FPADMET

FPADMET predicts ADMET (absorption, distribution, metabolism, excretion, and toxicity) properties using molecular fingerprint-based machine learning models to prioritize drug candidates in early-stage drug development.

Key Features:

• Molecular fingerprints: Implements a diverse set of 20 binary fingerprints derived from substructure keys, atom pairs, local path environments, and all-shortest paths.
• Machine learning: Leverages machine learning methodologies to improve ADMET prediction accuracy, including in data-scarce contexts.
• Random forest models: Trains fingerprint-based random forest models for ADMET endpoint prediction.
• Endpoint coverage: Provides predictive models tailored for more than 50 distinct ADMET and related endpoints.
• Benchmarking: Includes comprehensive evaluation comparing fingerprint-based models with conventional 2D/3D molecular descriptors.
• Developer: Developed by Vishsoft.

Scientific Applications:

• Early-stage candidate prioritization: Predicts ADMET endpoints to support prioritization of drug candidates during early drug development.
• Pharmacokinetics modeling: Models absorption, distribution, metabolism, and excretion properties for small molecules.
• Toxicology prediction: Predicts toxicity-related endpoints relevant to toxicological assessment.
• Method benchmarking: Enables comparative assessment of fingerprint-based models versus 2D/3D molecular descriptors across numerous endpoints.

Methodology:

Generates 20 binary fingerprints (substructure keys, atom pairs, local path environments, all-shortest paths) and trains fingerprint-based random forest models, with performance evaluated against 2D/3D molecular descriptors across more than 50 ADMET endpoints.