Learning causal networks from systems biology time course data: an effective model selection procedure for the vector...

Area of Science:

• Genomics
• Systems Biology
• Statistical Modeling

Background:

• Causal networks based on the vector autoregressive (VAR) process are valuable for modeling cellular regulatory interactions.
• Learning these networks is difficult with genomic data due to low sample size and high dimensionality.

Purpose of the Study:

• To present a novel and efficient approach for estimating VAR networks.
• To address the challenges of learning causal networks in high-dimensional genomic data.

Main Methods:

• A two-step procedure involving improved estimation of VAR regression coefficients using analytic shrinkage.
• Subsequent model selection through testing partial correlations.

Main Results:

• The proposed approach demonstrated superior performance in simulations for small sample sizes, particularly in true discovery rate.
• Analysis of Arabidopsis thaliana expression time series data yielded a biologically meaningful network.

Conclusions:

• The statistical learning of large-scale VAR causal models can be efficiently achieved using the proposed method.
• This procedure is effective even in challenging data scenarios common in genomics and proteomics.