Reconstruction of molecular network evolution from cross-sectional omics data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Understanding cancer evolution requires analyzing patient traits across disease stages.
• Reconstructing gene-gene interaction networks from omics data is complex, especially with non-homogeneous patient groups.
• Gene-gene interaction networks may change dynamically during cancer progression.

Purpose of the Study:

• To operationalize the estimation of stage-wise mixtures of Gaussian graphical models (GGMs) from high-dimensional omics data.
• To develop a robust methodology for identifying dynamic changes in gene-gene interaction networks across different cancer stages.
• To apply the developed method to identify gene-gene interaction network alterations in prostate cancer progression.

Main Methods:

• Utilized a fused ridge penalized Expectation-Maximization (EM) algorithm for fitting mixtures of GGMs.
• Employed cross-validation for selecting optimal fused ridge penalty parameters.
• Proposed estimation procedures were evaluated for consistency and performance through simulations.

Main Results:

• The proposed estimation procedures demonstrated consistency in estimating stage-wise GGMs.
• Simulation studies confirmed the performance of the methodology in various aspects.
• The method successfully identified gene-gene interaction network changes in the transition from normal to prostate cancer tissue.

Conclusions:

• The developed methodology provides a robust framework for analyzing dynamic gene-gene interaction networks across disease stages.
• This approach enhances the understanding of cancer evolution by revealing network alterations.
• The application to prostate cancer data illustrates the practical utility in identifying clinically relevant network changes.