Classifying Antiviral Peptides through the Application of Classification Algorithms

The functions of peptides have been studied in biology for many years. Scientists have constantly used different peptides with properties ranging from preventing dementia to helping with cell communication. In recent years, an emphasis on the use of antiviral peptides have become a huge part of medicine, slowly leading to breakthroughs in cures for viruses such as HIV. In this paper, peptides were predicted to have antiviral capabilities or not through the use of a k-mer bag of words model to have the data highlight each amino acid separately. From there, four models that consisted of a neural network, a decision tree, a random forest, and a logistic regression model were used to predict if the selected peptides were antiviral using the prevalence of each amino acid as a basis. As a result, it was found that certain amino acids create a higher probability of a peptide being antiviral, and the models used were able to predict with a high degree of accuracy if the sequence indicated an antiviral peptide or not. These results are open to additional testing and subjugation before consideration in practical applications in vaccine development or the investigation of different viruses. Once adjusted however, scientists will be able to predict the relationship between amino acid sequence and peptide properties, and from there will use that research to come up with new cures and breakthroughs in modern medicine.