ADMET Analysis with SwissADME

📑 In This Article

What is ADMET?
Lipinski's Rule of Five
Using SwissADME
Key Parameters
ADMETsar
Docking Workflow Integration

Overview

Approximately 90% of drug candidates that enter clinical trials fail — a significant portion not because of insufficient potency, but because of poor pharmacokinetics and toxicity. The good news: many of these failures can be predicted computationally, before any synthesis or animal testing. SwissADME is the most widely used free web tool for ADMET prediction — this guide walks you through every parameter and how to interpret results.

💊 What is ADMET?

AAbsorptionCan the drug enter the bloodstream?

DDistributionDoes it reach the target tissue?

MMetabolismHow is it broken down in the body?

EExcretionHow and when is it eliminated?

TToxicityDoes it cause harm to cells or organs?

These properties collectively determine a molecule's pharmacokinetic (PK) profile — whether a drug gets to where it needs to be, in sufficient concentration, for long enough, without causing harm.

📐 Lipinski's Rule of Five — The Foundation

Before SwissADME, the standard filter was Lipinski's Rule of Five (1997) — simple physicochemical limits derived from known oral drugs. A molecule violating more than one rule is unlikely to be orally bioavailable.

≤ 500 Da

Molecular Weight (MW)

Affects membrane permeability and bioavailability

LogP ≤ 5

Lipophilicity

Octanol-water partition — controls membrane crossing ability

HBD ≤ 5

H-Bond Donors

Excess donors significantly reduce oral absorption

HBA ≤ 10

H-Bond Acceptors

Affects solubility and membrane permeability

ℹ️

Common Exceptions

Biological macromolecules (peptides, natural products, antibiotics) are well-known exceptions. Rule of Five applies specifically to small-molecule oral drugs.

🌐 Using SwissADME Step by Step

Access SwissADME

Go to swissadme.ch — no registration required. Completely free.

Enter Your Molecule

Paste a SMILES string for your compound. Get SMILES from PubChem, ChEMBL, or draw in ChemDraw.

Run the Analysis

Click "Run" — results in seconds. Analyse up to 10 molecules simultaneously for batch comparison.

Interpret the Output

Results organized into: Physicochemistry, Lipophilicity, Water Solubility, Pharmacokinetics, Druglikeness, and Medicinal Chemistry.

💡

Getting a SMILES String

In PubChem: search → "Canonical SMILES" in Properties. In ChemDraw: Edit → Copy As → SMILES. Marvin Sketch also lets you draw and export SMILES online.

📊 Key Parameters to Evaluate

Physicochemistry

Parameter	Ideal Range	Why It Matters
Molecular Weight	150–500 Da	Affects membrane permeability and bioavailability
LogP (iLOGP)	0–5	Lipophilicity determines membrane crossing ability
H-Bond Donors	≤ 5	Excess H-bond donors reduce oral absorption
H-Bond Acceptors	≤ 10	Affects solubility and permeability
Rotatable Bonds	≤ 10	Too flexible = poor oral bioavailability
TPSA	< 140 Ų	High TPSA = poor permeability; <90 Ų required for CNS drugs

Water Solubility (ESOL)

Poor solubility is one of the most common reasons drug candidates fail. SwissADME predicts aqueous solubility using ESOL, Ali, and SILICOS-IT methods:

Solubility Class	LogS Value	Status
Highly soluble	LogS > −1	✅ Excellent
Soluble	−1 to −2	✅ Good
Moderately soluble	−2 to −4	⚠️ Acceptable
Poorly soluble	−4 to −6	❌ Risky
Insoluble	LogS < −6	❌ Problematic

Pharmacokinetics — Key Predictions

GI Absorption: Predicted as High or Low based on physicochemical properties

Blood-Brain Barrier (BBB): Critical for CNS drugs — predicted Yes/No

P-glycoprotein Substrate: P-gp efflux reduces effective drug concentration in cells

CYP Inhibition: Predicts which cytochrome P450 enzymes are inhibited — critical for drug-drug interactions

⚠️

CYP450 Inhibition Warning

If SwissADME predicts inhibition of CYP3A4, CYP2D6, or CYP2C9, this is a serious red flag. These enzymes metabolize most drugs on the market, and inhibition causes dangerous drug-drug interactions.

Druglikeness Scores

Lipinski Ro5

Rule of Five

0 or 1 violation = drug-like. Gold standard for oral small-molecule candidates.

Veber Rules

Oral Bioavailability

Rotatable bonds ≤10 + TPSA ≤140 Ų — predicts oral bioavailability independently of Lipinski.

Egan Filter

Passive Absorption

TPSA ≤ 131.6 Ų + LogP ≤ 5.88 — predicts human intestinal absorption.

Bioavailability Score

Oral Probability

Range 0.11 to 0.55 — probability of ≥10% oral bioavailability in rat models.

🔬 Going Deeper with ADMETsar

ADMETsar (admetsar.scbdd.com) provides more comprehensive toxicity predictions that complement SwissADME, particularly for safety-critical endpoints:

Ames Test Mutagenicity — bacterial mutagenicity prediction

Carcinogenicity Risk — rodent carcinogenicity prediction

hERG Cardiotoxicity — critical cardiac safety assessment

Acute Oral Toxicity — LD50 prediction (mg/kg)

Human Intestinal Absorption and Caco-2 permeability

Blood-Brain Barrier Penetration — quantitative score

🖼️

Fig 1. In silico ADMET screening filters thousands of compounds to identify the most promising candidates for synthesis — saving significant time and experimental resources before any wet lab work begins.

🔄 Integrating ADMET into Your Docking Workflow

Screen for Drug-Likeness First

Apply Lipinski/Veber filters to your virtual library before docking to save computation time and reduce false positives.

Dock the Filtered Library

Only dock compounds that pass ADMET pre-filters — dramatically reduces the hit list to meaningful candidates.

Run ADMET on Top Hits

For compounds with strong binding affinity, run full ADMET profiling to identify pharmacokinetic liabilities.

Prioritize by Multi-Parameter Score

Best candidates balance potency (docking score) with drug-likeness AND safety profile simultaneously.

🚩 Instant Rejection Criteria

Immediately deprioritize any compound that meets any of the following:

✕Violates Lipinski in 2+ rules

✕Has LogP > 5 (excessive lipophilicity)

✕Has TPSA > 140 Ų (unless CNS: >90 Ų)

✕Predicted hERG inhibitor (cardiac toxicity risk)

✕Fails mutagenicity — Ames test positive

ADMET Analysis with SwissADME:Predict Drug-Likeness Before Synthesis