2. Academic Validation
  3. Overfit deep neural network for predicting drug-target interactions

Overfit deep neural network for predicting drug-target interactions

  • iScience. 2023 Aug 15;26(9):107646. doi: 10.1016/j.isci.2023.107646.
Xiao Xiaolin 1 2 3,  Liu Xiaozhi 2 3,  He Guoping 4,  Liu Hongwei 5 6,  Guo Jinkuo 2 7,  Bian Xiyun 2 3,  Tian Zhen 8,  Ma Xiaofang 2 3,  Li Yanxia 2 3,  Xue Na 2 3,  Zhang Chunyan 2 3,  Gao Rui 2,  Wang Kuan 1,  Zhang Cheng 1,  Wang Cuancuan 1,  Liu Mingyong 2 9,  Du Xinping 1 2 7
Affiliations

Affiliations

  • 1 Department of Cardiology, Tianjin Fifth Central Hospital, Tianjin, China.
  • 2 Tianjin Key Laboratory of Epigenetics for Organ Development of Premature Infants, Tianjin Fifth Central Hospital, Tianjin, China.
  • 3 Central Laboratory, Tianjin Fifth Central Hospital, Tianjin, China.
  • 4 Geriatrics Department, Traditional Chinese Medicine Hospital of Binhai New Area, Tianjin, China.
  • 5 School of Clinical Medicine, North China University of Science and Technology, Tangshan, Hebei, China.
  • 6 Department of Anesthesiology, Tangshan Maternal and Child Health Hospital, Tangshan, Hebei, China.
  • 7 College of Food Science and Engineering, Tianjin University of Science & Technology, Tianjin, China.
  • 8 Deepwater Technology Research Institute, China National Offshore Oil Corporation, Tianjin, China.
  • 9 Department of Urology, Tianjin Fifth Central Hospital, Tianjin, China.
PMID: 37680476 DOI: 10.1016/j.isci.2023.107646
Abstract

Drug-target interactions (DTIs) prediction is an important step in drug discovery. As traditional biological experiments or high-throughput screening are high cost and time-consuming, many deep learning models have been developed. Overfitting must be avoided when training deep learning models. We propose a simple framework, called OverfitDTI, for DTI prediction. In OverfitDTI, a deep neural network (DNN) model is overfit to sufficiently learn the features of the chemical space of drugs and the biological space of targets. The weights of trained DNN model form an implicit representation of the nonlinear relationship between drugs and targets. Performance of OverfitDTI on three public datasets showed that the overfit DNN models fit the nonlinear relationship with high accuracy. We identified fifteen compounds that interacted with TEK, a receptor tyrosine kinase contributing to vascular homeostasis, and the predicted AT9283 and dorsomorphin were experimentally demonstrated as inhibitors of TEK in human umbilical vein endothelial cells (HUVECs).

Keywords

Biochemistry; Biological sciences; Mathematical biosciences; Structural biology.

Figures
Products