Prediction And Application Of Interaction Based On Sparse Representation

Protein as one of the basic substances of life,protein-protein interactions exist in various life activities,the important life activities are achieved through the interaction between proteins,such as metabolism of cell,regulation of hormone and enzyme catalysis.Therefore,the study of protein interactions is not only conducive to understanding the mechanism of life,but also of great significance to the development of new drugs.However,using traditional biological experiments to identify protein-protein interactions is not only costly,but also has a high rate of false positives and false negatives.With the rapid development of information technology,using intelligent computing to predict protein interactions is not only fast,but also highly accurate.It has become an important method for predicting protein interactions at present.The main works are as following:(1)Methods based on the feature extraction matrix.In order to efficiently and comprehensively extracted protein sequence important eigenvector matrix,the matrix used herein are principal component analysis and linear discriminant analysis and matrix protein sequence information through the numerical representation of feature extraction.The experimental comparison between H.pylori,S.cerevisiae and Human datasets with other feature extraction methods proves the effectiveness of the proposed feature extraction method.(2)Design of prediction model based on sparse representation.This paper uses sparse representation model and weighted sparse representation model to realize the function of protein interaction prediction.Through comparison experiments with support vector machines and other models,it is found that the model based on sparse representation can be well combined with the feature extraction method used in this paper and has excellent prediction results.(3)In order to meet the needs of researchers with weak computer programming skills,I designed and developed a complete protein interaction prediction system based on sparse representation using MATLAB GUI technology for everyone to learn and communicate.