Drug-Target Interaction Prediction Based on MACLDNN

View at publisher Summarise with VIRAT
Authors: Bingying Yao, Jianqiang Chen, Qing Ye
Cited by: 0
References: 33

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Drug-target interaction prediction can effectively predict potential targets of drugs or potential active drugs for targets of interest. However, the number of drug-target pairs formed by connecting drugs and targets far exceeds the of drug-target interactions, which makes it easy for DNN models for drug-target interaction prediction to overfit. To overcome the above problem, a drug-target interaction prediction method based on multi-auxiliary classification layer DNN is adopted. The multi-auxiliary classification layer DNN (MACLDNN) adds multiple auxiliary classification layers between the hidden layers, trains the parameters through multiple losses, increases the recognition of the middle and lower layers in the classifier, enhances the gradient signal of backpropagation, and provides additional regularization. In the experiment of finding effective drugs acting on new targets among known drugs, the average AUC of the multi-auxiliary classification layer DNN is 0.88%, 3.11%, 1.75%, and 2.95% higher than the best DNN-based comparative method on GPCR, IC, PF, and DB datasets, which verifies the effectiveness of the proposed method.

Fields of Study & Subjects

Computational Drug Discovery Methods Machine Learning in Bioinformatics Bioinformatics and Genomic Networks

Publication details

Journal / source
IEEE International Conference on Power, Electronics and Green Energy (ICPEGE)
Pages
1155–1161
Publication date
28 January 2026
Language
EN