2005 Fiscal Year Final Research Report Summary
Study on Nonlinear Methods for Structure Activity Relationship in Assessment of Health Effects of Chemical Substances
| Project/Area Number |
14209022
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (A)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
広領域
|
|---|
| Research Institution | Chiba Institute of Technology |
|---|
Principal Investigator |
TANABE Kazutoshi Chiba Institute of Technology, Department of Management Information Science, Professor, 社会システム科学部, 教授 (90344134)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
NAGASHIMA Umpei National Institute of Advanced Industrial Science and Technology, Research Institute of Computational Sciences, Senior Researcher, 計算科学研究部門, 総括研究員 (90164417)
UCHIMARU Tadafumi National Institute of Advanced Industrial Science and Technology, Research Institute of Computational Sciences, Chief Researcher, 計算科学研究部門, 主任研究員 (00151895)
TSUZUKI Seiji National Institute of Advanced Industrial Science and Technology, Research Institute of Computational Sciences, Chief Researcher, 計算科学研究部門, 主任研究員 (10357527)
MATSUMOTO Takatoshi Tohoku University, Institute of Multidisciplinary Research for Advanced Materials, Assistant, 多元物質科学研究所, 助手 (50343041)
NAKATA Munetaka Tokyo University of Agriculture and Technology, Graduate School of Bio-Applications and Systems Engineering, Professor, 大学院生物システム応用科学研究科, 教授 (40143367)
|
|---|
| Project Period (FY) |
2002 – 2005
|
| Keywords | Structure Activity Relationship / Neural Network / Carcinogenicity Prediction |
|---|
| Research Abstract |
The purpose of this study is to develop a method for predicting the carcinogenicity of diverse chemical substances only from information of their molecular structures on the basis of quantitative structure-activity relationship (QSAR). A three-layered neural network as a nonlinear QSAR model was constructed. For the 454 compounds used in the Predictive Toxicology Challenge (PTC) 2000-2001 contest, 37 kinds of molecular descriptors calculated with MO programs, and the carcinogenicity data were entered into the input and output layers, respectively. The data of 454 compounds was split into training (144 compounds), validation (143) and test (167) sets. To solve the problems such as over-training, over-fitting and local minimum in training the neural network with the error-back-propagation algorithm, various conditions of the network such as the training cycles and neuron numbers of the intermediate layer were optimized. The optimum model showed a correct classification rate close to 74 %, higher than any of the PTC contestants. In order to develop a method with higher predictability, experimental carcinogenicity data on about 400 compounds were collected from various sources such as IARC, NTP and others, and their reliabilities were ranked into nine categories. 70 kinds of molecular descriptors were calculated from their 3D structures for these compounds, and the relationship between carcinogenicity data and those descriptors was analyzed. The performance of the proposed model was assessed by applying the leave-one-out cross validation test. It was found that this method can predict the relative carcinogenicity of diverse chemicals with higher accuracy than those of existing methods.
|
