DEVELOPMENT OF A GRAPH STRUCTURE DATAMINING METHOD AND IDENTIFICATION SYSTEM OF ACTIVE MOLECULE SUBSTRU CTURES

在第一财年，研究了图结构数据挖掘的理论框架，开发了活性分子子结构识别的原型系统。本文研究了图结构数据在计算机中的表示和特征图模式的搜索原理。然后，对化学中可以引入我们工作的技术进行了调查，这些技术反映在原型系统中。最后，通过癌变和诱变化学成分数据中的子结构提取，对原型系统的基本性能进行了评价。在下一财年，对图结构数据挖掘的框架进行了扩展，使其在计算时间和内存消耗方面更加高效，并开发了用于活性分子子结构识别的真实规模系统。开发了高效计算时间和内存消耗的算法，并与常规化学技术进行了比较，设计了实际系统的功能。然后，对实际系统的原理和算法进行了修改和扩展，使图结构数据挖掘能够对海量图结构数据进行挖掘。在最后一个会计年度，在上一年开发的实际系统的基础上，进一步研究了化学分析方面需要实现的功能，并在实际系统中增加了一些可以实现的功能。然后，从化学工程和计算理论的角度，对实际系统的实用性和广泛适用性进行了评价。通过这些评估，验证了所开发的实际系统的实用性和高性能。目前正在与工业界合作开发商业系统。

In the first fiscal year, the theoretical framework of graph structure data mining was investigated, and a prototype system for active molecule substructure identification was developed. In this work, the representation of graph structure data in computers and search principle of characteristic graph patterns are studied. Them, the survey of techniques in chemistry which can be introduced to our work has been conducted, and these techniques were reflected in the prototype system. Finally, the basic performance of the prototype system has been evaluated through the substructure extraction in carcinogenetic and mutagenetic chemical component data.In the nest fiscal year, the framework of the graph structure data mining was extended to be more efficient in terms of computation time and memory consumption, and the real scale system for active molecule substructure identification has been developed. The algorithm for the efficient computation time and memory consumption was developed, and under the comparison with the conventional techniques in chemistry, the function of the real system was designed. Then, the principle and the algorithm of the real system was modified and extended to enable the graph structure data mining on the massive graph structure data.In the final fiscal year, further functions desired to be implemented in the view of chemical analysis were investigated based on the real system developed in the former year, and some functions which can be implemented feasibly were added to the real system. Then, from the view points of the chemical engineering and the computational theory, the practicality and the wide applicability of the real system have been evaluated. Through these evaluations, the practical and high performance of the developed real sysem has been confirmed. The effort to develop commercial system under collaboration with industries is currently underway.