CYP Inhibition

CYP Inhibition

Inhibition of CYPs is important for understanding drug toxicity and drug-drug interactions. The Metabolism Module includes generic inhibition models for nine recombinant CYPs: 1A2, 2A6, 2B6, 2C8, 2C9, 2C19, 2D6, 2E1, and 3A4.

The structures and biological activity data of corresponding CYP isoforms were extracted from the National Center for Biotechnology Information (NCBI) PubChem database AID 1851. According to the Pubchem Bioassay database, a compound was classified as a CYP Inhibitor if the AC50 value (compound concentration that leads to 50% of the activity of an inhibition control) is ≤10 μM and as a Non-inhibitor if the AC50 value is >57 μM. If the AC50 value falls between 10 and 57 μM, the compound is defined as inconclusive. Only the compounds defined as actives and inactives were considered for further analysis. The inconclusive compounds were excluded to prevent any uncertainty during model building.

These datasets were appended to our in-house collection of CYP inhibition data. All the datasets that were curated for regression models were used for building the classification models. Compounds with Ki less than 10 μM were labeled as “Inhibitors” and those with Ki greater than 30 μM were labeled as “Non-inhibitors”.

The combined datasets were pre-processed to eliminate entries containing non-covalent complexes, inorganic compounds and mixtures. Moreover, the entries containing salts were converted to their corresponding acids or bases, and water molecules were removed from hydrates. The duplicate entries were removed using the following criteria. If the duplicates had identical endpoints, then one entry was retained, and others eliminated.

For entries with conflicting end-points:

• If one of the duplicate entries contained an in-house (hand-curated) entry, the in-house data point was retained while that from the PubChem dataset was eliminated.

• If the duplicate entries originated from the PubChem dataset, both entries (with conflicting end-points) were removed from the final dataset.

The table below lists the statistics for each model.  Note that concordance, sensitivity and specificity are above 80% for the training and test sets of all models.  The CYP inhibition classification models are used as a filter for the CYP Ki models: i.e., if the classification model indicates that the compound does not inhibit that isoform, then the Ki model's prediction will be "NonInhibitor".