Robust gene coexpression networks using signed distance correlation

Area of Science:

• Genomics
• Bioinformatics
• Systems Biology

Background:

• Many genes lack functional annotations, hindering biological relationship inference.
• Gene coexpression networks are valuable for discovering gene functions but require reliable validation.
• Existing network construction methods face challenges due to annotation scarcity.

Purpose of the Study:

• To develop a principled method for constructing structurally stable gene coexpression networks.
• To infer biological relationships between genes using only gene expression data.
• To overcome limitations of current methods in the absence of functional information.

Main Methods:

• Introduced signed distance correlation as a novel measure of variable dependency.
• Applied signed distance correlation to generate gene coexpression networks.
• Developed a framework for self-consistent network generation purely from gene expression data.

Main Results:

• Gene coexpression networks generated using signed distance correlation are more stable.
• The proposed method captures more biological information compared to Pearson correlation or mutual information.
• Networks are robust and informative even without prior functional annotations.

Conclusions:

• Signed distance correlation provides a powerful and intuitive approach for gene coexpression network construction.
• This method enhances the reliability of gene function inference by generating stable and informative networks.
• The framework offers a valuable tool for systems biology research, particularly when functional annotations are limited.