Protein Tertiary Structure Prediction Based on Multiscale Recurrence Quantification Analysis and Horizontal Visibility Graph

Abstract

As protein types continue to increase, there are more and more methods for predicting protein structure. In this paper, a feature extraction method based on multiscale coarse-grained time series recurrence quantification analysis and horizontal visibility graph is proposed. First, the chaos game representation is used to map the protein secondary structure sequence into two time series. Multiscale coarse granulation time series. Then feature extraction by combining recurrence quantification analysis and horizontal visibility graph. Thereby a 30-D feature vector is obtained. This paper uses support vector machine to predict protein tertiary structure. In this paper, the prediction results of the two low homologous protein datasets were 95.33% and 93%, respectively.