vNN Web Server for ADMET Predictions.

Area of Science:

• Computational chemistry
• Drug discovery and development
• Toxicology

Background:

• Early assessment of pharmacokinetic and toxic properties is crucial in drug development to prevent costly failures.
• Existing in silico models for estimating these properties require improvements in speed, ease of use, and reliability.

Purpose of the Study:

• To develop advanced in silico models for predicting key absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties of drug candidates.
• To enhance the speed, ease of use, and reliability of ADMET property predictions in early drug development.

Main Methods:

• Utilized the vNN (vector Neural Network) method to create 15 distinct ADMET prediction models.
• Developed models to assess critical drug properties including cytotoxicity, mutagenicity, cardiotoxicity, drug-drug interactions, microsomal stability, and drug-induced liver injury.

Main Results:

• The study presents 15 vNN-based ADMET models capable of rapid and reliable prediction of essential drug properties.
• Summarizes the predictive performance of each model for various ADMET endpoints.

Conclusions:

• The developed vNN-based ADMET models offer a significant advancement for early-stage drug candidate assessment.
• These publicly available models and the vNN platform facilitate faster, more reliable drug development and custom model building.