A Tool for Predicting the Dark Side of Enzymes

Linfeng Liu, Soha Hassoun

Abstract: Early studies considered enzymes as specific catalysts following the classical rule, "one enzyme, one activity". Evidence, however, suggests that some enzymes have a promiscuous (dark) side and act on substrates other than the preferred substrate [1]. The ability to predict products due to enzyme promiscuity can yield beneficial results across many areas in biotechnology, including discovering novel pathways to synthesize target molecules, and identifying derivatives in of ingested foreign chemicals. In our group, we have recently used enzyme promiscuity to annotate measurements collected via untargeted metabolomics from a biological sample. We were able to experimentally validate our finding, thus providing the first demonstration of predicting and verifying in vivo enzyme promiscuity [2]. To this end, it would be advantageous to have a reliable tool for predicting metabolic derivatives due to promiscuous enzymes. In this work, we create a tool, PROXIMAL II, that generalizes the methodology in PROXIMAL [3]. PROXIMAL II allows the end user to specify a query molecules and a set of reactions and then predict potential derivative products. We construct biotransformation operators lookup table using Reaction Center, Difference Region, and Matched Region (RDM) patterns [4] and molecular substructures. The RDM patterns specify local regions of similarities/differences for reactant-product pairs based on chemical structure [5]. For each potential R pattern matched for the query molecule, a set of transformations are looked up in the table and applied to the query molecule, thus generating a set of possible transformed molecules. PROXIMAL II will provide a variety of predicting capabilities, based on EC number, KEGG reaction number, or organism.