ACD/LogP

Predict octanol-water partitioning coefficients from structure

The partition constant, P, is a measure of the propensity of a neutral molecule to differentially dissolve in two immiscible phases, and serves as a quantitative descriptor of lipophilicity. The logP prediction model provides an estimate of the value of the octanol-water partitioning coefficient (also referred to as K_OW) as the logarithmic ratio (logP) from structure.

LogP predictions are used as input variables in many of our PhysChem and ADME prediction algorithms including logD, Oral Bioavailability, Blood-Brain Barrier Permeation, Passive Absorption, and several Toxicity models, such as hERG inhibition and Aquatic toxicity.

Available Algorithms

The logP prediction model incorporates two different predictive approaches—Classic and GALAS (Global, Adjusted Locally According to Similarity). A Consensus logP based on these two models is also available. Experts can investigate each algorithm manually to decide which is more appropriate for particular chemical space, and provide colleagues with guidelines for use.

• Classic
  • The primary algorithm calculates logP using the principle of isolating carbons. Well-characterized logP contributions have been compiled for atoms, structural fragments, and intramolecular interactions derived from >12,000 experimental logP values. A secondary algorithm is applied when unknown fragments are present in the molecule. A detailed description of the original algorithm may be found at "Petrauskas, A., Kolovanov, E., ACD/Log P Method Description. Persp. in Drug Design, 19:1–19, 2000".
  • Source of experimental data—peer-reviewed scientific journals and the BioByte Star list.
• GALAS
  • Training set: 11,387 compounds; Internal validation: 4890 compounds
  • Source of experimental data—reference books (the Merck index, Therapeutic Drugs, Clarke's Isolation and Identification of Drugs), peer-reviewed scientific journals, and other public data sources such as handbooks and online databases.
  • Provides a quantitative estimate of reliability of prediction through the Reliability Index (RI). This number, between 0 and 1, allows you to judge the relevance of the internal training set to the chemical space being investigated by looking for similar structures and evaluating how well the model performs in the local chemical environment of the compound (0=poor reliability, either nothing similar exists in the training set, or the model produces inconsistent predictions for similar compounds; 1=excellent reliability, several identical entries present, and model predictions precisely match given experimental values).
• Consensus LogP
  • Uses both the Classic and GALAS algorithms.
  • Assigns dynamic adaptive coefficients to each model according to the corresponding indications of prediction quality. As a result, each model obtains larger weight in those regions of chemical space where it performs most reliably. This allows maximizing the Applicability Domain of the final model and obtaining maximal overall accuracy for the predicted result.

Training with Experimental Data

To improve logP prediction accuracy and make the model more relevant to in-house chemical space or a particular project, the Classic and GALAS algorithms feature the ability for the user to employ custom experimental data in the expansion of the Applicability Domain of a corresponding prediction method. Since Consensus logP predictions are based on both Classic and GALAS logP, this method does not have individual training options. Instead, whenever any of the underlying algorithms has any training data applied, it is automatically incorporated into Consensus prediction.

Real-world Applications

Pharmaceuticals—used in medicinal chemistry to assess drug likeness; in pharmacokinetics to help determine ADME profiles (the ability of a drug to be absorbed, successfully reach the intended target, be metabolized and excreted); and in pharmacodynamics to understand target receptor binding.

Agrochemicals—applied similarly to pharmaceuticals with the intention of developing herbicides and insecticides.

Environmental—to model migration of dissolved hydrophobic organics in soil and groundwater to help assess waterway pollution, and toxicity to animals and aquatic life.

Consumer Products—used in the formulation of cosmetics, dyes, household cleaners and many other products.

ACD/Labs Product Suites

The Percepta prediction modules are available as bundles to offer cost savings for multiple modules, and provide related modules as a package.

Other Products

ACD/LogP Model predictions are available as a browser-based thin client application on ACD/Percepta Portal; a thick client product (ACD/Percepta for the Desktop); and a batch calculator (ACD/Percepta Batch).