Related Experiment Video
Updated: Jun 29, 2025
11:38
High Content Screening Analysis to Evaluate the Toxicological Effects of Harmful and Potentially Harmful Constituents HPHC
Published on: May 10, 2016
12.2K
Graph Convolutional Network-Enhanced Model for Screening Persistent, Mobile, and Toxic and Very Persistent and Very
Qiming Zhao1, Yuting Zheng2, Yu Qiu1
1College of Environmental and Resource Sciences, and Women's Hospital, School of Medicine, Zhejiang University, Hangzhou 310058, China.
Environmental Science & Technology
|April 1, 2024
Summary
A new graph convolutional network (GCN) model enhances screening for persistent, mobile, and toxic (PMT) and very persistent and very mobile (vPvM) substances. This tool offers high accuracy for efficient environmental risk assessment.
Area of Science:
- Environmental Chemistry
- Computational Chemistry
- Toxicology
Background:
- Global chemical management faces challenges from emerging contaminants, particularly persistent, mobile, and toxic (PMT) and very persistent and very mobile (vPvM) substances.
- Existing screening models for PMT/vPvM substances are limited by small datasets, simplistic features, and conventional algorithms, hindering robust predictions.
- There is an urgent need for accessible, high-throughput screening tools to manage these hazardous chemicals effectively.
Purpose of the Study:
- To develop a robust and widely applicable computational model for the high-throughput screening of PMT/vPvM substances.
- To integrate molecular graph structures and descriptors for improved prediction accuracy.
- To create a publicly available online platform for assessing PMT/vPvM properties of chemical substances.
Main Methods:
- Construction of a graph convolutional network (GCN)-enhanced model incorporating feature fusion of molecular graphs and descriptors.
- Training the model on a large dataset of 213,084 substances adhering to the latest PMT classification criteria.
- Utilizing kernel density estimation to assess the model's application domains and suitability for high-throughput screening.
Main Results:
- The GCN-enhanced model achieved a high accuracy rate of 86.6% and a low false-negative rate of 6.8% for PMT/vPvM substance screening.
- The model effectively leverages the correlation between critical molecular descriptors and PMT/vPvM properties.
- An online server, the PMT/vPvM profiler, was developed with a user-friendly web interface (http://www.pmt.zj.cn/) for public access.
Conclusions:
- The developed GCN-enhanced model provides a significant advancement in the efficient evaluation of PMT/vPvM substances.
- The publicly accessible PMT/vPvM profiler facilitates global high-throughput screening and management of these hazardous chemicals.
- This study addresses the need for robust screening tools, contributing to strengthened global chemical management strategies.
