ИСТИНА

The pharmacokinetic properties and toxicity of potential drugs and other organic compounds (ADMET properties: absorption, distribution, metabolism, excretion, toxicity) critically affect their efficacy, pharmacological profile, administration protocol and safety. Their optimization is an important aspect of drug discovery and development, and the ability to predict these properties for new structures and identify specific structural features responsible for increasing or decreasing them can substantially improve its speed and efficiency as well as minimize the risks of adverse effects. We have developed a general methodology for the prediction of ADMET parameters based on the application of artificial neural networks and fragmental descriptors to extensive experimental data sets. During the model construction, the GPU-based deep learning and double cross-validation are used to achieve optimal performance and model predictivity. The resulting models are often superior in the prediction accuracy and the applicability domain to the models previously published in the literature. As an additional guidance for ADMET optimization, a graphic map highlighting the parts of a molecule that make positive or negative contributions to the predicted property is generated based on the influence of each fragmental descriptor. These models are implemented in an integrated online service available on the Internet (http://qsar.chem.msu.ru/admet/) that supports convenient prediction and analysis of important properties such as lipophilicity, blood-brain barrier permeability, human intestinal absorption, hERG-mediated cardiac toxicity, Ames mutagenicity, aquatic toxicity, etc. This integrated prediction system may be used in the research in various areas of medicinal chemistry, pharmacology, and toxicology.