Prediction Of Protein-protein Interaction And DNA Methylation Sites Based On Increment Of Diversity Combined With Quadratic Discriminant Analysis

Protein is the material basis of all life. The study of protein-protein interactions is helpful to understand the molecular mechanism of biological processes and diseases. In this article, based on the compositional characteristics of protein sequences, the method of increment of diversity combined with quadratic discriminant analysis (IDQD) was used to predict human protein-protein interactions including 1 962 pairs. The results showed that the prediction accuracy values of self-consistent test and 3-fold cross-validation test are higher than 75% and 64%, respectively. The method was also used to predict Saccharomyces cerevisiae protein-protein interactions, which included 4 548 interaction pairs. The total accuracy (TA) and the correlation coefficient (CC) values of 5-fold cross-validation test are 61% and 0.22, respectively.At the same time, we used the increment of diversity combined with quadratic discriminant to predict the human DNA methylation sites. With a window size of 79 nucleotides (39 nucleotides at each side of the central cytosine) and the attribute of 8-mer frequencies, the IDQD module performed very well and better than other models. This module achieved a TA and CC of 80.03% and 0.60, respectively, when evaluated using a 5-fold cross-validation.The results of the two works showed that the method of increment of diversity combined with quadratic discriminant analysis could be used to solve the problems of the complexity biology. It is a very good classifier.