ssPA

ssPA performs single-sample pathway analysis by transforming molecular-level metabolomics data into pathway-level scores to identify patient-specific pathway signatures.

Key Features:

• Single-sample pathway analysis: Enables pathway scoring at the individual-sample level and supports multi-group comparisons and downstream pathway-based machine learning analyses.
• Implemented methods: Implements ssGSEA, GSVA, Singular Value Decomposition (SVD) as used in PLAGE, z-score calculations, ssClustPA, and kernel Principal Component Analysis (kPCA).
• Benchmarking and evaluation: Benchmarked on semi-synthetic metabolomics data, reporting that GSEA-based and z-score approaches have superior recall while clustering and dimensionality-reduction methods (ssClustPA, kPCA) achieve higher precision at moderate-to-high effect sizes.
• Case study application: Applied to inflammatory bowel disease mass spectrometry data to uncover subtype-specific pathway signatures by clustering pathway scores and generating pathway-based patient subtype-specific correlation networks.
• Pathway-level transformation: Converts molecular omics measurements into pathway-level insights to facilitate interpretation of complex metabolomics datasets relative to conventional approaches such as ORA and GSEA.

Scientific Applications:

• Metabolomics research: Characterizes pathway-level alterations in metabolomic datasets to inform studies of disease mechanisms.
• Patient-specific analysis: Identifies individual-specific pathway signatures to support personalized or stratified analyses.
• Subtype discovery and ML: Provides pathway scores usable for clustering, subtype discovery, and pathway-based machine learning models.

Methodology:

Implements ssGSEA, GSVA, SVD (PLAGE), z-score, ssClustPA, and kPCA; benchmarking was performed on semi-synthetic metabolomics data; includes clustering of pathway scores and visualization of pathway-based patient subtype-specific correlation networks.