Efficiency due to data balancing in the prediction of essential proteins

Abstract

Essential proteins are important for the survival of the organism. Those play a vital role to prepare antibiotics, disease diagnosis, understanding the organism structure, etc. That is why essential proteins now become one of the hot topics in current research. In recent times, many methods were proposed to identify the essential proteins. But the identification rate of the essential proteins is still low. Therefore, an efficient approach is required to identify the essential proteins with high performance. In this paper, a machine learning-based efficient method is proposed to identify the essential proteins. To identify the essential proteins, both balanced and imbalanced datasets are used here. For data balancing, different techniques are used to observe the efficiency of the methods. Both topological and biological features are used in this method. The S. Cerevisiae dataset is used to evaluate the proposed method. Another dataset of species E. coli is used to validate the performance of this method. Three classifiers such as Random Forest, LightGBM, XGBoost, and an ensemble method of these three classifiers are used to predict the essential proteins. The best performance is achieved from the balanced dataset from the SMOT+ENN and Random Forest classifier. The results are balanced both in accuracy and F1-score which is also best compared to the existing related methods.