Searching for functional gene modules with interaction component models.

Area of Science:

• Systems Biology
• Bioinformatics
• Computational Biology

Background:

• Identifying functional gene modules and protein complexes is crucial for understanding cellular mechanisms.
• Existing clustering methods often fail to account for uncertainty in gene expression and protein interaction data.
• Proteins can participate in multiple functional roles, necessitating models that capture overlapping clusters.

Purpose of the Study:

• To develop a generative probabilistic model for discovering functional gene modules.
• To incorporate both gene expression and protein-protein interaction data, addressing data uncertainty.
• To enable the modeling of overlapping functional modules where proteins can have multiple roles.

Main Methods:

• Formulated a generative probabilistic model for protein-protein interaction networks.
• Integrated gene expression data into the probabilistic model using two distinct approaches.
• Applied the model to analyze two datasets from yeast (Saccharomyces cerevisiae).

Main Results:

• The model successfully identifies interaction components interpretable as overlapping functional modules.
• Demonstrated superior performance in discovering biologically relevant modules with enriched functional classes compared to existing methods.
• Validated findings on two yeast datasets, highlighting the model's effectiveness.

Conclusions:

• Combining gene expression and protein interaction data via a probabilistic generative model significantly improves module discovery.
• The proposed model identifies biologically relevant modules more effectively than alternative methods.
• The model inherently handles overlapping modules, reflecting the multifaceted roles of proteins in cellular functions.