Data mining techniques to predict protein secondary structures

Abstract

Protein secondary structure prediction is a key step in prediction of protein tertiary structure. There have emerged many methods based on machine learning techniques, such as neural networks (NN) and support vector machines (SVM), to focus on the prediction of the secondary structures. In this paper a new method, DM-pred, was proposed based on a protein clustering method to detect homologous sequences, a sequential pattern mining method to detect frequent patterns, features extraction and quantification approaches to prepare features and SVM method to predict structures. When tested on the most popular secondary structure datasets, DM-pred achieved a Q3 accuracy of 78.20% and a SOV of 76.49% which illustrates that it is one of the top range methods for protein secondary structure prediction.