Classifying Protein Sequences Using Convolutional Neural Network

Abstract

One of the major challenges in bioinformatics is the classification and identification of protein structure and function. Large amounts of Ribonucleic Acid (RNA) data cannot be managed using traditional laboratory methods. For this, proteins should be separated according to their structure and families. Therefore, proteins need to be classified to define their biological families and functions. In traditional machine learning approaches, various feature extraction algorithms are used to classify proteins. In manual feature extraction, the selected features directly affect performance. Therefore, in the proposed method of this study, protein sequences were digitized by the amino acid composition technique. The digitized protein sequences were converted to spectrograms, and automatic feature extraction was performed using two-dimensional Convolutional Neural Network (CNN) models (VGG19, ResNet). The extracted features were classified with Support Vector Machine (SVM) and k-Nearest Neighbors (k-NN). As a result, the accuracy of 95.03% was achieved in the classification of protein sequences using ResNet.