Prior knowledge driven Granger causality analysis on gene regulatory network discovery | JoVE Visualize

Area of Science:

Systems Biology
Computational Biology
Genomics

Background:

Discovering gene regulatory networks from time-series gene expression data is crucial.
Traditional Granger Causality (GC) models struggle with datasets where the number of genes (n) far exceeds the number of time points (T), leading to false positives.
Existing methods like pairwise GC (PGC) and regularization strategies are insufficient when n>>T.

Purpose of the Study:

To address the limitations of existing methods for inferring gene regulatory networks from time-series expression data, especially when n>>T.
To propose a novel Granger Causality-based method that incorporates prior biological knowledge to improve accuracy.
To enhance the discovery of true positive regulatory interactions.

Main Methods:

Developed CGC-2SPR (Causality using two-step prior Ridge regularization), a new method integrating prior biological knowledge with gene expression data.
Incorporated prior biological knowledge to guide the Granger Causality model.
Utilized Monte Carlo Significance Estimation (MCSE) to statistically validate discovered network edges.

Main Results:

CGC-2SPR demonstrated significant improvements in accuracy compared to widely used GC (PGC, Ridge, Lasso) and MI-based (MRNET, ARACNE) methods in simulation experiments.
Application to the yeast metabolic cycle dataset revealed more true positive regulatory edges using CGC-2SPR than existing methods.
The combination of prior knowledge and gene expression data showed a synergistic effect ('1+1>2') in network discovery.

Conclusions:

Integrating heterogeneous biological knowledge with gene expression data enhances causality modeling for regulatory network discovery.
The proposed CGC-2SPR method offers a more accurate and robust approach for inferring gene regulatory networks, particularly in n>>T scenarios.
Functional prediction based on discovered networks identified a potential role for the Abm1 gene in yeast's response to glucose levels, highlighting the method's utility in biological discovery.