Causal Network Models for Predicting Compound Targets and Driving Pathways in Cancer

Area of Science:

• Computational Biology
• Systems Biology
• Genomics

Background:

• Gene-expression data analysis is crucial for understanding transcriptional responses to compounds.
• Current methods often use differentially expressed genes (DEGs) for pathway enrichment, overlooking upstream causal factors.
• Identifying causal targets and pathways is essential for drug discovery and cancer research.

Purpose of the Study:

• To develop and validate graph-based models for predicting causal targets from compound-microarray data.
• To compare different network traversal approaches for identifying compound targets.
• To enhance pathway enrichment analysis by utilizing causal nodes instead of DEGs.

Main Methods:

• Development of graph-based models to predict causal targets from compound-microarray data.
• Evaluation of various network topology traversal methods, including a consensus minimum-rank score (SigNet).
• Application of the approach to integrated datasets from The Cancer Genome Atlas (TCGA) for triple-negative breast cancer analysis.

Main Results:

• The consensus minimum-rank score (SigNet) outperformed individual methods in ranking compound targets.
• Larger, non-canonical networks proved more effective than linear canonical interactions.
• Pathway enrichment using causal nodes identified relevant pathways more frequently than using DEGs.
• Critical and novel pathways, including growth, DNA damage, cytoskeleton remodeling, immune response, and apoptosis pathways, were uncovered in triple-negative breast cancer.

Conclusions:

• The developed graph-based approach effectively links transcriptional profiles to compound targets and driving pathways in cancer.
• This method offers a more accurate way to identify causal relationships and relevant pathways compared to traditional DEG-based enrichment.
• The findings provide valuable insights for drug discovery and understanding cancer biology.