A Binary Classifier for Prediction of the Types of Metabolic Pathway of Chemicals

Author(s): Yemin Fang, Lei Chen*

Journal Name: Combinatorial Chemistry & High Throughput Screening
Accelerated Technologies for Biotechnology, Bioassays, Medicinal Chemistry and Natural Products Research

Volume 20 , Issue 2 , 2017

Become EABM
Become Reviewer


Background: The study of metabolic pathway is one of the most important fields in biochemistry. Good comprehension of the metabolic pathway system is helpful to uncover the mechanism of some fundamental biological processes. Because chemicals are part of the main components of the metabolic pathway, correct identification of which metabolic pathways a given chemical can participate in is an important step for understanding the metabolic pathway system. Most previous methods only considered the chemical information, which tried to deal with a multilabel classification problem of assigning chemicals to proper metabolic pathways.

Methods: In this study, the pathway information was also employed, thereby transforming the problem into a binary classification problem of identifying the pair of chemicals and metabolic pathways, i.e., a chemical and a metabolic pathway was paired as a sample to be considered in this study. To construct the prediction model, the association between chemical pathway type pairs was evaluated by integrating the association between chemicals and association between pathway types. The support vector machine was adopted as the prediction engine.

Results: The extensive tests show that the constructed model yields good performance with total prediction accuracy around 0.878.

Conclusion: The comparison results indicate that our model is quite effective and suitable for the identification of whether a given chemical can participate in a given metabolic pathway.

Keywords: Metabolic pathway, chemical-chemical interaction, bipartite graph, chemical-chemical interaction, Kuhn-Munkres algorithm, protein-protein interaction, support vector machine.

Rights & PermissionsPrintExport Cite as

Article Details

Year: 2017
Page: [140 - 146]
Pages: 7
DOI: 10.2174/1386207319666161215142130
Price: $65

Article Metrics

PDF: 12