Comprehensible Enzyme Function Classification Using Reactive Motifs with Negative Patterns.

The main objective of this research work is to build a comprehensible enzyme function classification by using high-coverage and high-precision reactive motifs. Reactive motifs are directly extracted from the protein active and binding sites. Main advantage of the reactive motifs is their high-coverage, however their generalization make them over-generalized with low-precision quality. In this paper, a method for generating reactive motifs with negative patterns is proposed. Reactive motifs with negative patterns are able to control the level of motif generalization. As result, non over-generalized reactive motifs with high-precision are generated. Each of the reactive motifs is associated with a specific enzyme function. They can directly predict enzyme function of an unknown protein sequence. Without use of a complex classification model, set of voting methods are proposed and used to construct a comprehensible enzyme function classification. Essential clues of the enzyme mechanism are provided to the biologist end-users.