A generative model of identifying informative proteins from dynamic PPI networks

Area of Science:

• Bioinformatics
• Systems Biology
• Computational Biology

Background:

• Informative proteins are crucial for cellular functions and translating bioinformatics into clinical practice.
• Existing methods for identifying informative biomarkers often lack dynamic biological process considerations, being heuristic and arbitrary.
• Understanding dynamic protein-protein interaction networks (PPINs) is key to discovering functional proteins.

Purpose of the Study:

• To develop a novel generative model for identifying informative proteins.
• To systematically analyze the topological variety of dynamic protein-protein interaction networks (PPINs).
• To improve the identification of critical proteins for potential biomarker research.

Main Methods:

• A deep feature generation model is employed to learn common representations of multiple PPINs.
• The model reconstructs original PPINs, and reconstruction errors are analyzed to locate informative proteins.
• The approach systematically analyzes the topological variety within dynamic PPINs.

Main Results:

• The generative model effectively identifies informative proteins by analyzing reconstruction errors from dynamic PPINs.
• Experiments on yeast cell cycles and prostate cancer data demonstrate the method's effectiveness.
• The identified informative proteins show potential for biomarker research in dynamic biological processes.

Conclusions:

• The proposed generative model offers a systematic approach to identifying informative proteins from dynamic PPINs.
• This method overcomes limitations of heuristic biomarker identification by considering network dynamics.
• The findings highlight critical protein members in dynamic biological processes, paving the way for biomarker discovery.