Network methods for describing sample relationships in genomic datasets: application to Huntington's disease

Area of Science:

• Genomics
• Bioinformatics
• Systems Biology

Background:

• Genomic datasets are common, but sample relationships are often overlooked.
• Under-appreciation of sample variation complicates genomic data analysis.
• Gene co-expression network analysis is particularly affected by uncharacterized sample variation.

Purpose of the Study:

• Characterize sample relationships in genomic data using network methods.
• Develop a robust approach to identify outlying samples without relying on clustering.
• Introduce metrics for quantifying sample relationship consistency across studies and platforms.

Main Methods:

• Applied network analysis to human brain tissue microarray data.
• Developed a novel method for outlier sample detection.
• Introduced correlation between connectivity and clustering coefficient (cor(K,C)) as a homogeneity measure.
• Utilized cor(K,C) to identify disease-specific effects and gene modules.

Main Results:

• Demonstrated network methods for characterizing sample relationships in genomic data.
• Identified cor(K,C) as a sensitive indicator of biological sample homogeneity.
• Showcased cor(K,C)'s ability to distinguish biologically meaningful sample subgroups.
• Revealed Huntington's disease impact on caudate nucleus samples and specific co-expressed gene modules.

Conclusions:

• Systematic exploration of sample relationships is crucial for large genomic datasets.
• Introduced a standardized R software platform for interactive sample network exploration.
• The findings highlight a new strategy for investigating disease effects on gene sets.