ADMET Prediction
Overview
In drug discovery, even pharmacophores with strong predictive potential can face late-stage failure due to poor ADMET properties, despite undergoing an extensive and resource-intensive development process. Early-stage ADMET profiling has therefore become a critical aspect of lead optimization. While many high-throughput in vitro models are available for this purpose, in silico approaches are increasingly favored for their cost-effectiveness and rapid prediction capabilities,eliminating the need for time-consuming and expensive laboratory experiments. Herein, using the AI-powered ADMET prediction module one can predict key pharmacokinetic and toxicity properties of small molecules using a single SMILES input. This task enables fast evaluation of a compound’s ADMET profile, including physicochemical parameters, absorption, distribution, metabolism, excretion, and toxicity.
Input Format
Input: Single valid SMILES string
Example: CC(=O)OC1=CC=CC=C1C(=O)O
Once submitted, the system returns a detailed prediction table for the molecule.
Prediction Categories
The output of this module is categorized as follows:
Physicochemical properties that are essential for assessing a molecule’s drug-likeness and pharmacokinetic behavior. These properties include molecular weight, logP (partition coefficient), topological polar surface area (TPSA), number of hydrogen bond donors and acceptors, and rotatable bonds.
Absorption-related properties are critical for determining how well a drug candidate can enter the bloodstream after administration. It evaluates parameters such as human intestinal absorption (HIA), Caco-2 permeability, P-glycoprotein (P-gp) substrate or inhibitor status, and bioavailability.
Distribution-related properties that determine how a drug candidate disperses throughout the body after absorption. These include the volume of distribution (VDss), which reflects the degree of drug penetration into tissues, and plasma protein binding (PPB), which influences the amount of free, active drug available in circulation. The tool also evaluates blood-brain barrier (BBB) permeability and central nervous system (CNS) penetration—key parameters for drugs targeting neurological conditions.
Metabolism-related properties, which are vital for understanding how a drug is processed by the body. It evaluates whether a compound is a substrate or inhibitor of key cytochrome P450 (CYP450) enzymes such as CYP3A4, CYP2D6, CYP2C9, and others. These enzymes play a central role in the biotransformation of drugs and can significantly affect a drug’s half-life, efficacy, and potential for drug-drug interactions.
excretion-related properties that help evaluate how efficiently a drug candidate is eliminated from the body. These include total clearance, which combines renal and hepatic pathways, and whether a compound is a substrate or inhibitor of renal transporters such as OCT2. Proper excretion ensures that the drug does not accumulate to toxic levels and maintains a desirable dosing frequency.
A comprehensive toxicity predictions, which are crucial for assessing the safety profile of a drug candidate before experimental testing. It evaluates endpoints such as AMES mutagenicity, hepatotoxicity, cardiotoxicity (e.g., hERG inhibition), and LD50 values. These predictions help identify compounds that may cause genetic damage, liver injury, heart rhythm disturbances, or general acute toxicity.
Physicochemical Parameters
Property |
Safe Limit |
Notes |
|---|---|---|
Molecular weight |
< 500 Da |
Lipinski rule of five |
logP |
< 5 |
Optimal lipophilicity for absorption |
TPSA |
< 140 Ų |
Good oral bioavailability |
H-bond donors |
≤ 5 |
Lipinski rule |
H-bond acceptors |
≤ 10 |
Lipinski rule |
Lipinski violations |
0 or 1 |
Out of 4 rules |
QED |
> 0.4 |
Drug-likeness score (0-1) |
Stereo centers |
≤ 3 |
Complexity affects synthesis |
Solubility (logS) |
> -5.0 |
More soluble is better |
Hydration Free Energy |
< 0 |
Negative preferred |
Lipophilicity (AZ) |
< 3 |
Optimized for balance |
Absorption
Property |
Safe Limit |
Notes |
|---|---|---|
HIA (Human Intestinal Absorption) |
> 70% |
Good oral bioavailability |
Caco-2 permeability |
> -5.15 (log units) |
Reflects intestinal uptake |
PAMPA |
> -5.0 (log units) |
Passive permeability |
Bioavailability |
> 0.3 |
Fraction reaching systemic circulation |
Pgp-substrate |
No |
Avoids efflux and poor absorption |
Pgp-inhibitor |
No |
Avoid drug-drug interactions |
Distribution
Property |
Safe Limit |
Notes |
|---|---|---|
BBB penetration |
> 50 percentile (CNS) or < 50 (non-CNS) |
Central or peripheral activity |
PPBR (Plasma Protein Binding) |
20-90% |
25-75th percentile preferred |
VDss |
< 0.7 L/kg |
Higher for tissue targeting |
Metabolism
Property |
Safe Limit |
Notes |
|---|---|---|
CYP Interaction (substrate/inhibitor) |
< 0.5 |
Low interaction reduces drug–drug interaction risk |
Excretion
Property |
Safe Limit |
Notes |
|---|---|---|
Clearance (hepatocyte) |
< 20 µL/min/10⁶ cells |
Lower values = slower clearance |
Clearance (microsome) |
< 30 µL/min/mg protein |
Microsomal stability |
Half-life |
> 2 hours |
Long enough for convenient dosing |
Toxicity
Property |
Safe Limit |
Notes |
|---|---|---|
AMES test |
Negative |
No mutagenicity |
DILI (Drug-Induced Liver Injury) |
Low (< 0.3) |
Less risk of liver toxicity |
ClinTox |
Low |
Fewer clinical toxicity flags |
Carcinogenicity |
Low |
Avoid cancer risk |
hERG inhibition |
< 0.3 |
Prevents QT prolongation |
LD50 (Zhu) |
> 300 mg/kg |
Safer at therapeutic doses |
Skin Reaction |
No |
Avoid hypersensitivity |
NR/SR toxicity (e.g., NR-AhR, SR-MMP) |
Negative |
Avoid endocrine and stress pathway toxicity |
Usage Instructions
Navigate to the ADMET Prediction section of the platform.
Paste the SMILES string of your compound into the input field.
Click Run ADMET prediction to run the analysis.
Results will be displayed in a structured table with downloadable CSV option.
Summary
The ADMET prediction module evaluates drug-likeness by predicting key properties across physicochemical, absorption, distribution, metabolism, excretion, and toxicity categories. It helps identify compounds with favorable pharmacokinetics and safety profiles early in drug development. Classification of all the properties are based on DrugBank’s prediction score and precentile guidlines.