Abstract
In silico ADMET (absorption, distribution, metabolism, excretion, and toxicity) models are important tools in combating late-stage attrition in the drug discovery process. This work shows how ADMET models can be combined to tailor predictions depending on one's needs. We demonstrate how the judicious use of data and considered combination of predictions can produce models that provide truly useful answers. This approach is illustrated with the prediction of hERG channel blocking and cytochrome P450 2D6 inhibition, where combination of two predictive models (with >80% of compounds correctly predicted) resulted in models with even better predictive values (with >90% of compounds correctly predicted for those classes of interest).
MeSH terms
-
Bayes Theorem
-
Calcium Channel Blockers / chemistry
-
Cation Transport Proteins / chemistry
-
Cytochrome P-450 Enzyme Inhibitors
-
Cytochrome P-450 Enzyme System / chemistry
-
Cytochrome P450 Family 2
-
Drug Design*
-
Enzyme Inhibitors / chemistry
-
Ether-A-Go-Go Potassium Channels
-
Humans
-
Neural Networks, Computer
-
Phenethylamines / chemistry
-
Potassium Channels, Voltage-Gated / chemistry
-
Quantitative Structure-Activity Relationship*
-
Steroid 21-Hydroxylase / antagonists & inhibitors
-
Steroid 21-Hydroxylase / chemistry
-
Sulfonamides / chemistry
Substances
-
Calcium Channel Blockers
-
Cation Transport Proteins
-
Cytochrome P-450 Enzyme Inhibitors
-
Enzyme Inhibitors
-
Ether-A-Go-Go Potassium Channels
-
KCNH6 protein, human
-
Phenethylamines
-
Potassium Channels, Voltage-Gated
-
Sulfonamides
-
Cytochrome P-450 Enzyme System
-
Cyp2c6v1 protein, rat
-
Cytochrome P450 Family 2
-
Steroid 21-Hydroxylase
-
dofetilide