ADMET

ADMET is the composite pharmacokinetic and toxicological framework—standing for Absorption, Distribution, Metabolism, Excretion, and Toxicity—that quantifies how a living organism processes a xenobiotic small molecule or biologic and whether the compound achieves therapeutic unbound concentrations at its biological target without eliciting organ injury. In pharmaceutical research and drug discovery, high in vitro potency (such as sub-nanomolar dissociation constants K_d) does not assure clinical efficacy if low intestinal permeability, hepatic first-pass clearance, or off-target ion-channel blockade prevents safe systemic exposure.

Historically, benchmark retrospective analyses by Ismail Kola and John Landis (2004) demonstrated that in 1991, unfavorable pharmacokinetics and bioavailability accounted for ~40% of all clinical trial failures across pharmaceutical pipelines. The integration of high-throughput in vitro microsome and cell-monolayer assays alongside in silico chemoinformatics models—using molecular descriptors, quantitative structure-activity relationship (QSAR) predictors, and graph neural networks—reduced pharmacokinetic-driven clinical attrition to less than 10% by the 2010s, shifting optimization of ADMET properties into early hit-to-lead discovery stages.

Intestinal Absorption, Solubility, and Efflux Transporters

For an orally administered drug to enter systemic circulation, it must dissolve in gastrointestinal fluids and traverse the intestinal epithelium. Systemic oral bioavailability (F) is the product of three sequential fractions:

F = f_a × f_g × f_h

where f_a is the fraction absorbed across the apical enterocyte membrane, f_g is the fraction escaping gut-wall metabolism, and f_h = 1 − (CL_h / Q_h) is the fraction escaping first-pass hepatic extraction (where CL_h is hepatic blood clearance and Q_h ≈ 20 mL·min⁻¹·kg⁻¹ is human hepatic blood flow).

Passive transcellular diffusion depends on thermodynamic and kinetic aqueous solubility (S > 50–100 μg/mL), lipophilicity (Log D_7.4 between 1.0 and 3.0), and ionization state governed by the Henderson–Hasselbalch equation for weak acids and bases (p*K*_a). Permeability is quantified experimentally across human colorectal adenocarcinoma (Caco-2) or canine kidney (MDCK) cell monolayers as the apparent permeability coefficient (P_app):

P_app = (d*Q* / d*t*) / (A · C₀)

where d*Q*/d*t* is the steady-state molar transport rate across membrane surface area A at donor concentration C₀. Compounds with P_app > 10 × 10⁻⁶ cm/s generally achieve f_a > 80%. However, apical ATP-binding cassette (ABC) efflux pumps—principally P-glycoprotein (P-gp / MDR1 / ABCB1) and Breast Cancer Resistance Protein (BCRP / ABCG2)—extrude substrates back into the gut lumen. An efflux ratio ER = P_app(B→A) / P_app(A→B) > 2.0 indicates active P-gp transport that can limit oral absorption and central nervous system penetration.

Volume of Distribution, Plasma Protein Binding, and Blood–Brain Barrier

Once in systemic blood, a drug partitions between plasma proteins—primarily human serum albumin (HSA) for acidic drugs and α₁-acid glycoprotein (AAG) for basic drugs—and peripheral tissues. According to the free drug hypothesis, only the unbound fraction (f_u) in plasma can diffuse across capillary endothelia to engage pharmacological targets or undergo glomerular filtration. The steady-state volume of distribution (V_d,ss) relates the amount of drug in the body to its plasma concentration (V_d,ss = V_p + V_t · [f_u / f_u,t]), ranging from ~0.04 L/kg for macromolecules restricted to plasma volume (V_p) to >10 L/kg for lipophilic amines sequestered in lysosomal and adipose compartments.

Crossing the Blood–Brain Barrier (BBB)—formed by cerebral microvascular endothelial cells sealed by tight junctions (claudin-5 and occludin)—is evaluated by the brain-to-blood concentration ratio, log BB = log(C_brain / C_blood), or the unbound brain-to-plasma partition coefficient (K_p,uu,brain). Travis Wager and colleagues at Pfizer formulated the Central Nervous System Multiparameter Optimization (CNS MPO) score across six physicochemical properties (MW ≤ 360 Da, Clog*P* ≤ 3.0, Clog*D* ≤ 2.0, TPSA 40–90 Ų, HBD ≤ 0.5, and most basic p*K*_a ≤ 8.0), showing that 74% of marketed CNS drugs achieve a CNS MPO score ≥ 4.0 out of 6.0.

Phase I Cytochrome P450 Oxidation and Phase II Conjugation

Hepatic biotransformation converts lipophilic xenobiotics into hydrophilic metabolites suitable for renal or biliary elimination through two enzymatic phases. Phase I functionalization reactions—comprising aliphatic/aromatic hydroxylation, N- and O-dealkylation, epoxidation, and sulfoxidation—are catalyzed predominantly by the heme-thiolate monooxygenase superfamily Cytochrome P450 (CYP) in the hepatic smooth endoplasmic reticulum. Using NADPH and O₂, CYP enzymes execute the stoichiometry:

RH + O₂ + NADPH + H⁺ → ROH + H₂O + NADP⁺

Out of 57 human CYP genes, five isoforms metabolize >90% of clinical drugs: CYP3A4/5 (~50% of drugs), CYP2D6 (~20%, polymorphic across poor and ultra-rapid metabolizers), CYP2C9 (~12%), CYP2C19 (~8%), and CYP1A2 (~4%). Non-CYP Phase I enzymes include flavin-containing monooxygenases (FMOs), aldehyde oxidase (AO), and carboxylesterases. Phase II conjugation reactions subsequently attach polar endogenous cofactors—such as glucuronic acid via UDP-glucuronosyltransferases (UGTs), sulfate via sulfotransferases (SULTs), glutathione via glutathione S-transferases (GSTs), or acetyl groups via N-acetyltransferases—increasing water solubility and reducing pharmacological activity.

Metabolic stability is measured in vitro in human liver microsomes (HLM) or cryopreserved hepatocytes to determine intrinsic clearance (CL_int = V_max / K_m). Under the well-stirred hepatic model, in vivo hepatic blood clearance is scaled as:

CL_h = (Q_h · f_u,b · CL_int) / (Q_h + f_u,b · CL_int)

Compounds that act as reversible or mechanism-based time-dependent inhibitors (TDI) of CYP3A4 or CYP2D6 (or nuclear receptor PXR/CAR inducers) trigger clinical drug–drug interactions (DDIs) by elevating or depleting co-administered drug concentrations.

Renal and Biliary Excretion and Half-Life

Excretion removes intact parent drugs and polar Phase I/II metabolites primarily via the kidneys and hepatobiliary tract. Net renal clearance (CL_r) reflects the balance of three nephron processes:

CL_r = (f_u · GFR + CL_sec) × (1 − F_reabs)

where the human glomerular filtration rate (GFR) is ~125 mL/min, CL_sec is active proximal tubular secretion mediated by basolateral Organic Anion/Cation Transporters (OAT1/3, OCT2) and apical MATE1/2-K or MRP2/4 pumps, and F_reabs is passive distal tubular reabsorption. Polar conjugates with MW > 400–500 Da are frequently excreted into bile via canalicular BSEP and MRP2 transporters. The terminal elimination half-life (t_1/2) is governed by both volume of distribution and systemic clearance:

t_1/2 = (0.693 · V_d) / CL_tot

Off-Target Cardiotoxicity, Hepatotoxicity, and Genotoxicity

Preclinical toxicity screening targets three primary organ and genetic liabilities that historically forced post-marketing drug withdrawals:

1. hERG Cardiotoxicity: Blockade of the cardiac voltage-gated potassium channel K_v11.1 (encoded by the KCNH2 / hERG gene) impairs the delayed rectifier K⁺ current (I_Kr) during ventricular repolarization. This prolongs the electrocardiographic QT interval and precipitates polymorphic ventricular tachycardia (torsades de pointes), which caused the withdrawal of terfenadine, cisapride, and grepafloxacin. Basic lipophilic amines contain the classic hERG pharmacophore (π-stacking with Tyr652 and Phe656 inside the central pore cavity); regulatory guidelines (ICH S7B) require an unbound safety margin of hERG IC₅₀ / C_max,u ≥ 30.
2. Drug-Induced Liver Injury (DILI): Reactive electrophilic metabolites—such as N-acetyl-p-benzoquinone imine (NAPQI) formed by CYP2E1 oxidation of acetaminophen, or epoxides and acyl glucuronides—deplete hepatic glutathione (GSH) and covalently modify mitochondrial proteins, or inhibit the Bile Salt Export Pump (BSEP), causing cholestatic or hepatocellular necrosis.
3. Mutagenicity and Phospholipidosis: Genotoxicity is evaluated via the Ames reverse-mutation assay in Salmonella typhimurium histidine-auxotrophic strains (TA98, TA100) ± S9 metabolic activation and the in vitro micronucleus clastogenicity test, alongside Structural Alerts (Ashby–Tennant toxicophores such as aromatic nitro, aniline, and Michael acceptor groups) and pan-assay interference compounds (PAINS).

Modern in silico ADMET platforms—including SwissADME, pkCSM, ADMETlab 3.0, and multi-task D-MPNN (Chemprop) models trained on millions of endpoints in PubChem and ChEMBL—predict >50 ADMET endpoints simultaneously during virtual screening prior to chemical synthesis.

Why can increasing a drug's lipophilicity (Log P) simultaneously improve target binding yet degrade its overall ADMET profile?

Increasing lipophilicity (Log P > 4–5) often enhances in vitro binding affinity via the hydrophobic effect (releasing ordered water molecules from a non-polar receptor pocket, gaining ~0.7 kcal/mol per added methylene group). However, high lipophilicity creates a cascade of ADMET liabilities: it lowers aqueous solubility below 10 μg/mL, increases binding to hydrophobic active sites of hepatic CYP3A4 (accelerating first-pass metabolic clearance CL_int), enhances promiscuous blockade of the cardiac hERG K⁺ channel, and promotes tissue accumulation (phospholipidosis). Medicinal chemists therefore optimize Lipophilic Ligand Efficiency (LLE = pIC₅₀ − Log P ≥ 5).

How does the free drug hypothesis govern the interpretation of plasma protein binding (f_u)?

The free drug hypothesis states that at steady state, the unbound drug concentration in plasma (C_u = f_u · C_plasma) equilibrates with the unbound drug concentration in extracellular tissue water surrounding the pharmacological target (absent active transport). As demonstrated by Dennis Smith, Li Di, and Edward Kerns (2010), attempting to reduce plasma protein binding through structural modification rarely increases in vivo unbound exposure (AUC_u) for cleared oral drugs because a higher unbound fraction (f_u) proportionally increases hepatic intrinsic clearance (CL_h ≈ f_u · CL_int), leaving steady-state C_u unchanged.

What is mechanism-based (time-dependent) Cytochrome P450 inhibition, and why does it cause clinical drug–drug interactions?

Unlike reversible competitive inhibition—where an inhibitor detaches from the CYP active site as plasma concentrations decline—mechanism-based (time-dependent) inhibition (TDI) occurs when a CYP isoform oxidizes a substrate (such as furanocoumarins in grapefruit juice, macrolide antibiotics like clarithromycin, or ritonavir) into a reactive intermediate that forms a covalent bond with the prosthetic heme iron or apoprotein. This permanently inactivates the enzyme molecule. Recovery of metabolic capacity requires de novo hepatic synthesis of new CYP protein (with a turnover half-life of 24–72 hours), causing sustained multi-fold spikes in co-administered drug exposure.

1. Kola, I., & Landis, J. (2004). "Can the Pharmaceutical Industry Reduce Attrition Rates?" Nature Reviews Drug Discovery, 3(8), 711–716. https://doi.org/10.1038/nrd1470
2. van de Waterbeemd, H., & Gifford, E. (2003). "ADMET in Silico Modelling: Towards Prediction Paradise?" Nature Reviews Drug Discovery, 2(3), 192–204. https://doi.org/10.1038/nrd1032
3. Wager, T. T., Hou, X., Verhoest, P. R., & Villalobos, A. (2010). "Moving Beyond Rules: The Development of a Central Nervous System Multiparameter Optimization (CNS MPO) Approach to Enable Alignment of Druglike Properties." ACS Chemical Neuroscience, 1(6), 435–449. https://doi.org/10.1021/cn100008c
4. Smith, D. A., Di, L., & Kerns, E. H. (2010). "The Effect of Plasma Protein Binding on In Vivo Efficacy: Misconceptions in Drug Discovery." Nature Reviews Drug Discovery, 9(12), 929–939. https://doi.org/10.1038/nrd3287