Identification of epistatic effects using a protein-protein interaction database.

Area of Science:

• Genetics
• Bioinformatics
• Molecular Biology

Background:

• Epistasis, or gene-gene interaction, is crucial for complex human traits but statistical and molecular definitions often diverge.
• Current statistical epistasis detection methods may not fully capture biological interactions at the gene product or DNA level.
• High-dimensional data on protein-protein interactions (PPI) and gene expression are increasingly available.

Purpose of the Study:

• To bridge the gap between statistical and molecular models of epistasis.
• To demonstrate incorporating PPI information into the statistical analysis of copy number variation (CNV) interactions.
• To identify biologically plausible epistatic interactions influencing gene expression.

Main Methods:

• Integrated human PPI data with common CNV data from HapMap samples.
• Identified CNV pairs overlapping with genes involved in known PPIs.
• Performed statistical analysis of CNV-CNV interactions on gene expression levels, using PPI data to guide hypothesis testing.

Main Results:

• Identified 37 CNV pairs overlapping with genes in PPI networks.
• Two CNV pairs showed sufficient variation for epistasis analysis.
• Five epistatic effects were detected (P < 10^-6), including a significant CNV-CNV interaction associated with TP53TG3 expression (P = 2 × 10^-20).

Conclusions:

• Incorporating PPI data enhances the biological relevance of statistical epistasis testing.
• This approach successfully identified a CNV-CNV interaction with a significant impact on TP53TG3 gene expression.
• The findings suggest a molecular mechanism involving protein binding and transcriptional regulation, aligning statistical and molecular epistasis models.