Development of a computational toxicology system that realizes toxicity prediction based on the mechanism of toxicity

<p>Computational toxicology in drug safety research is gaining importance. Computational toxicology provides a low-cost, high-throughput toxicity screening method used primarily in the very early stages of drug discovery. Computational toxicology can also expect to complement and extend traditional drug safety evaluations. In addition, although the Three Rs principle regarding the use of animals is required to respect in toxicity research, computational toxicology may contribute to the reduction of animals in toxicity research. </p><p>With the remarkable progress of computing resources and machine learning methodologies in recent years, computational toxicology approaches have diversified. These approaches include multivariate analysis/pattern recognition, which processes complex relationships among multiple factors, and expert systems that incorporate expert knowledge as rules. In particular, the latest technology, deep learning, is expected to enable diversification and deepening of toxicological interpretation using big data. On the other hand, in risk assessment and management of pharmaceuticals, it is necessary to interpret highly complicated/diverse toxicity mechanisms and utilize a large amount of data/knowledge obtained in related fields in an integrated manner. This knowledge also includes knowledge shared only among some experts as subjective and nonverbal tacit knowledge. Therefore, to effectively utilize the computational toxicology system, it expects to develop a new knowledge system that appropriately navigates the necessary knowledge from various systematic knowledge.</p><p>The presentation will introduce our efforts to develop the computational toxicology system. And the challenges and prospects of computational toxicology will discuss.</p><p>Acknowledgment: These researches are supported by AMED under Grant Number JP19nk0101103(Informa) and JP21nk0101111(DAIIA).</p>