Predicting the sub-class of enzyme by applying graph mining based on sequence structure of enzyme

Abstract

This article is a description of the way to apply graph mining technology to disaggregate the amino acid sequence of the enzyme - belonging to the same already named sub-class - into a set of respective maximal frequent subgraphs. The subgraphs can have one or many vertexes. When predicting the sub-class of a new enzyme, one just needs to disaggregate the amino acid sequence of that enzyme, then matches it with each maximal frequent subgraph in the data base. The predicted sub-class is based on the one with the highest scores after matching. The test developed on the sub-class of Oxidoreductase EC 1.2.1.1 and Hydrolase EC 3.1.1.3 gave good results. It left us with the remark that when enlarging the scale of learning set, all the named enzymes should be chosen. This aims to create a set of maximal frequent subgraphs with high reliability.