ssNPA

ssNPA quantifies sample-specific pathway perturbations by learning causal gene networks from control samples and measuring network-neighborhood deregulation in individual gene expression profiles.

Key Features:

• Instance-Specific Analysis: Analyzes single-cell RNA-seq and disease samples to identify network-level perturbations specific to each individual sample.
• Causal Graph Learning: Learns a causal graph directly from control samples, avoiding reliance on external pathway databases.
• Network Neighborhood Deregulation Quantification: Quantifies deregulation by measuring the prediction error of a gene's expression based on its Markov blanket (the set of genes directly influencing the target gene).
• Performance Evaluation: Evaluated on liver development single-cell RNA-seq data and two TCGA datasets, demonstrating improved recovery of cell timing and identification of patient clusters with significant survival differences relative to alternative methods.

Scientific Applications:

• Pathway Subtyping: Enables subtyping of samples based on unique pathway perturbations derived from gene network deregulation.
• Personalized Medicine: Characterizes pathway perturbations in individual samples to inform patient-specific diagnosis and treatment decisions.
• Research Applications: Applicable to analysis of single-cell RNA-seq and cancer genomics for studying dynamic changes in gene expression networks and identifying therapeutic targets.

Methodology:

Constructs a causal graph from control samples and assesses deregulation by evaluating prediction errors within each gene's Markov blanket.