Integration of differential expression and network structure for 'omics data analysis.

Differential expression (DE) analysis has been routinely used to identify molecular features that are statistically significantly different between distinct biological groups. In recent years, differential network (DN) analysis has emerged as a powerful approach to uncover molecular network structure changes from one biological condition to the other where the molecular features with larger topological changes are selected as biomarkers. Although a large number of DE and a few DN-based methods are available, they have been usually implemented independently. DE analysis ignores the relationship among molecular features while DN analysis does not account for the expression changes at individual level. Therefore, an integrative analysis approach that accounts for both DE and DN is required to identify disease associated key features. Although, a handful of methods have been proposed, there is no method that optimizes the combination of DE and DN. We propose a novel integrative analysis method, DNrank, to identify disease-associated molecular features that leverages the strengths of both DE and DN by calculating a weight using resampling based cross validation scheme within the algorithm. First, differential expression analysis of individual molecular features is carried out. Second, a differential network structure is constructed using the differential partial correlation analysis. Third, the molecular features are ranked in the order of their significances by integrating their DE measures and DN structure using the modified Google's PageRank algorithm. In the algorithm, the optimum combination of DE and DN analyses is achieved by evaluating the prediction performance of top-ranked features utilizing support vector machine classifier with Monte Carlo cross validation. The proposed method is illustrated using both simulated data and three real data sets. The results show that the proposed method has a better performance in identifying important molecular features with respect to predictive discrimination. Also, as compared to existing feature selection methods, the top-ranked features selected by our method had a higher stability in selection. DNrank allows the researchers to identify the disease-associated features by utilizing both expression and network topology changes between two groups.