Interpreting transcriptional changes using causal graphs: new methods and their practical utility on public networks

Area of Science:

• Transcriptional data analysis
• Systems biology
• Bioinformatics

Background:

• Inferring regulatory cascades from transcriptional data is crucial for understanding molecular and environmental perturbations.
• Commercial tools exist but lack freely available, efficient algorithms for academic use with public networks.

Purpose of the Study:

• To propose a novel statistical method for inferring upstream regulators from gene expression patterns.
• To provide an efficient algorithm suitable for public interaction networks with mixed edge types.

Main Methods:

• Developed a new statistical method to infer upstream regulators based on differential gene expression.
• Created a novel algorithmic approach for efficient statistical inference.
• Implemented the method in the R package QuaternaryProd.

Main Results:

• The method effectively infers upstream regulators using public interaction networks.
• Demonstrated applicability in in-vitro experiments, stem-cell differentiation, and animal disease models.
• The Quaternary test statistic incorporates diverse evidence for regulator relevance.

Conclusions:

• Closed a gap in using causal networks for analyzing differentially expressed genes.
• The approach supports both signed and unsigned causal networks, addressing ambiguities.
• The implemented method efficiently estimates p-values for upstream regulators in current biological contexts.