Development of Reaction Prediction System (RESUK)

Abstract

We have been developing three different types of organic synthesis design systems that contribute to organic chemist's creative research. The AIPHOS (Artificial Intelligence for Planning and Handling Organic Synthesis) is a superior system to create new synthetic pathways using the advantages of empirical knowledge-based approach and logic-centered approach. The KOSP (Knowledgebase-Oriented Synthesis Planning) uses strategies similar to the AIPHOS. The KOSP has a function of recognizing a strategy site by using a knowledgebase converted from the reaction database (RDB). Thus the KOSP is more empirical system than the AIPHOS. The TOSP (Transform-Oriented Synthesis Planning) is empirical approach type system that uses a transform database built from existing reactions. On the other hand, we have been developing reaction evaluation system that predicts reaction products by using a transform database abstracted from the RDB. The reaction evaluation system can predict reaction products about a synthetic precursor proposed by organic synthesis design systems. From the result, we can check reaction products other than a target product. And this system can evaluate synthetic pathways based on the safety and the generating condition of by-product. We have developed a client/server (C/S) system that integrated three organic synthesis design systems mentioned above and the reaction evaluation system. This C/S system can search synthetic pathways by using each system through a common interface. The purpose of this work is to improve the convenience of reaction prediction of the C/S system by developing a function of reaction prediction between two reactants, reaction formula sorting function and the function of multi-step reaction prediction.