ADMETboost: a web server for accurate ADMET prediction.

Area of Science:

• Computational chemistry
• Pharmacology
• Machine learning in drug discovery

Background:

• Absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties are critical determinants of drug efficacy and safety.
• Accurate prediction of ADMET properties is essential for efficient drug discovery pipelines.

Purpose of the Study:

• To develop and validate a highly accurate machine learning model for predicting ADMET properties.
• To provide a publicly accessible web server (ADMETboost) for ADMET prediction.

Main Methods:

• Utilized an ensemble of molecular features, including fingerprints and descriptors.
• Employed an extreme gradient boosting (XGBoost) tree-based machine learning model.
• Evaluated model performance on the Therapeutics Data Commons ADMET benchmark dataset.

Main Results:

• The developed model achieved top rankings in the Therapeutics Data Commons ADMET benchmark.
• The model ranked first in 18 out of 22 prediction tasks.
• The model achieved a top 3 ranking in 21 out of 22 tasks, demonstrating high accuracy.

Conclusions:

• The extreme gradient boosting model provides accurate ADMET property predictions.
• The ADMETboost web server offers a valuable resource for researchers in drug discovery.
• This approach enhances the prediction of drug efficacy and safety profiles.