ISSN 1000-1239 CN 11-1777/TP

Journal of Computer Research and Development ›› 2019, Vol. 56 ›› Issue (9): 1881-1888.doi: 10.7544/issn1000-1239.2019.20180830

Previous Articles     Next Articles

Predicted Results Evaluation and Query Verification of Drug-Target Interaction

Yu Donghua1, Guo Maozu1,2,3, Liu Xiaoyan1, Cheng Shuang4   

  1. 1(School of Computer Science and Technology, Harbin Institute of Technology, Harbin 150001); 2(School of Electrical and Information Engineering, Beijing University of Civil Engineering and Architecture, Beijing 100044); 3(Beijing Key Laboratory of Intelligent Processing for Building Big Data, Beijing 100044); 4(Institute of Materials, China Academy of Engineering Physics, Mianyang, Sichuan 621900)
  • Online:2019-09-10
  • Supported by: 
    This work was supported by the National Natural Science Foundation of China (61871020, 61571164, 61671189), the Scientific Research Key Project of Beijing Educational Committee (KZ201810016019), and the Special Funds for Basic Research Business in Municipal Universities of Beijing Architectural University (X18197, X18198, X18203).

Abstract: The drug-target interaction prediction is one of the important assistant approaches in drug discovery and design, however, experimental identification and validation of potential drug-target encoded by the human genome is both costly and time-consuming. Therefore, querying and validating the predicted drug-target interaction in databases is an important assessment of prediction methods. In this paper, the query and validation method of drug-target interaction named as DTcheck (drug-target check) is developed and designed with Web spider based on KEGG, DrugBank, ChEMBL databases, which realizes efficient query and validation function for drug-target pair providing both KEGG DRUG ID and KEGG GENES ID. ID mapping function is also designed in DTcheck, which can map Uniprot ID from DrugBank and ChEMBL into KEGG GENE ID. DTcheck expands 907, 766, 458, 40 pairs of new drug-target interaction for Enzyme, IC (ion channel), GPCR (G-protein-coupled receptor), NR (nuclear receptor) standard datasets, respectively. Moreover, combined with query and validation result, the analysis of the prediction results of the BLM (bipartite local models) method shows that evaluation of Top N is more reasonable than AUC (area under curve) value for the prediction method of drug-target interaction. It also shows that the BLMd method with low AUC value is superior to the BLMmax method with high AUC value in predicting the drug-target interaction.

Key words: drug-target interaction prediction, query and validation, drug-target data, AUC evaluation, Top N evaluation

CLC Number: