What is ADMET Predictor™?

ADMET Predictor is sophisticated, yet very easy to use, computer software for advanced predictive modeling of ADMET properties. The "ADMET" acronym is commonly used in the pharmaceutical industry to indicate all the phenomena associated with Absorption, Distribution, Metabolism, Elimination, and Toxicity of chemical substances in the human body. ADMET Predictor not only rapidly estimates a number of vital ADMET properties (listed below) from molecular structures, but is also capable of building predictive models of new properties from user's data via its integrated ADMET Modeler™ module. The program predictions can be further utilized as inputs for our software products GastroPlus™ and MedChem Studio™, or used internally by a flexible screening filters ADMET Risk that can include any combination of predicted properties. All predictions for single structures, as well as structure editing, can also be directly accessed from MedChem Designer™.

What makes ADMET Predictor unique?

  • ACCURACY: its predictive models have been consistently ranked as the most accurate by independent third party comparisons (see the "Publications" link for details).
  • SPEED: it can process up to ~200,000 compounds per hour on a notebook PC.
  • EASE OF USE: using the program is as simple as opening a file containing molecular structures.
  • OUTPUT OPTIONS: it provides a variety of output channels to store, transfer, or visualize its results quickly and efficiently.
  • BATCH CAPABILITY: it has a built-in batch mode and can run as a Pipeline Pilot™ component.
  • NOVEL DESCRIPTORS: it features a unique collection of rapid estimates of quantum mechanical descriptors of electronic density and reactivity at both atomic and molecular levels - a result of federally funded 3-year research project.

What does it predict?

Physico-Chemical and Biopharmaceutical Module

  • pKa (ionization constants; a multiprotic model)
  • Human effective permeability in jejunum (Peff)
  • MDCK apparent permeability (Papp)
  • Corneal permeability
  • Human skin permeability
  • Solubility
    • - Native solubility (solubility in pure water)
    • - Native pH at saturation in pure water
    • - Intrinsic solubility in pure water
    • - Salt solubility factor
    • - Water solubility at user-specified pH
    • - Solubility in the simulated gastrointestinal fluids: FaSSGF, FaSSIF, and FeSSIF
  • Supersaturation ratio (a tendency to supersaturate in water)
  • logP (two models: artificial neural network ensemble and Moriguchi)
  • logD (estimation of octanol-water distribution coefficient at user-defined pH)
  • Air-water partition coefficient (i.e., Henry's Law constant)
  • Diffusion coefficient in water (Hayduk-Laudie formula)
  • Molal volume (Schroeder formula)
  • Blood-brain barrier permeation (two models: qualitative and quantitative as logarithm of the BBB partition coefficient).
  • Human plasma protein binding as percent unbound
  • Human volume of distribution
  • Blood-to-plasma concentration ratio
  • NEW: Fraction unbound in human liver microsomes
  • Inhibition of the hepatic OATP1B1 transporter in human
  • UPGRADED: Likelihood of efflux by P-glycoprotein
  • Likelihood of inhibiting efflux by P-glycoprotein
  • Activity models
    • - Inhibition of HIV integrase mediated strand transfer
    • - Inhibition of HIV integrase mediated 3' processing

Metabolism Module

  • Likely sites of metabolic attack by five CYP P450 enzymes: 1A2, 2A6, 2B6, 2C8, 2C9, 2C19, 2D6, 2E1, 3A4 (nine models)
  • Classification of whether a molecule will be a substrate of one of the five CYP P450 enzymes: 1A2, 2A6, 2B6, 2C8, 2C9, 2C19, 2D6, 2E1, 3A4 (nine models).
  • UPGRADED: Michaelis-Menten kinetic Km constants for hydroxylation reaction catalyzed by five CYP P450 enzymes: 1A2, 2C9, 2C19, 2D6, 3A4 (six models; 1) now at atomic resolution; 2) 3A4 also features a model build from human liver microsomes data)
  • UPGRADED: Michaelis-Menten kinetic Vmax constants for hydroxylation reaction catalyzed by five CYP P450 enzymes: 1A2, 2C9, 2C19, 2D6, 3A4 (six models; 1) now at atomic resolution; 2) 3A4 also features a model build from human liver microsomes data)
  • UPGRADED: Intrinsic clearance, CLint, resulting from metabolic activity of five CYP P450 enzymes: 1A2, 2C19, 2C9, 2D6, and 3A4 (six models; 1) now at atomic resolution; 2) 3A4 also features a model build from human liver microsomes data)
  • NEW: Ooverall intrinsic clearance in human liver microsomes
  • General inhibitory properties against four CYP P450 enzymes: 1A2, 2C9, 2C19, 2D6, and 3A4 (five models)
  • Specific inhibitory properties against 3A4-mediated metabolism of midazolam and testosterone (two models)
  • Specific inhibitory constant Ki against 3A4-mediated metabolism of midazolam and testosterone (two models)
  • Classification of whether a molecule will be glucuronidated by one of the nine isoforms of the Uridine 5'-Diphosphate-Glucuronosyltransferase (UGT): 1A1, 1A3, 1A4, 1A6, 1A8, 1A9, 1A10, 2B7, and 2B15.

Toxicity Module

  • Estrogen receptor toxicity (two models: qualitative and, if toxic, quantitative ratio of IC50 estradiol/IC50 compound)
  • Androgen receptor toxicity (two models: qualitative and, if toxic, quantitative ratio of IC50 androgen/IC50 compound)
  • Maximum recommended therapeutic dose (MRTD)
  • Environmental bioconcentration factor (a.k.a. bioaccumulation factor)
  • Likelihood of a chemical's biodegradation in the environment
  • Fathead minnow lethality as LC50
  • Acute toxicity in Daphnia magna (water fleas) as pLC50
  • Allergenic skin sensitization
  • Allergenic respiratory sensitization.
  • Mutagenic chromosomal aberrations.
  • Phospholipidosis.
  • Reproductive / developmental toxicity.
  • Acute toxicity in rats as LD50
  • Carcinogenicity in rats as TD50
  • Carcinogenicity in mice as TD50
  • Qualitative filter of mutagenicity in 10 strains of Salmonella bacteria
  • hERG-encoded K+ channel affinity (two models: qualitative and quantitative as pIC50)
  • Acute toxicity in Tetrahymena pyriformis as pIGC50 (growth inhibition)
  • Human liver adverse effects of drugs as the binary likelihood of causing elevation in diagnostic liver enzymes: alkaline phosphatase, SGOT, SGPT, LDH, and GGT.

Simulation Module

  • Fraction absorbed in human (by simulation at 1 mg, 10 mg, 100 mg, and 1000 mg dose levels)
  • Optimal dose in human, in mg, matching desired efficacious concentration in blood plasma

Customizable ADMET Risk filters

  • Risk of low absorption from an oral dose (three models: one derived from focused subset of the World Drug Index, one trained on measured fraction absorbed in human, and one identical to the Lipinski's Rule of 5)
  • Risk of mutagenicity
  • Risk of overall toxicity
  • UPGRADED: Risk of metabolic liability
  • UPGRADED: Global ADMET Risk summarizing all of the above in one score

If your quantity of interest is not listed above, then you can use the integrated ADMET Modeler module to easily build and append an appropriate predictive model to ADMET Predictor!

What other features does the program offer?

Model Building Engine

  • A sophisticated ADMET Modeler™ module featuring an array of powerful modeling algorithms plus SALI analysis (Structure-Activity Landscape Index) for assessing model quality.

Descriptor Sensitivity Analysis

  • Interpretation of model predictions in structural terms for guided design of molecule's derivatives with desired ADMET properties

Associative modeling

  • Extending the scope of ADMET predictions using your own data, without the need to retrain the underlying models (a.k.a. "LIBRARY Mode")

Detailed Structure Depiction

  • Graphical map of atomic charge, polarizability, E-state, and reactivity distributions, plus atomic sites of likely metabolic attack and atom-specific Michaelis-Menten kinetic constants of CYP-mediated metabolism.

Graphical User Interface with 2D Graphing

  • Rich collection of data visualization tools, including distribution histograms, and correlation graphs

Multidimensional State-of-the-Art Graphics

  • Data display in a rotatable 3D scatter plot. Individual points may be colored by a fourth variable.
  • An integrated Miner3D graphics component.

What input formats does it support?

  • SMILES strings (2D predictive models only)
  • ISIS/Base™ Structure Data Format (SDF)
  • ISIS/Base Reaction Data Format (RDF)
  • ISIS/Draw™ Molecular Structure Format (MOL)
  • ChemDraw™ Structure Molfile (MOL)
  • Tripos/SYBYL™ Structure Format (MAC)


  • ADMET Predictor comes now with fully integrated ADMET Modeler module - a software application for building powerful predictive models based on Artificial Neural Networks, Support Vector Machines, Multiple Linear Regression, and Kernel Partial Least Squares.
  • ADMET Predictor outputs are 100% compatible with GastroPlus - a software application developed by Simulations Plus for simulating the oral absorption, distribution, metabolism and excretion of drugs.
  • MedChem Studio - our intuitive cheminformatics platform can run ADMET Predictor as its module providing on-the-fly property predictions.
  • MedChem Designer - our molecule drawing tool that is capable of running all ADMET Predictor models at a mouse click.
  • Input and output of ADMET Predictor are fully compatible with SDF, RDF and MOL file formats used by ISIS/Base and ISIS/Draw.
  • ADMET Predictor can be run in batch mode from command line, which makes it suitable for incorporation into automated data flow control software, such as Pipeline Pilot or KNIME. (Suitable Pipeline Pilot component and KNIME metanode are distributed with the program.)

ISIS/Base, ISIS/Draw, and Pipeline Pilot are trademarks of BIOVIA