pygsva: Gene Set Variation Analysis in Python

Gene Set Variation Analysis (GSVA) is a powerful gene set enrichment method designed for single-sample analysis. It enables pathway-centric analyses of molecular data by shifting the functional unit of analysis from individual genes to gene sets. This approach is particularly useful for bulk microarray, RNA-seq, and other molecular profiling data types, providing a pathway-level view of biological activity.

Overview

GSVA transforms an input gene-by-sample expression data matrix into a gene-set-by-sample expression data matrix, representing pathway activities. This transformed data can then be utilized with classical analytical methods such as:

• Differential Expression
• Classification
• Survival Analysis
• Clustering
• Correlation Analysis

Additionally, GSVA enables pathway comparisons with other molecular data types, such as microRNA expression, binding data, copy-number variation (CNV), or single nucleotide polymorphisms (SNPs).

Performance

Version 0.2.0 includes significant performance optimizations with vectorized NumPy operations:

For very large datasets (19,000 genes × 1,500 samples), GSVA with kcdf='Gaussian' completes in ~4 minutes (previously 52 minutes).

Methods

The pygsva package provides Python implementations of four single-sample gene set enrichment methods:

1. PLAGE (Pathway Level Analysis of Gene Expression)

• Reference: Tomfohr, Lu, and Kepler (2005)
• Standardizes expression profiles over the samples.
• Performs Singular Value Decomposition (SVD) on each gene set.
• The coefficients of the first right-singular vector are returned as pathway activity estimates.

2. Z-Score Method

• Reference: Lee et al. (2008)
• Standardizes expression profiles over the samples.
• Combines standardized values for each gene in a gene set.

3. ssGSEA (Single-Sample Gene Set Enrichment Analysis)

• Reference: Barbie et al. (2009)
• Calculates enrichment scores as the normalized difference in empirical cumulative distribution functions (CDFs) of gene expression ranks inside and outside the gene set.
• By default, the pathway scores are normalized by dividing them by the range of calculated values.

4. GSVA (Default Method)

• Reference: Hänzelmann, Castelo, and Guinney (2013)
• A non-parametric method using empirical CDFs of gene expression ranks inside and outside the gene set.
• Calculates an expression-level statistic to bring gene expression profiles with varying dynamic ranges to a common scale.

Installation

# Install from PyPI
pip install pygsva

# Or install from source
git clone https://github.com/guokai8/pygsva
cd pygsva
pip install .

Usage

from pygsva import *

# Load example data
hsko = load_hsko_data()
pbmc = load_pbmc_data()
gene_sets = {key: group.iloc[:, 0].tolist() for key, group in hsko.groupby(hsko.iloc[:, 2])}

# GSVA (default method)
param = gsvaParam(pbmc, gene_sets=gene_sets, kcdf="Gaussian", n_jobs=4)
result = gsva(param)

# For faster computation on large datasets, use kcdf="none"
param_fast = gsvaParam(pbmc, gene_sets=gene_sets, kcdf="none", n_jobs=4)
result_fast = gsva(param_fast)

# ssGSEA
result_ssgsea = ssgsea(pbmc, gene_sets, n_jobs=4)

# PLAGE
param_plage = plageParam(pbmc, gene_sets=gene_sets, min_size=2)
result_plage = gsva(param_plage)

# Z-score
param_zscore = zscoreParam(pbmc, gene_sets)
result_zscore = gsva(param_zscore)

Key Parameters

• kcdf: Kernel for cumulative distribution function estimation
  • "Gaussian": Standard GSVA with Gaussian kernel (default)
  • "Poisson": For count data (RNA-seq)
  • "none": Empirical CDF only (fastest, recommended for large datasets)
• n_jobs: Number of CPU cores for parallel processing (-1 for all cores)
• min_size/max_size: Filter gene sets by size
• use_sparse: Use sparse matrix operations for memory efficiency

References

If you use any of the methods in this package, please cite the corresponding articles:

1. Tomfohr, Lu, and Kepler (2005) - Pathway Level Analysis of Gene Expression (PLAGE)
2. Lee et al. (2008) - Z-Score Method
3. Barbie et al. (2009) - Single Sample Gene Set Enrichment Analysis (ssGSEA)
4. Hänzelmann, Castelo, and Guinney (2013) - Gene Set Variation Analysis (GSVA)

License

This project is licensed under the MIT License. See the LICENSE file for details.

Contact

For any questions please contact guokai8@gmail.com or https://github.com/guokai8/pygsva/issues