Identification of differentially expressed subnetworks based on multivariate ANOVA.

Area of Science:

• Bioinformatics
• Systems Biology
• Genomics

Background:

• High-throughput protein-protein interaction (PPI) data is increasingly integrated with genome-wide data.
• Identifying phenotype-related PPI subnetworks using gene expression data is a key research area.
• Effective integration requires robust search algorithms and scoring methods.

Purpose of the Study:

• To propose a novel multivariate analysis of variance (MANOVA)-based scoring method.
• To integrate protein-protein interaction (PPI) data with gene expression data.
• To identify phenotype-related differentially expressed PPI subnetworks.

Main Methods:

• A MANOVA-based scoring method was developed.
• A greedy search algorithm was employed to identify subnetworks with maximum scores.
• The method was applied to human microarray datasets.

Main Results:

• The MANOVA-based method successfully identified phenotype-related functional pathways and complexes.
• Compared to t statistic and mutual information methods, MANOVA yielded larger subnetworks.
• Subnetworks identified by MANOVA consisted of highly correlated proteins.

Conclusions:

• A MANOVA-based scoring method combined with a greedy search effectively integrates PPI and expression data.
• This method is recommended for sensitive detection of large, significant PPI subnetworks.
• The identified subnetworks are linked to phenotype-related pathways and complexes.