SEQUENCE BASED ENZYME CLASSIFICATION USING NEURAL NETWORKS
No Thumbnail Available
Date
2021
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
UMT, Lahore
Abstract
As a group of vital biocatalysts, enzymes boost the efficiency of chemical reactions. In the mid of 20th century, the Enzyme Commission (EC) number system classified enzymes into 6 major classes which are oxidoreductases, transferases, hydrolases, lyases, isomerases, and ligases denoted by (EC1), (EC2), (EC3), (EC4), (EC5) and (EC6) respectively. In August 2018, the International Union of Biochemistry and Molecular Biology (IUBMB) added a new enzyme class Translocases(EC7) replacing ATPases (EC 3.6.3.-) after which it has become necessary to conduct researches according to the latest classification. There is a need to improve enzyme classification method because it is a prominent area of research providing us with the opportunity to study more the structure and function of enzyme molecule if we know the class of enzyme which requires strenuous efforts to determine using biological experiments. With the substantial advancement of computational power as well as ever-increasing data, deep learning techniques have become exceedingly popular owing to their splendid achievements in computer vision queries. Therefore, making different computational models to predict the enzyme class has become a suitable and viable process. In this research, a branch of RNN, bidirectional long short-term memory (BiLSTM) model has been proposed for this purpose with the overall accuracy of 86% which comes up with a good approach for researchers to further explore sequence classification and discover even better and easier way out to predict enzyme class using amino acid sequence only. An innovative enzyme class prediction data app has also been built using Streamlit framework.