obs_prerank_enrich() Enrichment: Gene List with Rank Scores for Observations
Step 1: Prepare rank_df
Begin by preparing a list of genes along with their preranked scores (e.g., gene expression). Store this information in a DataFrame called rank_df.
- The index of
rank_dfshould be the gene symbols. - The columns of
rank_dfshould correspond to one observation ID (e.g., sample or cell identifier).
This format allows obs_prerank_enrich() to perform enrichment analysis using observation-level ranked gene-level statistics.
%load_ext autoreload
import pandas as pd
rank_df = pd.read_csv("grea/data/GSE13268_expr.csv", index_col=0)
rank_df.head()
Step 2: Preparing Gene Set Libraries
There are several ways to prepare gene set libraries for use in GREA:
Option 1: Use Built-in Libraries
Simply specify the libraries you're interested in as a list. For example:
libraries = ['KEGG_2019_Mouse', 'WikiPathways_2024_Mouse']
You can use the grea.library.list_libraries() function to view all available pathway libraries included in GREA.
Option 2: Load from GMT File
You can load external gene set libraries from .gmt files using:
libraries = read_gmt('your_library_file.gmt')
Option 3: Define a Custom Library
Create your own gene set library using a Python dictionary, where each key is a pathway name and the corresponding value is a list of genes:
libraries = {
'term1': ['A2ml1', 'A1cf'],
'term2': ['A3galt2', 'A4galt']
}
from grea.library import list_libraries
print(list_libraries())
Step 3: Run Enrichment
To perform enrichment analysis, call the grea.obs_prerank_enrich(rank_df, libraries) function.
The observation-level enrichment do not estimate the p-value for efficiency.
The function returns a GREA object containing all enrichment results, including enrichment scores and statistical significance for each library term.
%autoreload
from grea import grea
libraries = ['WikiPathways_2024_Mouse']
obj = grea.obs_prerank_enrich(rank_df, libraries)
obj
Step 4: Check Enrichment Results
The GREA object stores all enrichment results, including enrichment scores and statistical significance for each library term. GREA supports three types of enrichment scores, each reflecting a different scoring strategy:
'KS-ES': Kolmogorov–Smirnov-based Enrichment Score, capturing the peak deviation between hit and miss distributions.'KS-ESD': KS-based enrichment Score Difference, the sum of the maximum positive and negative deviations from the running score.'RC-AUC': Area Under the Recovery Curve, summarizing early enrichment of target genes along the ranking.
You can select the appropriate metric depending on your analysis goal or data characteristics.
To retrieve the enrichment results as a long DataFrame, use the get_enrich_results(metric) function, as a wide DataFrame, use the get_enrich_score(metric).
%autoreload
df = obj.get_enrich_results(metric='KS-ES')
df
%autoreload
df = obj.get_enrich_score(metric='KS-ES')
df
%autoreload
df = obj.get_enrich_results(metric='KS-ESD')
df
%autoreload
df = obj.get_enrich_score(metric='KS-ESD')
df
%autoreload
df = obj.get_enrich_results(metric='RC-AUC')
df
%autoreload
df = obj.get_enrich_score(metric='RC-AUC')
df
Step 5: Visualize Enrichment Results
To visualize the enrichment results, use the pl_running_sum(metric, term, obs_id) function by specifying the desired metric, term, and observation ID.
%autoreload
term = 'WikiPathways_2024_Mouse|TCA Cycle WP434'
obs_id = 'GSM334850'
fig = obj.pl_running_sum('KS-ES', term, obs_id)
fig
%autoreload
term = 'WikiPathways_2024_Mouse|TCA Cycle WP434'
obs_id = 'GSM334850'
fig = obj.pl_running_sum('KS-ESD', term, obs_id)
fig
%autoreload
term = 'WikiPathways_2024_Mouse|TCA Cycle WP434'
obs_id = 'GSM334850'
fig = obj.pl_running_sum('RC-AUC', term, obs_id)
fig
