ADMETlab 3.0: an updated comprehensive online ADMET prediction platform enhanced with broader coverage, improved...

Area of Science:

• Computational chemistry
• Medicinal chemistry
• Drug discovery

Background:

• The drug discovery process requires efficient evaluation of Absorption, Distribution, Metabolism, Excretion, and Toxicity (ADMET) properties.
• Previous versions of ADMETlab offered valuable tools but had limitations in coverage and performance.
• There is a growing need for accessible and robust platforms to predict drug-like properties.

Purpose of the Study:

• To introduce ADMETlab 3.0, an updated web server for comprehensive ADMET and physicochemical property evaluation.
• To address the limitations of previous versions by enhancing coverage, performance, and functionality.
• To provide a robust platform with API access and uncertainty estimates for improved drug candidate selection.

Main Methods:

• ADMETlab 3.0 utilizes a multi-task Deep Molecular Probabilistic Neural Network (DMPNN) architecture integrated with molecular descriptors.
• The platform incorporates 119 features, a 31-feature increase from the previous version.
• An Application Programming Interface (API) has been implemented for programmatic data access.

Main Results:

• ADMETlab 3.0 demonstrates superior performance in accuracy and robustness for predicting ADMET parameters.
• The server offers simultaneous calculation of endpoints with high speed.
• The dataset size has increased 1.5 times, now exceeding 400,000 entries.
• Uncertainty estimates are provided with prediction results.

Conclusions:

• ADMETlab 3.0 represents a significant advancement in computational tools for drug discovery.
• The platform provides enhanced decision support for identifying promising drug candidates.
• Public access to ADMETlab 3.0 is available without registration, facilitating broader adoption.