Using Quadratic Discriminant Analysis to Predict Protein Secondary Structure Based on Chemical Shifts

Background: Prediction of the protein three-dimensional structure is one of the most important and hot topics in the field of bioinformatics. However, the prediction of the secondary structure of a protein from its amino acid’s sequence is an important step towards the prediction of its three-dimensional structure. Many approaches have been proposed for the prediction of protein secondary structure and yielded better results. However, these algorithms were primarily based on the features of the amino acid sequences. Objective: In this paper, we introduced a new model for predicting the secondary structure of proteins. Method: We used chemical shifts as a novel feature and combined with the quadratic discriminant analysis method in predicting the secondary structure of proteins. Results: Finally, the three-state overall prediction accuracy of 85.7% was obtained in the ten-fold crossvalidated test, and the accuracies of alpha helices, beta stands and coil reached 95.2%, 83.7%, 77.8% respectively. Moreover, to determine the importance of chemical shifts of six nuclei, we used the leave one out feature and combined another five nuclei as features, the results showed that the chemical shift of each nuclei play a different role in the prediction of protein secondary structure, and the maximum overall accuracy reached 87.3% (Q3) in using C Cα Cβ Hα N as features. Conclusion: Our model outperformed other state-of-the-art method in term of predictive accuracy. Our results showed that the quadratic discriminant analysis method by using chemical shifts as features is indeed a good choice for protein secondary structures.