NetDiff - Bayesian model selection for differential gene regulatory network inference.

Area of Science:

• Computational Biology
• Systems Biology
• Genomics

Background:

• Understanding cellular process changes is crucial for disease research.
• Gene regulatory networks (GRNs) are key to cellular function.
• Identifying differential networks aids in distinguishing biological states.

Purpose of the Study:

• To develop a novel computational methodology for inferring differential gene regulatory networks.
• To improve the efficiency and robustness of network inference from gene expression data.
• To identify differential network interactions in complex diseases.

Main Methods:

• Utilized a Bayesian model selection approach for comparing network structures.
• Employed Gaussian graphical models (GGMs) to represent gene regulatory networks.
• Applied variational inference for computationally efficient learning of GGMs.

Main Results:

• The proposed method demonstrated higher robustness compared to independent analyses.
• Achieved fewer false positive predictions of differential edges in synthetic data.
• Successfully identified differential network interactions in amyotrophic lateral sclerosis (ALS) patient data.

Conclusions:

• The novel methodology provides a computationally efficient and robust tool for differential GRN inference.
• This approach can significantly advance the understanding of molecular mechanisms in various biological conditions.
• Facilitates the discovery of novel biomarkers and therapeutic targets in diseases like ALS.