Related Experiment Video
Updated: Sep 23, 2025
09:47
Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
1.3K
Interpretable-ADMET: a web service for ADMET prediction and optimization based on deep neural representation
Yu Wei1, Shanshan Li1,2, Zhonglin Li1,2
1State Key Laboratory of Medicinal Chemical Biology, Frontiers Science Center for Cell Responses, College of Pharmacy and Tianjin Key Laboratory of Molecular Drug Research, Nankai University, Tianjin 300353, China.
Bioinformatics (Oxford, England)
|May 13, 2022
Summary
This study introduces interpretable-ADMET, a web service predicting 59 absorption, distribution, metabolism, excretion, and toxicity properties. It identifies key molecular substructures and generates novel drug candidates for optimized lead discovery.
Area of Science:
- Computational chemistry
- Drug discovery
- Pharmacology
Background:
- Lead compound discovery and optimization is challenging for non-experts to intuitively understand substructure contributions to molecular properties.
- Accurate prediction of absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties is crucial in drug development.
Purpose of the Study:
- To develop a user-friendly web service, interpretable-ADMET, for predicting and interpreting ADMET properties.
- To aid non-expert pharmacologists in understanding structure-property relationships.
- To facilitate the optimization of lead compounds in drug discovery.
Main Methods:
- Development of a web service utilizing graph convolutional neural network and graph attention network algorithms.
- Implementation of 90 qualitative classification and 28 quantitative regression models.
- Integration of gradient-weighted class activation mapping for substructure identification.
- Inclusion of an optimization module based on matched molecular pair rules.
Main Results:
- The interpretable-ADMET service predicts 59 ADMET-associated properties for 80,167 chemical compounds, with 250,729 associated entries.
- Provides interpretable models to identify substructures crucial for specific properties.
- Offers an automated module for generating novel virtual drug candidates.
Conclusions:
- Interpretable-ADMET serves as a valuable tool for drug discovery, particularly in the lead optimization phase.
- The service enhances the ability to interpret structure-property relationships and generate optimized compounds.
- Facilitates a more intuitive understanding of molecular contributions to ADMET profiles.
