Skin Sensitization Study by Quantitative Structure-Activity Relationships (QSAR)

Kazuhiro Sato1, Tomohiro Umemura1, Tarou Tamura1, Yukinori Kusaka1, Kohji Aoyama2, Atsushi Ueda3, Kohichi Harada3, Keiko Minamoto3, Takemi Otsuki4, Kunihiko Yamashita5, Tatsuya Takeshita6, Eiji Shibata7, Kunio Dobashi8, Satomi Kameo9, Muneyuki Miyagawa10, Masaaki Kaniwa11, Yoko Endo12, and Kohtaro Yuta13,14

1Department of Environmental Health, School of Medicine, University of Fukui, Fukui, Japan,
2Department of Environmental Medicine, Kagoshima University, Graduate School of Medical and Dental Sciences, Kagoshima, Japan,
3Department of Environmental Health, Faculty of Medical and Pharmaceutical Sciences, Kumamoto University, Kumamoto, Japan,
4Department of Hygiene, Kawasaki Medical School, Kurashiki, Japan,
5Daicel Chemical Industries, Ltd., Himeji, Japan,
6Department of Public Health, School of Medicine, Wakayama Medical University, Wakayama, Japan,
7Department of Hygiene, Aichi Medical University, Nagakute, Japan,
8School of Health Sciences, Faculty of Medicine, Gunma University, Maebashi, Japan,
9Department of Public Health, Gunma University, Maebashi, Japan,
10National Institute of Occupational Safety and Health, Kawasaki, Japan,
11Division of Medical Devices, National Institute of Health Sciences, Tokyo, Japan,
12Research Center for Occupational Poisoning, Tokyo Rosai Hospital, Tokyo, Japan,
13Fujitsu Limited, Tokyo, Japan (14present: Research Center for Environmental Risk, National Institute for Environmental Studies, Tsukuba, Japan

AATEX 14(3):940-946, 2009

In silico assessment of skin sensitization is increasingly needed owing to the problems concerning animal welfare, as well as excessive time consumed and cost involved in the development and testing of new chemicals. Skin sensitization positive/negative prediction models with discriminant function were gener-ated and parameter analysis was discussed on the basis of QSAR technology.
Samples used in this research were selected from the list of "Maximale Arbeitsplatz-Konzentration" (MAK) and "Biologischer Arbeitsstoff-Toleranz-Wert" (BAT) values 2008, Deutschen Forschungsge-meinschaft (DFG) for positive samples (skin sensitizers) and from the classification results of the Japanese Globally Harmonized System of Classification and Labeling of Chemicals (GHS) Inter-ministerial Committee of the National Institute for Technology and Evaluation for negative skin sensitizers (controls). A total of 291 compounds (122 positive sensitizers and 169 negative sensitizers) were used in this study. Parameters were generated from 2-D and 3-D structures of compounds. All of the approximately 800 parameters generated were reduced to 47 parameter sets and 32 parameter sets by feature selection. Vari-ous linear and non-linear discriminant analysis methods were applied using 2 parameter sets. All data analyses were performed using ADMEWORKS/ModelBuilder software.
Perfect classification ratios (100%) were achieved using Support Vector Machine and AdaBoost for 32 parameters. The highest prediction ratio of 81.44% by Leave-Ten-Out Cross-Validation was achieved with Neutral Network for 47 parameter sets. Log P was not found to be important.
This is the first QSAR model for skin sensitization from Japan. Future studies of this QSAR model are needed to improve its efficacy.

Key wards: skin sensitization, QSAR, animal study, occupational exposure limit


(AATEX: Altern. Animal Test. EXperiment.: Alternatives to Animal Testing and EXperimentation)