2.7.3. Predicting Metabolites
The metabolites generated by enzymatic attack on a xenobiotic are more or less obvious for a given site of attack for some molecules, but not all. In particular, the metabolites formed initially can be unstable, breaking down or tautomerizing more or less readily to yield a more stable structure.
ADMET Predictor uses ANNE-based substrate classification and site of metabolism (SOM) models to pick out atoms subject to CYP and AOX oxidation and then applies a set of transformations to those atoms to generate the initial oxidation product as well as subsequent spontaneous and enzymatic transformations such as tautomerization and the oxidation of quinols.
Predicting the glucuronides produced by UGTs also depends on substrate classification models, but is done without the benefit of statistically-derived site models because the amount of good data required to build such models is not available. Instead we rely on an expert system - i.e., a hierarchy of substructure queries to identify the most likely glucuronidation site for each isoform and generate the expected metabolite from that.
Esterases are too indiscriminate in terms of what they bind to support classification models. Fortunately, they are relatively specific in terms of the kinds of substructures that they will attack. Here the expert system consists of a block of queries and transforms that are applied exhaustively.
Products for all three types of metabolism - oxidation (CYPs and AOX), glucuronidation (UGTs) and hydrolysis (esterases) - are generated similarly in either of two ways:
Graphical metabolic maps for individual molecules can be generated in MedChem Designer, which is accessible from the spreadsheet in several ways: by double clicking on a structure cell in the spreadsheet; by right-clicking on a row selection and then clicking on the Display in MedChem Designer option from the context menu that pops up; or via the keyboard shortcut Shift+D.
MedChem Designer only generates metabolite maps for one compound at a time and only for one metabolic cycle at a time, however. Going deeper tends to produce metabolic maps that are too cluttered to be useful, even when display of all intermediates is suppressed. Getting to the second level of metabolism - i.e., to see metabolites of a metabolite - requires selecting a specific metabolite, then using the 
Checking the metabolism of metabolites at some point is important because many common kinds of metabolites are themselves subject to further metabolism. Aldehydes, for example, are typically oxidized to the corresponding carboxylic acids too rapidly to exert much of a biological effect. The fact that they are unlikely to accumulate in vivo is an important consideration in drug development, since stable aldehydes tend to be mutagenic.
Metabolites can also be generated directly from the spreadsheet. Doing so allows multiple rows to be processed simultaneously, generating a new spreadsheet or CTK file containing the predicted metabolites along with details about where each came from.