This Python package provides a computationally performant Gene Set Enrichment Analysis (GSEA) implementation of the pre-rank algorithm [1]. GSEApy was used as the reference for the running sum and enrichment score calculation [2]. The algorithm estimates the enrichment score (ES) distribution of the null model by fitting data to gamma distibutions instead of calculating permutations for each gene set. blitzGSEA calculates p-values with much higher accuracy than other reference implementations available in Python.

Gene set libraries can directly be loaded from Enrichr (https://maayanlab.cloud/Enrichr). For this use the blitzgsea.enrichr.get_library() function. All libraries can also be listed with blitzgsea.enrichr.print_libraries().

blitzGSEA provides plotting functions to generate publication ready figures similar to the original GSEA-P software. blitzgsea.plot.running_sum() plots an enrichment plot for a single gene set and blitzgsea.plot.top_table() plots the top n gene sets in a compact table.

blitzGSEA is currently only available as a Python package in this GitHub repository. You can install the blitzGSEA Python package and its dependencies through pip by using the following command:

$ pip install git+https://github.com/MaayanLab/blitzgsea.git

blitzGSEA depends on two input files. 1) a gene signature and 2) a gene set library. The gene set library is a dictionary with the name of the gene set as key and a list of gene ids as values. Gene set libraries can be loaded directly from Enrichr. The signature should be a pandas dataframe with two columns [0,1]. The first column should contain the gene ids (matching the gene ids in the gene set library).

The gene set library is a dictionary with the gene set names as key and lists of gene ids as values.

{
'ERBB2 SIGNALING PATHWAY (GO:0038128)': ['CDC37',
                                          'PTPN12',
                                          'PIK3CA',
                                          'SOS1',
                                          'CPNE3',
                                          'EGF',
                                          ...
                                         ],
'HETEROTYPIC CELL-CELL ADHESION (GO:0034113)': ['DSC2',
                                                 'ITGAV',
                                                 'ITGAD',
                                                 'LILRB2',
                                                 ...
                                                ],
...
}

This short example will download two files (signature and gene set library). The gene set library consists of KEGG pathways and the signature is an example signature of differential gene expression of muscle samples from young and old donors. Differential gene expression was computed with Limma Voom.

import blitzgsea as blitz
import urllib.request
import pandas as pd

# download example gene expression signature
url = "https://github.com/MaayanLab/blitzgsea/raw/main/testing/ageing_muscle_gtex.tsv"
urllib.request.urlretrieve(url, "ageing_muscle_gtex.tsv")

# read signature as pandas dataframe
signature = pd.read_csv("ageing_muscle_gtex.tsv")

# list available gene set libraries in Enrichr
blitz.enrichr.print_libraries()

# use enrichr submodule to retrieve gene set library
library = blitz.enrichr.get_library("KEGG_2021_Human")

# run enrichment analysis
result = blitz.gsea(signature, library)

The main function of blitzgsea.gsea() supports several optional parameters. The default parameters should work well for most use cases.

blitzGSEA supports several plotting functions. blitzgsea.plot.running_sum() and blitzgsea.plot.top_table() can be used after enrichment has been performed. blitzgsea.plot.running_sum() shows the running sum of an individual gene set. It has a compact mode in which the image will be more readable if small. blitzgsea.plot.top_table() shows the top n enriched gene sets and displays the results in a table, with normalized enrichment score (NES) and the distribution of hits relative to the gene ranking of the signature.

import blitzgsea as blitz
import urllib.request
import pandas as pd

# download example gene expression signature
url = "https://github.com/MaayanLab/blitzgsea/raw/main/testing/ageing_muscle_gtex.tsv"
urllib.request.urlretrieve(url, "ageing_muscle_gtex.tsv")

# read signature as pandas dataframe
signature = pd.read_csv("ageing_muscle_gtex.tsv")

# list available gene set libraries in Enrichr
blitz.enrichr.print_libraries()

# use enrichr submodule to retrieve gene set library
library = blitz.enrichr.get_library("KEGG_2021_Human")

# run enrichment analysis
result = blitz.gsea(signature, library)

# plot the enrichment results and save to pdf
fig = blitz.plot.running_sum(signature, "CELL ADHESION MOLECULES", library, result=result, compact=False)
fig.savefig("running_sum.png", bbox_inches='tight')

fig_compact = blitz.plot.running_sum(signature, "CELL ADHESION MOLECULES", library, result=result, compact=True)
fig_compact.savefig("running_sum_compact.png", bbox_inches='tight')

fig_table = blitz.plot.top_table(signature, library, result, n=15)
fig_table.savefig("top_table.png", bbox_inches='tight')

The resulting plots will look like the examples below:

This is the sample shuffling algorithm from GSEApy. It performs a t-test to build signatures for phenotype shuffled groups. The input is a gene expression dataframe, which should be normalized for library size. groups is a list containing 0 or 1 describing the corresponding group for the samples in exprs. The index of exprs are the gene ids matching the gene set library.

blitz.shuffle.gsea(exprs, library, groups, permutations=50, seed=1)

Python 3.6+

When using blitzgsea please cite the following reference:

Lachmann, Alexander, Zhuorui Xie, and Avi Ma’ayan. "blitzGSEA: efficient computation of gene set enrichment analysis through gamma distribution approximation." Bioinformatics (2022). https://academic.oup.com/bioinformatics/advance-article/doi/10.1093/bioinformatics/btac076/6526383?login=false

[1] Lachmann, Alexander, Zhuorui Xie, and Avi Ma’ayan. "blitzGSEA: efficient computation of gene set enrichment analysis through gamma distribution approximation." Bioinformatics (2022).

[2] Subramanian, Aravind, Heidi Kuehn, Joshua Gould, Pablo Tamayo, and Jill P. Mesirov. "GSEA-P: a desktop application for Gene Set Enrichment Analysis." Bioinformatics 23, no. 23 (2007): 3251-3253.

[3] Fang, Z., A. Wolf, Y. Liao, A. McKay, F. Fröhlich, J. Kimmel, L. Xiaohui, and sorrge. "zqfang/gseapy: gseapy-v0. 10.3." (2021).