Bayesian graphical models for computational network biology.

Area of Science:

• Computational network biology
• Bioinformatics
• Biostatistics

Background:

• Probabilistic graphical models offer a robust framework for characterizing molecular interactions and their uncertainties.
• Reciprocal graphs (RG) are a powerful, yet underutilized, class of graphical models suitable for feedback mechanisms in molecular networks.

Purpose of the Study:

• To review probabilistic graphical models and recent advancements in RG approaches for network modeling.
• To extend RG methods for integrated analysis of multi-omics data (DNA, RNA, protein).

Main Methods:

• Review of directed, undirected, and reciprocal graphical models.
• Extension of RG models to incorporate multi-platform data (DNA copy number, methylation, gene and protein expression).
• Application of the extended RG method to The Cancer Genome Atlas ovarian cancer dataset.

Main Results:

• Demonstration of RG models' utility in analyzing complex molecular networks.
• Identification of significant findings within the ovarian cancer multi-omics data through the extended RG approach.

Conclusions:

• The extended RG approach provides a principled and efficient method for integrating diverse molecular data types.
• This work highlights the potential of RG models in biostatistics and bioinformatics for uncovering biological insights.