Differential network analysis reveals the genome-wide landscape of estrogen receptor modulation in hormonal cancers

Area of Science:

• Systems Biology
• Bioinformatics
• Genomics

Background:

• Mutual information (MI)-based algorithms identify gene interactions but are computationally intensive and limited for genome-wide analysis.
• Existing methods often require prior knowledge, restricting their application in large-scale studies.

Purpose of the Study:

• To introduce the Modulated Gene/Gene set Interaction (MAGIC) analysis for systematic, genome-wide identification of interaction networks.
• To develop a computationally efficient and adaptable method for analyzing modulated interactions in clinical cohorts.

Main Methods:

• MAGIC analysis utilizes a novel statistical test with conjugate Fisher transformations of correlation coefficients.
• The method is designed for fast computation and adaptability to diverse clinical datasets.
• Simulated datasets were used to compare MAGIC's performance against MI-based methods.

Main Results:

• MAGIC demonstrated significantly improved computational efficiency and superior performance compared to MI-based methods in simulations.
• Application to breast cancer revealed novel gene and gene set interaction networks modulated by estrogen receptor (ER).
• An ER+-dependent interaction between TGFβ and NFκB was identified and associated with patient survival, validated in independent datasets.

Conclusions:

• MAGIC provides a systematic framework for constructing modulated interaction networks across the whole genome.
• The method effectively identifies key modulators and functional interactions, applicable to various cancers like breast and ovarian cancer.
• MAGIC offers a powerful tool for systems biology research, with a MATLAB implementation available.