Abstract—Clarifying the interaction between cytochrome P450 (P450) and food-related compounds that affect the metabolic activity of P450 allows to effectively predict the toxicity of food-related compounds. Currently, we are developing a fluorescence P450 sensor to evaluate the metabolic reactions of food-related compounds. The amount of fluorescent metabolic products for various P450 species can be easily measured using sensors.
In this paper, a method of constructing models to predict toxicity of food-related compounds from the amount of fluorescent metabolic product using a machine learning technique is proposed. Since the precision of the measurement of the amount of fluorescent metabolic products is not high enough to quantitatively predict the toxicity value, multiple regression analysis is not always appropriate. We consider the toxicity prediction problem to be a toxicity classification problem. In this framework, however, it is difficult to determine the precise boundary values for separating one toxicity class from another. In our proposed method, fuzzy sets are introduced not only to the attributes for dividing instances but also to the classes that give the prediction results to solve the difficulty of determining the boundary values. The performance of our proposed method was confirmed by comparing the mean error and the cosine similarity with those of other methods.

Index Terms—Fluorescence sensor, fuzzy decision tree, P450, toxicity prediction.

Daishi Yajima and Takenao Ohkawa are with the Graduate School of System Informatics, Kobe University, Kobe, Japan (e-mail: yajima@cs25.scitec.kobe-u.ac.jp, ohkawa@kobe-u.ac.jp ).
Kouhei Muroi is with the Graduate School of Agricultural Science, Kobe University, Kobe, Japan (e-mail: a0873_3826@yahoo.co.jp).
Hiromasa Imaishi is with the Research Center for Environmental Genomics, Kobe University, Kobe, Japan (e-mail: himaish@kobe-u.ac.jp).

 

Cite: Daishi Yajima, Takenao Ohkawa, Kouhei Muroi, and Hiromasa Imaishi, "Predicting Toxicity of Food-Related Compounds Using Fuzzy Decision Trees," International Journal of Bioscience, Biochemistry and Bioinformatics vol. 4, no. 1, pp. 33-38, 2014.