Please wait a minute...
Data Analysis and Knowledge Discovery
Current Issue | Archive | Adv Search |
Building A Drug ADMET Classification Prediction Model Based on Graph Attention Network
Gu Yao-wen,Zhang Bo-wen,Zheng Si,Yang Feng-chun1,Li Jiao
(Institute of Medical Information, Chinese Academy of Medical Sciences, Beijing 100020,China) (XtalPi AI Research Center, Beijing 100089, China)
Export: BibTeX | EndNote (RIS)      

[Objective] This study aims to build a model for drugs’ ADMET properties (Absorption, Distribution, Metabolism, Excretion, Toxicity), which can evaluate drug properties in virtual screening.

[Methods] We proposed a Graph Attention Network (GAN) based method for drug ADMET classification prediction.  We constructed molecular graphs as the molecular structure features, with integrating drug ADMET properties from open databases and scientific publications.  . Furthermore, our GAN-based model was compared with three machine learning models and two graph neural network models.

[Results] We collected and integrated 149,457 available ADMET data with result of 9 datasets. The ADMET prediction model based on the graph attention network has an average accuracy of 0.825 and an average F1-Score of 0.67 in the 9 datasets. Compared with the baseline models of machine learning and graph and neural network, the maximum improvement of ours in average Accuracy and average F1-score were 6.3% and 26.0%.

[Limitations] Data cleaning steps can be refined, and the model prediction performance can be further improved by designing a pre-training architecture.

[Conclusions] The GAN-based model proposed in this study is well-performed in drug ADMET prediction, and it can be used in virtual drug screening process to provide reference for computer-aided drug design and drug discovery.

Key words Graph Neural Network      Graph Attention Network      Multi-source Heterogeneous Data      ADMET      Virtual Screening      
Published: 29 June 2021
ZTFLH:  R961  

Cite this article:

Gu Yao-wen, Zhang Bo-wen, Zheng Si, Yang Feng-chun, Li Jiao. Building A Drug ADMET Classification Prediction Model Based on Graph Attention Network . Data Analysis and Knowledge Discovery, 0, (): 1-.

URL: 2021.0233     OR

[1] Gu Yaowen, Zhang Bowen, Zheng Si, Yang Fengchun, Li Jiao. Predicting Drug ADMET Properties Based on Graph Attention Network[J]. 数据分析与知识发现, 2021, 5(8): 76-85.
[2] Wang Song, Yang Yang, Liu Xinmin. Discovering Potentialities of User Ideas from Open Innovation Communities with Graph Attention Network[J]. 数据分析与知识发现, 2021, 5(11): 89-101.
  Copyright © 2016 Data Analysis and Knowledge Discovery   Tel/Fax:(010)82626611-6626,82624938