Metabolic Forest: Predicting the Diverse Structures of Drug Metabolites

Abstract

Adverse drug metabolism often severely impacts patient morbidity and mortality. Unfortunately, drug metabolism experimental assays are costly, inefficient, and slow. Instead, computational modeling could rapidly flag potentially toxic molecules across thousands of candidates in the early stages of drug development. Most metabolism models focus on predicting sites of metabolism (SOMs): the specific substrate atoms targeted by metabolic enzymes. However, SOMs are merely a proxy for metabolic structures: knowledge of an SOM does not explicitly provide the actual metabolite structure. Without an explicit metabolite structure, computational systems cannot evaluate the new molecule's properties. For example, the metabolite's reactivity cannot be automatically predicted, a crucial limitation because reactive drug metabolites are a key driver of adverse drug reactions (ADRs). Additionally, further metabolic events cannot be forecast, even though the metabolic path of the majority of substrates includes two or more sequential steps. To overcome the myopia of the SOM paradigm, this study constructs a well-defined systemtermed the metabolic forestfor generating exact metabolite structures. We validate the metabolic forest with the substrate and product structures from a large, chemically diverse, literature-derived dataset of 20 736 records. The metabolic forest finds a pathway linking each substrate and product for 79.42% of these records. By performing a breadth-first search of depth two or three, we improve performance to 88.43 and 88.77%, respectively. The metabolic forest includes a specialized algorithm for producing accurate quinone structures, the most common type of reactive metabolite. To our knowledge, this quinone structure algorithm is the first of its kind, as the diverse mechanisms of quinone formation are difficult to systematically reproduce. We validate the metabolic forest on a previously published dataset of 576 quinone reactions, predicting their structures with a depth three performance of 91.84%. The metabolic forest accurately enumerates metabolite structures, enabling promising new directions such as joint metabolism and reactivity modeling.

Bioactivation of Isoxazole-Containing Bromodomain and Extra-Terminal Domain (BET) Inhibitors

“…We built a graph database with Neo4j to connect molecules with observed direct and downstream metabolites up to three metabolic steps away. We queried the graph database to find conjugated electrophile species involving an isoxazole ring by applying our previously developed structure inference model, the Metabolic Forest [11], to identify metabolite pairs formed between the isoxazole-containing molecule and direct or downstream metabolites. For identified molecule pairs, the upstream isoxazole-containing molecule was annotated as forming a conjugated electrophile.…”

“…We calculated the accuracy of the reactivity model based on the ROC-AUC score using the highest atom score for each molecule and proper identification of the atom reactive toward glutathione, as recorded in AMD. Lastly, we used the Metabolic Forest model to predict glutathione adduct structures based on the glutathione rule within its conjugation rule set [11].…”

“…Importantly, the quinone model is capable of predicting the formation of other reactive metabolites within the broader class of conjugated electrophiles, despite its name. Second, we recently 2 of 19 reported a novel approach for predicting the structures of metabolites arising from bioactivations [11]. The information could facilitate their experimental identification as well as the construction of possible metabolic pathways.…”

Modeling the Bioactivation and Subsequent Reactivity of Drugs

“…50 Since the reaction type is accounted for during prediction, we can infer the structure of the resultant product using a structure inference model. 51 Next, to each inferred structure we apply a previously developed model for predicting reactivity to protein or GSH. Once we have predictions regarding metabolism of the input molecule and reactivity of its inferred metabolites, we can use these predictions to train a deep neural network that will predict bioactivation at both the molecule-and pathway-level.…”

MetaClass, a Comprehensive Classification System for Predicting the Occurrence of Metabolic Reactions Based on the MetaQSAR Database

“…In general, artificial intelligence (AI) methods can find at least three major applications focused on drug metabolism [9] since they can predict: (1) the potential sites of metabolism (regardless of the involved reactions) [10], (2) the metabolic reaction(s) that a given substrate undergoes [11] and (3) the formed metabolites [12]. These three applica-tions can be also seen as the progressive steps of an ideal workflow that allows the comprehensive prediction of the entire metabolic fate of a given compound [13].…”