Enzyme Function Prediction using Deep Learning

Abstract

Understanding an Enzyme's function is one of the most crucial problem domains in computational biology. Enzymes are a key component in all organisms as they speed up key chemical reactions and help in fighting diseases and are even used in the manufacturing and pharmaceutical industries. Their applications are wide and therefore the necessity to discover new enzymatic proteins is economically rewarding. Wet lab experiments are time-consuming and resource expensive for determining these function(s). Due to this importance and the cumbersome nature of the experiments used to elucidate their function, we propose a computational approach to predict an enzyme's function up to the fourth level of the Enzyme Commission (EC) Number. Previous studies are unable to predict the enzyme function at the fourth level (precise function) mainly due to the lack of examples in each class. We use innovative deep learning approaches along with an efficient hierarchical classification scheme to predict an enzyme's function (EC number), up to the 4th level. On a dataset of 11,343 enzymes and 401 classes, we achieved an Accuracy and F1 Score of 85% and 80%, respectively, on the 4th level. Moreover, our approach classifies enzymes from level 0 to level 4 using only 22 models as compared to 146 models used by assigning a model to each class whose lower levels are predicted.