Discovery of meaningful associations in genomic data using partial correlation coefficients

Area of Science:

• Systems Biology
• Bioinformatics
• Network Inference

Background:

• Inferring biochemical interactions from large-scale omics data (transcriptomics, proteomics, metabolomics) is a key challenge in systems biology.
• Existing methods often struggle to differentiate direct from indirect interactions.
• Understanding biochemical networks is crucial for deciphering cellular functions.

Purpose of the Study:

• To develop and evaluate a novel method for constructing Undirected Dependency Graphs from large-scale biochemical data.
• To infer the underlying network topology by distinguishing direct and indirect interactions.
• To apply the method to real biological data and generate functional hypotheses.

Main Methods:

• Utilizing partial correlation analysis to model relationships between biochemical compounds.
• Constructing approximate Undirected Dependency Graphs to represent network structures.
• Evaluating the method's performance using simulated data with varying noise levels.

Main Results:

• The partial correlation analysis demonstrated good statistical power and a low False Discovery Rate, even with noisy data.
• Application to yeast gene expression data successfully inferred small gene networks.
• Inferred networks contained known functionally related genes and uncharacterized open reading frames (ORFs), suggesting potential new gene functions.

Conclusions:

• Partial correlation analysis is a robust method for inferring biochemical interaction networks from large-scale omics data.
• The approach effectively identifies network topology and aids in discovering gene functions, including those of uncharacterized ORFs.
• The developed software is available for Windows and Linux systems, facilitating broader application in systems biology research.