Using Machine Learning Approaches for Computational Prediction of Human and Hepatitis C Virus Proteins

Abstract

Background:

Hepatitis C virus (HCV) infection affects around 170 million individuals worldwide, with an estimated 3% of the world’s population presently afflicted. More than 350,000 people are killed each year by HCV throughout Asia and the rest of the globe due to liver disorders such as cirrhosis, chronic hepatitis, and hepatocellular carcinoma (HCC). Understanding viral-host protein interactions are essential for understanding viral infection, disease etiology, and the development of innovative therapeutics. This is due to the inherent limits of laboratory techniques for finding host-virus protein-protein interactions (PPIs). There seems to be a strong computational effect on the research of cellular infection.

Materials and Methods:

In this study, we predicted the interaction between human and HCV proteins using an ensemble learning technique.Support vector machines (SVMs) nuclear liner and radial are the cornerstones of our model, as are K-Nearest Neighbors (KNN) and Random Forest (RF). Four different feature vectors were used to encode human and HCV proteins: the tripeptide composition (TPC), The composition of k-spaced acid pairs (CKSAAP), the amino acid autocorrelation-autocovariance (AAutoCor), and the conjoint triad (CT).

 Results:

The predictive power of the suggested technique is evaluated using a benchmark dataset that contains both consistently positive and negative PPIs. A support vector machine (Radial-SVM) model was used to predict which human proteins interact with HCV. To achieve accuracy and specificity of 84.9 and 88.3 percent, we employed tenfold cross-validation and principal component analysis (PCA).

Conclusion:

Our technique correctly predicts PPIs based on human and HCV proteins. The discovery of HCV-human protein interaction networks, enriched pathways, gene ontology, and functional categories has improved our knowledge of HCV infection.