Function Prediction of Human Proteins Using Machine Learning Algorithms

Abstract

Due to the rapid pace of research, accumulation of biological data is happening at an overwhelming rate. Sophisticated computation techniques are required to extract the useful information from this enormous amount of protein data such that the knowledge is practically useful and easily interpretable. For instance, drug discoverers need biological or computational methods to predict the functions of proteins, responsible for various diseases in human body. Since traditional biological methods were time consuming and comparatively expensive, various computational methods have been introduced in the respective research areas. In this research, we have tried to generate machine learning models that predict the protein function of unknown proteins and analyse their performance to obtain a model with highest accuracy. Sequence annotations such as Molecule Processing, Amino Acid modifications and other structural features like Active Site, Beta strand, Chain, etc. along with protein mass and length have been considered for prediction of protein functions. Feature selection has been performed in order to further improve the accuracy. Classification of proteins into 6 groups has been done according to the enzyme nomenclature scheme.