QSAR, ADMET and Predictive Toxicology




Understanding and quantifying structure-activity relationships can significantly impact lead optimization and drug development by minimizing tedious and costly experimentation.

Build, validate and apply your own models based on a wide range of approaches, and keep improving them as new data become available.

Assess the potential risk posed by unfavorable pharmacokinetic properties and potential toxicity using BIOVIA’s distributed models, extend them to better cover your proprietary chemical space and use the comprehensible indications to navigate around such liabilities.

  • Comprehensive and consistent data preparation:
    • Ligands: remove duplicates and handle tautomers and ionization
    • Prepare response property (scaling and binning)
    • Split your data into training and test sets with the appropriate methods
  • Choose from a large number of physicochemical, topological, electronic, geometric. fingerprint and Quantum Mechanics based descriptors
  • Create statistical models including Bayesian, MLR (Multiple Linear Regression), PLS (Partial Least Squares), and GFA (Genetic Functional Analysis)
  • Analyze and validate models using model applicability domains (MAD), automatic test set validation, cross validation and statistical metrics
  • Identify Matched Molecular Pairs (MMPs) transformations and study activity cliffs

Get an early assessment of your compounds by calculating the predicted ADMET (absorption, distribution, metabolism, excretion and toxicity) properties for collections of molecules such as synthesis candidates, vendor libraries, and screening collections. Use the results to eliminate compounds with unfavorable ADMET characteristics and evaluate proposed structural refinements, designed to improve these properties prior to synthesis.
ADMET descriptors include:

  • Human intestinal absorption
  • Aqueous solubility
  • Blood brain barrier penetration
  • Plasma protein binding
  • CYP2D6 binding
  • Hepatotoxicity
  • Filter sets of small molecules for undesirable function groups based on published SMARTS rules   

Evaluate your compounds’ performance in experimental assays and animal models. Compute and validate assessments of the toxic and environmental effects of chemicals solely from their molecular structure. TOPKAT® (TOxicity Prediction by Komputer Assisted Technology) employs robust and cross-validated Quantitative Structure Toxicity Relationship (QSTR) models for assessing various endpoints and utilizing the patented Optimal Predictive Space validation method to assist in interpreting the results.

Please note:
Details on our extensible TopKat models have been published in QSAR Model Report Format (QMRF) on the European Commission Joint Research Center (JRC)’s "QSAR Model Database"*.

  • Ames mutagenicity
  • Rodent carcinogenicity (NTP and FDA data)
  • Weight of evidence carcinogenicity
  • Carcinogenic potency TD50
  • Developmental toxicity potential
  • Rat oral LD50
  • Rat maximum tolerated dose
  • Rat inhalation toxicity LC50
  • Rat chronic LOAEL
  • Skin irritancy and sensitization
  • Eye irritancy
  • Aerobic biodegradability
  • Fathead minnow LC50
  • Daphnia magna EC50

*Learn more about the JRC here; for information on BIOVIA-specific models, search “BIOVIA” on this page