generalizability of our initial application built on MetaCyc documents enriched with chemical reactions to a general set of articles related to bacteria.