GSVA

GSVA estimates variation in pathway activity across samples by transforming a gene-by-sample expression matrix into a gene-set-by-sample matrix for sample-wise gene set enrichment analysis.

Key Features:

• Non-parametric and unsupervised: Uses a non-parametric, unsupervised approach that does not require parametric assumptions or predefined sample labels.
• Matrix transformation: Converts a gene-by-sample expression matrix into a gene-set-by-sample matrix to enable pathway-level scoring per sample.
• Per-sample enrichment scoring: Produces sample-wise pathway or signature scores for evaluation of pathway activity at the individual-sample level.
• Compatibility with microarray and RNA-seq: Applicable to both microarray and RNA-seq expression datasets.
• Noise and dimensionality reduction: Summarizes gene expression into pathway or signature summaries to reduce noise and dimensionality relative to single-gene analyses.
• Increased sensitivity: Improves power to detect subtle changes in pathway activity across samples compared to single-gene methods.
• Robustness to heterogeneity: Suitable for highly heterogeneous datasets and modeling pathway activity across diverse sample populations.

Scientific Applications:

• Differential pathway activity analysis: Compare pathway activity between conditions or groups using per-sample pathway scores.
• Survival analysis: Use sample-wise pathway scores as predictors in survival models to associate pathway activity with clinical outcomes.
• Pathway-centric modeling: Provide pathway-level summaries as inputs for building pathway-centric models of biology.
• Sample-level biological interpretation: Enable nuanced biological interpretation at the individual-sample resolution.

Methodology:

Applies a non-parametric, unsupervised algorithm that transforms a gene-by-sample expression matrix into a gene-set-by-sample matrix and summarizes gene expression into pathway or signature scores.