Automatic prediction of enzyme functions from domain compositions using enzyme reaction prediction scheme

Proteins perform most important biochemical reactions in organisms, such as the catalysis, signal transduction, and transport of nutrients. The urgent need of automatic annotation is due to the advent of high-throughput sequencing techniques in the post-genomic era. Proteins consist of domains which are elementary building units of protein folding, function, and evolution. The evidence of protein function is convincible to deduce from its domain composition. For enzyme function prediction, efficiency and reliability become more and more important in the recent researches. This study proposed an enzyme reaction prediction scheme with a learning model for enzyme function predictions to avoid the exponential enumeration problem of frequent item-sets in the association rule algorithm. Our work also contributed to the prediction of multiple reactions due to the nature of enzymes.