A compound can have perfect binding affinity in silico, but if it's toxic, metabolized too quickly, or can't reach its target, it will never become a drug. ADMET properties — Absorption, Distribution, Metabolism, Excretion, and Toxicity — are the gatekeepers of pharmaceutical success.

Why ADMET Matters

Statistics tell the story:

• 90% of drug candidates fail in clinical trials
• 40% fail due to poor pharmacokinetics (ADME issues)
• 30% fail due to toxicity
• Average cost per failed Phase II drug: $100-200 million

Predicting ADMET early saves time, money, and lives. Computational ADMET prediction has become a first-line filter in modern drug discovery.

The Five Pillars of ADMET

A — Absorption

Can the drug get into the bloodstream?

Most drugs are taken orally, so they must cross the intestinal epithelium. Key factors:

• Lipinski's Rule of Five: MW <500, logP <5, H-bond donors <5, acceptors <10
• Permeability: Caco-2 cell assays predict intestinal absorption
• Solubility: Must dissolve in gut fluid (aqueous solubility >60 µg/mL preferred)
• Efflux pumps: P-glycoprotein can pump drugs back out

BioPipeline prediction: Calculates Lipinski violations, logP, and molecular weight automatically.

D — Distribution

Can the drug reach its target tissue?

Once absorbed, drugs must distribute to target organs. Challenges:

• Blood-brain barrier (BBB): CNS drugs must cross; peripheral drugs should NOT
• Plasma protein binding: High binding (>99%) limits free drug availability
• Volume of distribution: How widely drug spreads (Vd <0.1 L/kg = stays in blood; >1 L/kg = extensive tissue distribution)
• Tissue accumulation: Can cause toxicity if concentrated in organs

Computational prediction: logP and PSA (polar surface area) indicate BBB permeability.

M — Metabolism

How is the drug chemically modified in the body?

The liver's cytochrome P450 (CYP) enzymes are the main culprits:

• CYP3A4: Metabolizes ~50% of drugs — avoid if possible
• CYP2D6: ~25% of drugs — genetic polymorphisms affect activity
• CYP2C9, 2C19, 1A2: Other major isoforms

Why it matters:

• Too fast → drug cleared before it works (short half-life)
• Too slow → drug accumulates, causing toxicity
• Drug-drug interactions → one drug inhibits/induces CYP, altering another drug's levels

BioPipeline prediction: Flags CYP substrate/inhibitor likelihood using structural alerts.

E — Excretion

How is the drug eliminated?

Two main routes:

• Renal (kidneys): Hydrophilic drugs filtered by glomeruli
• Hepatic (liver → bile → feces): Lipophilic drugs conjugated and excreted

Half-life considerations:

• Too short (<2 hours) → frequent dosing needed (poor compliance)
• Too long (>24 hours) → risk of accumulation, slow to reach steady state
• Ideal: 6-12 hours for once or twice daily dosing

T — Toxicity

Does the drug cause harm?

Types of toxicity:

• Hepatotoxicity: Liver damage (most common reason for drug withdrawal)
• Cardiotoxicity: QT interval prolongation → fatal arrhythmias
• Nephrotoxicity: Kidney damage (especially with chronic use)
• Mutagenicity: DNA damage → cancer risk
• hERG inhibition: Cardiac potassium channel → sudden death

Computational toxicology: QSAR models predict structural alerts (e.g., Michael acceptors, quinones, nitroaromatics).

Lipinski's Rule of Five: The Gold Standard

In 1997, Christopher Lipinski analyzed marketed oral drugs and found patterns:

• Molecular weight: <500 Da
• LogP: <5 (lipophilicity)
• H-bond donors: <5 (OH, NH groups)
• H-bond acceptors: <10 (O, N atoms)

Interpretation: Violating >1 rule increases likelihood of poor absorption.

Modern extensions:

• Veber's rules: Rotatable bonds <10, PSA <140 Ų
• Pfizer's 3/75 rule: LogP <3 and PSA >75 reduces toxicity
• CNS drugs: MW <450, PSA <90, logP 2-5

Real-World ADMET Failures

Terfenadine (Seldane) — Withdrawn 1998

Problem: CYP3A4 substrate → when taken with erythromycin or grapefruit juice, accumulates → hERG inhibition → fatal cardiac arrhythmias.

Rofecoxib (Vioxx) — Withdrawn 2004

Problem: Cardiovascular toxicity not caught in preclinical testing. $2.5 billion liability, 38,000 deaths attributed.

Trovafloxacin (Trovan) — Withdrawn 1999

Problem: Severe hepatotoxicity in <1% of patients, but risk too high for antibiotic use.

How BioPipeline Helps

BioPipeline automatically calculates 20+ ADMET descriptors for every compound:

• Lipinski compliance — pass/fail with violation count
• LogP and PSA — predict absorption and BBB permeability
• CYP450 flags — substrate/inhibitor likelihood for major isoforms
• Structural alerts — toxic moieties (e.g., nitro groups, epoxides)
• Synthetic accessibility — how hard to make in the lab

Best Practices for ADMET-Aware Design

1. Filter early: Apply Lipinski and Veber rules before docking
2. Balance properties: Don't over-optimize potency at the expense of ADMET
3. Check CYP liabilities: Avoid CYP3A4 substrates if possible
4. Monitor toxicophores: Remove Michael acceptors, aldehydes, azos
5. Validate computationally: Use multiple prediction tools (consensus improves accuracy)
6. Test experimentally: In vitro ADME assays (Caco-2, microsomal stability) before advancing

The Bottom Line

A molecule with -15 kcal/mol binding affinity means nothing if it:

• Can't be absorbed (poor solubility)
• Doesn't reach the target (BBB impermeability)
• Is rapidly metabolized (short half-life)
• Causes liver toxicity

ADMET-driven drug design integrates these properties from day one. With computational predictions now 70-85% accurate, there's no excuse for late-stage ADMET failures.

Start your next project with ADMET in mind — your clinical trial will thank you.