A Classification-based Approach to Prediction of Dengue Virus and Human Protein-Protein Interactions using Amino Acid Composition and Conjoint Triad Features

• Main Authors: Dey, Lopamudra, Mukhopadhyay, Anirban
• Format: Conference Proceeding
• Language: English
• Subjects: amino acid composition; conjoint triad; Dengue; KNN; protein-protein interaction; SVM
• ISSN: 2642-6102
• DOI: 10.1109/TENSYMP46218.2019.8971382

Dengue Virus (DENV) is one of the most significant mosquito-borne viral infections of humans. Accoding to WHO, every year more than 50 million people are affected by DENV, resulting in 20,000 deaths. Protein-protein interactions play an important role in the cellular process of dengue virus infection in the human body. Although recently some studies have predicted protein-protein interactions (PPI) between human and DENV, many still remain to be identified. In this paper, we have predicted PPI between dengue and its human host combining amino acid composition and the conjoint triad of human protein sequences as a feature vector. We have concentrated on three well-known supervised machine learning methods, viz., Support Vector Machine (SVM), Naive Bayes (NB) and K-nearest neighbor (KNN) for prediction. SVM-based method achieved better accuracy, specificity and F1 score over other algorithms. Subsequently, unknown target human proteins of DENV infection are predicted using SVM to generate 411 new dengue-human PPI. Furthermore, the Gene Ontology (GO) and KEGG pathway of these predicted human proteins are analyzed. Both known and predicted human proteins share similar GO annotations. The pathways of the predicted human proteins are also found to be supported by recent literature. Identification of such interactions may accelerate a way for predicting new drugs to prevent dengue-related diseases.