Research on data mining from protein structure databases

点击翻译按钮获取中文摘要

We have developed some methods of assisting knowledge discovery from protein structure database, which include three-dimensional data processing, structure data mining, text mining, and information filtering for protein functional analysis.1. Profile extraction from protein structure data and its application to retrieval of protein with similar interactionSince function of a protein depends on the interaction in its local site, a profile is automatically generated from information of location and properties of atoms in the interaction sites of proteins with the same interaction. Profile-based retrieval method has also been explored, in which the pattern matching between generated profile and protein structure data is introduced.2. Surface motif discovery from protein surface dataA method of mining protein molecular surface database for extracting surface motifs, which are patterns of shape and physical properties that commonly appear in protein molecular surfaces, has been proposed. We also have proposed a method of filtering surface motifs by using TF-IDF evaluation and negative instances of interaction sites.3. Protein functional information extraction from literature by using structure dataA method of extracting protein functional information from literature related to protein structure analysis has been proposed. In this method, interaction-related sentences are selected by using distance between protein and its partner based on structure data, and then, functional information is extracted by text processing and iterative learning using feedback information from the operator.