dce

dce infers differential causal effects to identify dysregulated edges within signaling pathways by comparing normal and cancerous cells for applications in cancer research.

Key Features:

• Causality-Based Framework: Employs causal inference to pinpoint individual pathway edges that differ between normal and cancerous cellular states.
• Confounding Adjustment: Accounts for confounding, including unobserved dense confounding where latent variables such as batch effects or cell cycle states influence multiple covariates simultaneously.
• Robustness to Technical Artifacts: Minimizes the impact of technical artifacts to enhance reliability of detected pathway dysregulations.
• Performance Validation: Demonstrated superior performance on synthetic datasets and CRISPR knockout screens and validated latent confounding adjustment using GTEx data.
• Discovery Potential: Applied to TCGA breast cancer data to recover known genes and nominate novel genes implicated in breast cancer progression.

Scientific Applications:

• Cancer Research: Identifies dysregulated pathways and pathway edges to elucidate molecular mechanisms of cancer development and progression.
• Genomic Studies: Adjusts for latent confounding to support analyses of complex gene–environment and gene–gene interactions in genomic datasets.
• Drug Target Identification: Provides insights into pathway components and edges whose dysregulation may inform therapeutic target selection.

Methodology:

Extends traditional causal inference techniques to accommodate unobserved dense confounding, compares normal and cancerous cells to estimate differential causal effects on pathway edges, and employs statistical modeling and computational algorithms to isolate biological signals from noise.