Predict Molecular Interaction Network of Norway Rats Using Data Integration

Abstract

The emergence of systems biology enables us to simulate and analyze organism’s microscope features from the level of genome, proteome and interactome. This article utilized data integration method to predict molecular interaction network of Norway rat following the basic principles of systems biology. This research selects microarray related with cardiac hypertrophy, and built the downstream studies on 730 differentially expressed genes.4 heterogeneous kinds of data type including microarray expression, gene sequence, subcellular localization of protein and orthologous data are selected to make the overall model more comprehensive. After processed by specific algorithms, the 4 data types are transformed to 5 types of evidence: Pearson correlation coefficient, SVM model recognition, similarities between gene sequences, distance between proteins and orthologous alignment. A widely used machine learning algorithm, support vector machines (SVM) is introduced here to help deal with single evidence preparation and multiple evidence integration. This article finds that the prediction accuracy of data integration is obviously higher than that of single evidence. Data integration promised that heterogeneous data types could enhance each other’s advantages by weakening each other’s disadvantages so as to deliver more objective and comprehensive understanding of molecular interactions.