Odds ratios differ between enrichr and gseapy.enrich()

[136s]

Hi,

As mentioned in the isssue #132 , the formula for calculating the odds ratio differs from the Enrichr definition.

Setup

I am reporting a problem with GSEApy version, Python version, and operating
system as follows:

import sys; print(sys.version)
import platform; print(platform.python_implementation()); print(platform.platform())
import gseapy; print(gseapy.__version__)

3.11.0 | packaged by conda-forge | (main, Oct 25 2022, 06:12:32) [MSC v.1929 64 bit (AMD64)]
CPython
Windows-10-10.0.19045-SP0
1.0.6

Steps to reproduce

from pathlib import Path
import subprocess

import gseapy as gp
import pandas as pd

def gene_list():
    gene_list_link = "https://raw.githubusercontent.com/zqfang/GSEApy/master/tests/data/gene_list.txt"
    return (
        pd.read_csv(gene_list_link, header=None, sep="\t")
        .squeeze()
        .str.strip()
        .to_list()
    )


def msigdb_hallmark_2020():
    kegg_gmt_link = "https://maayanlab.cloud/Enrichr/geneSetLibrary?mode=text&libraryName=MSigDB_Hallmark_2020"
    gmt_path = Path("MSigDB_Hallmark_2020.gmt")
    subprocess.run(["curl", "-o", gmt_path.as_posix(), kegg_gmt_link])
    gene_sets = gp.read_gmt(gmt_path.as_posix())
    gmt_path.unlink()
    return gene_sets


background = 20000
col_filter = [
    "Term",
    "Overlap",
    "P-value",
    "Adjusted P-value",
    "Odds Ratio",
    "Combined Score",
]

gseapy.enrichr() behaviour

enr1 = gp.enrichr(
    gene_list=gene_list(), gene_sets="MSigDB_Hallmark_2020", background=background
)
enr1.results[col_filter].head()

	Term	Overlap	P-value	Adjusted P-value	Odds Ratio	Combined Score
0	IL-6/JAK/STAT3 Signaling	19/87	1.1972245657484395e-09	5.986122828742197e-08	6.844693831437825	140.61232377322565
1	TNF-alpha Signaling via NF-kB	27/200	3.220897630403587e-08	5.368162717339312e-07	3.8415676480045455	66.270962847295
2	Complement	27/200	3.220897630403587e-08	5.368162717339312e-07	3.8415676480045455	66.270962847295
3	Inflammatory Response	24/200	1.6358898982890469e-06	2.0448623728613085e-05	3.343017806935333	44.54010811384466
4	heme Metabolism	23/200	5.533816327081873e-06	5.533816327081873e-05	3.181358113561503	38.50917198328045

gseapy.enrich() behaviour

enr2 = gp.enrich(
    gene_list=gene_list(), gene_sets=msigdb_hallmark_2020(), background=background
)
enr2.results[col_filter].sort_values("P-value").head()

	Term	Overlap	P-value	Adjusted P-value	Odds Ratio	Combined Score
22	IL-6/JAK/STAT3 Signaling	19/87	1.1974768437897092e-09	5.987384218948546e-08	5.678650945983	116.65680534113953
41	TNF-alpha Signaling via NF-kB	27/200	3.2213958391716315e-08	5.368993065286052e-07	3.511026893013104	60.56825441128164
10	Complement	27/200	3.2213958391716315e-08	5.368993065286052e-07	3.511026893013104	60.56825441128164
23	Inflammatory Response	24/200	1.6360465440779507e-06	2.0450581800974384e-05	3.1159207583102773	41.51412228865055
47	heme Metabolism	23/200	5.5342613169804326e-06	5.5342613169804326e-05	2.984896280199822	36.13083360370597

Enrichr behaviour

Enrichr result

The odds ratios for Enrichr and gseapy.enrichr() look the same, but the odds ratios for gseapy.enrich() do not.

Thank you.

[zqfang]

Let see how issue #132 goes first

[zqfang]

Thank you so much for your contribution. the code has been merge into master. A new quick patch version (1.1.1) has just released.

[136s]

Thanks for your quick merge and release!