[Perspective of predictive toxicity assessment of in vivo repeated dose toxicity using structural activity relationship].

Tens of thousands of existing chemicals have been widely used for manufacture, agriculture, household and other purposes in worldwide. Only approximately 10% of chemicals have been assessed for human health hazard. The health hazard assessment of residual large number of chemicals for which little or no information of their toxicity is available is urgently needed for public health. However, the conduct of traditional toxicity tests which involves using animals for all of these chemicals would be economically impractical and ethically unacceptable. (Quantitative) Structure-Activity Relationships [(Q)SARs] are expected as method to have the potential to estimate hazards of chemicals from their structure, while reducing time, cost and animal testing currently needed. Therefore, our studies have been focused on evaluation of available (Q)SAR systems for estimating in vivo repeated toxicity on the liver. The results from our preliminary analysis showed the distribution for LogP of the chemicals which have potential to induce liver toxicity was bell-shape and indicating the possibility to estimate liver toxicity of chemicals from their physicochemical property. We have developed (Q)SAR models to in vivo liver toxicity using three commercially available systems (DEREK, ADMEWorks and MultiCASE) as well as combinatorial use of publically available chemoinformatic tools (CDK, MOSS and WEKA). Distinct data-sets of the 28-day repeated dose toxicity test of new and existing chemicals evaluated in Japan were used for model development and performance test. The results that concordances of commercial systems and public tools were almost same which below 70% may suggest currently attainable knowledge of in silico estimation of complex biological process, though it possible to obtain complementary and enhanced performance by combining predictions from different programs. In future, the combinatorial application of in silico and in vitro tests might provide more accurate information which support regulatory decisions. At the same time, an appropriate strategy to use (Q)SAR for of the efficiency and accuracy in chemical management is necessary.