Using the CoRegTor package

In this tutorial we demonstrate the use of the python package CoRegTor to predict coregulatory cluster for a target gene using gene expression data. We also describe how different parameter can be adjusted.

Installing the tool

# Install coregtor if not already installed, then import it
try:
    from coregtor import RunCoRegTor 
except ImportError:
    %pip install coregtor
    from coregtor import RunCoRegTor 

import io
import pandas as pd
import requests
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import seaborn as sns

Preparing the data

In this tutorial we directly fetch gene expression data for from the GTEx portal and the list of human transcription factors (from here). Please note that CoRegTor requires expression data with rows as samples and columns are gene names.

For simplicity, we use the “Description” col as gene name rather than their EnsemblId col name.

# 1. Fetch GTEx gene expression data
gtex_url = "https://storage.googleapis.com/adult-gtex/bulk-gex/v11/rna-seq/tpms-by-tissue/gene_tpm_adult_gtex_v11_liver.gct.gz"
response = requests.get(gtex_url, verify=False)
response.raise_for_status()

# 2. Load into DataFrame skipping GCT metadata rows
df_raw = pd.read_csv( io.BytesIO(response.content), sep='\t', skiprows=2, compression='gzip')

# 3. Drop Ensembl ID ('Name'), aggregate expression by Gene Symbol ('Description'), and transpose
df_matrix = (df_raw.drop(columns=['Name']).groupby('Description').mean().T)
df_matrix.columns.name = "Sample"

# 4. Fetch strict human TF list 
tf_url = "https://raw.githubusercontent.com/aertslab/pySCENIC/master/resources/hs_hgnc_curated_tfs.txt"
tf_response = requests.get(tf_url, verify=False)
tf_response.raise_for_status()

# 5. Read strict TF list into DataFrame and convert to Python list
df_tfs = pd.read_csv(io.StringIO(tf_response.text),header=None,names=['TF_Symbol'])

# Extract cleaned list of sequence-specific TFs
tf_list = df_tfs['TF_Symbol'].dropna().str.strip().tolist()

df_matrix.head(2)

Sample	5S_rRNA	5_8S_rRNA	7SK	A1BG	A1BG-AS1	A1CF	A2M	A2M-AS1	A2ML1	A2ML1-AS1	...	ZXDC	ZYG11A	ZYG11AP1	ZYG11B	ZYX	ZYXP1	ZZEF1	ZZZ3	hsa-mir-8069-1	snoZ196
GTEX-111FC-1226-SM-HAV27	0.0	0.0	0.0	193.103	3.90616	20.3111	40.9842	3.22847	0.276122	0.0	...	8.29606	0.537917	0.0	3.90875	26.7844	0.0	4.09897	1.95510	0.0	0.0
GTEX-11DXY-0526-SM-5EGGQ	0.0	0.0	0.0	442.279	5.75803	39.8927	208.5500	4.47446	0.127329	0.0	...	13.84280	2.242130	0.0	10.09060	29.6591	0.0	5.29482	3.86624	0.0	0.0

2 rows × 73321 columns

print(f"{len(tf_list)} TFs")
found = len(set(tf_list).intersection(set(df_matrix.columns)))
print(f"{found} TFs found in data")

1390 TFs
1383 TFs found in data

import os
base =  os.path.expandvars("$HOME/projects/coregtor-examples/input")
# df_matrix.to_csv(base/"liver_gtex_tpm.csv",index=False)
filtered_tf_list1 = filter_tfs_by_mean_and_variance(df_matrix, tf_list, min_mean =2, min_var = 0.75)
filtered_tf_list1

Total TFs evaluated: 1383
TFs passing filter (Mean >= 2, Variance >= 0.75): 569

['ADNP',
 'ADNP2',
 'AEBP2',
 'AHCTF1',
 'AHR',
 'AR',
 'ARID3B',
 'ARID3C',
 'ARID5A',
 'ARID5B',
 'ARNT',
 'ASCL1',
 'ATF1',
 'ATF2',
 'ATF3',
 'ATF4',
 'ATF5',
 'ATF6',
 'ATF6B',
 'ATF7',
 'ATOH7',
 'ATOH8',
 'BACH1',
 'BATF',
 'BBX',
 'BCL6',
 'BCL6B',
 'BHLHA15',
 'BHLHE40',
 'BPTF',
 'BRF2',
 'CDC5L',
 'CEBPA',
 'CEBPB',
 'CEBPD',
 'CEBPG',
 'CEBPZ',
 'CENPB',
 'CIC',
 'CLOCK',
 'CREB3',
 'CREB3L2',
 'CREB3L4',
 'CREBL2',
 'CREBZF',
 'CREM',
 'CTCF',
 'CUX1',
 'CUX2',
 'CXXC1',
 'DBP',
 'DDIT3',
 'DEAF1',
 'DMRTA1',
 'DNMT1',
 'DNTTIP1',
 'DR1',
 'E2F3',
 'E2F4',
 'E2F6',
 'E4F1',
 'EBF4',
 'EGR1',
 'ELF1',
 'ELF2',
 'ELF3',
 'ELK1',
 'ELK3',
 'ELK4',
 'EPAS1',
 'ERF',
 'ESR1',
 'ESRRA',
 'ETS1',
 'ETS2',
 'ETV3',
 'ETV5',
 'ETV6',
 'FBXL19',
 'FLI1',
 'FOS',
 'FOSB',
 'FOSL1',
 'FOSL2',
 'FOXA1',
 'FOXA2',
 'FOXA3',
 'FOXF1',
 'FOXJ2',
 'FOXJ3',
 'FOXK1',
 'FOXK2',
 'FOXN2',
 'FOXN3',
 'FOXO1',
 'FOXO3',
 'FOXO4',
 'FOXP1',
 'FOXP4',
 'FOXQ1',
 'GABPA',
 'GATA4',
 'GATA6',
 'GLI4',
 'GLIS2',
 'GMEB2',
 'GRHL1',
 'GTF2B',
 'GTF2I',
 'GTF2IRD1',
 'GTF3A',
 'GZF1',
 'HAND2',
 'HBP1',
 'HES1',
 'HES4',
 'HES6',
 'HEY2',
 'HEYL',
 'HHEX',
 'HIC1',
 'HIF1A',
 'HINFP',
 'HIVEP1',
 'HIVEP2',
 'HLF',
 'HLX',
 'HMBOX1',
 'HMG20A',
 'HMG20B',
 'HMGA1',
 'HMGN3',
 'HNF1A',
 'HNF1B',
 'HNF4A',
 'HNF4G',
 'HOMEZ',
 'HSF1',
 'HSF2',
 'HSF4',
 'IKZF5',
 'IRF1',
 'IRF2',
 'IRF3',
 'IRF5',
 'IRF6',
 'IRF7',
 'IRF8',
 'IRF9',
 'JAZF1',
 'JDP2',
 'JUN',
 'JUNB',
 'JUND',
 'KDM2A',
 'KDM2B',
 'KDM5B',
 'KLF10',
 'KLF11',
 'KLF12',
 'KLF13',
 'KLF15',
 'KLF16',
 'KLF2',
 'KLF3',
 'KLF4',
 'KLF6',
 'KLF7',
 'KLF9',
 'KMT2A',
 'LBX2',
 'LHX2',
 'MAF',
 'MAFB',
 'MAFF',
 'MAFG',
 'MAFK',
 'MAX',
 'MAZ',
 'MBD1',
 'MBD2',
 'MBD4',
 'MBNL2',
 'MECP2',
 'MEF2A',
 'MEF2B',
 'MEF2D',
 'MEIS2',
 'MLX',
 'MLXIP',
 'MLXIPL',
 'MNT',
 'MSC',
 'MTERF1',
 'MTF1',
 'MXD1',
 'MXD3',
 'MXD4',
 'MXI1',
 'MYC',
 'MYCL',
 'MYNN',
 'MYPOP',
 'MYRF',
 'MZF1',
 'NFAT5',
 'NFATC3',
 'NFE2L1',
 'NFE2L2',
 'NFE2L3',
 'NFIA',
 'NFIB',
 'NFIC',
 'NFIL3',
 'NFIX',
 'NFKB1',
 'NFKB2',
 'NFYA',
 'NFYB',
 'NFYC',
 'NPAS2',
 'NR1D1',
 'NR1D2',
 'NR1H2',
 'NR1H3',
 'NR1H4',
 'NR1I2',
 'NR1I3',
 'NR2C1',
 'NR2F1',
 'NR2F2',
 'NR2F6',
 'NR3C1',
 'NR3C2',
 'NR4A1',
 'NR4A2',
 'NR5A2',
 'NRF1',
 'ONECUT1',
 'ONECUT2',
 'PATZ1',
 'PBX2',
 'PBX3',
 'PHF21A',
 'PKNOX1',
 'PLAGL2',
 'POU5F1',
 'POU6F1',
 'PPARA',
 'PPARD',
 'PPARG',
 'PRDM10',
 'PRDM15',
 'PRDM4',
 'PROX1',
 'PURA',
 'RARA',
 'RBAK',
 'RBPJ',
 'RELA',
 'RELB',
 'REST',
 'RFX1',
 'RFX5',
 'RLF',
 'RORA',
 'RORC',
 'RREB1',
 'RUNX1',
 'RXRA',
 'RXRB',
 'SALL1',
 'SATB1',
 'SCMH1',
 'SETBP1',
 'SETDB1',
 'SIX5',
 'SMAD1',
 'SMAD3',
 'SMAD4',
 'SMAD5',
 'SNAI1',
 'SNAI2',
 'SNAPC4',
 'SNAPC5',
 'SOX12',
 'SOX13',
 'SOX18',
 'SOX4',
 'SOX5',
 'SOX7',
 'SOX9',
 'SP1',
 'SP100',
 'SP110',
 'SP2',
 'SP3',
 'SPI1',
 'SREBF1',
 'SREBF2',
 'SRF',
 'STAT1',
 'STAT2',
 'STAT3',
 'STAT4',
 'STAT5A',
 'STAT5B',
 'STAT6',
 'TBP',
 'TBPL1',
 'TBX15',
 'TBX2',
 'TBX3',
 'TCF12',
 'TCF3',
 'TCF7',
 'TCF7L1',
 'TCF7L2',
 'TCFL5',
 'TEAD1',
 'TEAD2',
 'TEAD3',
 'TEAD4',
 'TEF',
 'TERF1',
 'TERF2',
 'TFAP4',
 'TFCP2',
 'TFDP1',
 'TFDP2',
 'TFE3',
 'TFEB',
 'TGIF1',
 'TGIF2',
 'THAP1',
 'THAP11',
 'THAP12',
 'THAP5',
 'THRA',
 'THRB',
 'TOPORS',
 'TP53',
 'UBP1',
 'USF1',
 'USF2',
 'VEZF1',
 'XBP1',
 'XPA',
 'YBX1',
 'YY1',
 'ZBED1',
 'ZBTB1',
 'ZBTB11',
 'ZBTB14',
 'ZBTB16',
 'ZBTB17',
 'ZBTB18',
 'ZBTB2',
 'ZBTB22',
 'ZBTB25',
 'ZBTB33',
 'ZBTB38',
 'ZBTB39',
 'ZBTB4',
 'ZBTB40',
 'ZBTB42',
 'ZBTB43',
 'ZBTB44',
 'ZBTB45',
 'ZBTB47',
 'ZBTB48',
 'ZBTB5',
 'ZBTB7A',
 'ZBTB7B',
 'ZEB1',
 'ZFP1',
 'ZFP62',
 'ZFP90',
 'ZFP91',
 'ZFPM1',
 'ZFX',
 'ZHX1',
 'ZHX2',
 'ZHX3',
 'ZIC1',
 'ZKSCAN1',
 'ZKSCAN5',
 'ZKSCAN8',
 'ZNF10',
 'ZNF12',
 'ZNF121',
 'ZNF131',
 'ZNF133',
 'ZNF141',
 'ZNF143',
 'ZNF146',
 'ZNF160',
 'ZNF174',
 'ZNF18',
 'ZNF181',
 'ZNF182',
 'ZNF189',
 'ZNF195',
 'ZNF202',
 'ZNF205',
 'ZNF207',
 'ZNF211',
 'ZNF212',
 'ZNF213',
 'ZNF217',
 'ZNF219',
 'ZNF22',
 'ZNF224',
 'ZNF226',
 'ZNF23',
 'ZNF232',
 'ZNF24',
 'ZNF248',
 'ZNF25',
 'ZNF251',
 'ZNF254',
 'ZNF263',
 'ZNF266',
 'ZNF267',
 'ZNF274',
 'ZNF275',
 'ZNF276',
 'ZNF280D',
 'ZNF281',
 'ZNF282',
 'ZNF3',
 'ZNF30',
 'ZNF302',
 'ZNF316',
 'ZNF317',
 'ZNF319',
 'ZNF32',
 'ZNF324',
 'ZNF326',
 'ZNF329',
 'ZNF331',
 'ZNF335',
 'ZNF33A',
 'ZNF33B',
 'ZNF341',
 'ZNF343',
 'ZNF354A',
 'ZNF354B',
 'ZNF358',
 'ZNF362',
 'ZNF367',
 'ZNF37A',
 'ZNF384',
 'ZNF385A',
 'ZNF394',
 'ZNF397',
 'ZNF408',
 'ZNF410',
 'ZNF414',
 'ZNF419',
 'ZNF429',
 'ZNF432',
 'ZNF433',
 'ZNF436',
 'ZNF438',
 'ZNF44',
 'ZNF441',
 'ZNF444',
 'ZNF467',
 'ZNF48',
 'ZNF480',
 'ZNF496',
 'ZNF503',
 'ZNF507',
 'ZNF513',
 'ZNF514',
 'ZNF516',
 'ZNF517',
 'ZNF524',
 'ZNF529',
 'ZNF544',
 'ZNF548',
 'ZNF550',
 'ZNF552',
 'ZNF554',
 'ZNF558',
 'ZNF561',
 'ZNF563',
 'ZNF564',
 'ZNF574',
 'ZNF579',
 'ZNF580',
 'ZNF581',
 'ZNF584',
 'ZNF587',
 'ZNF589',
 'ZNF592',
 'ZNF595',
 'ZNF598',
 'ZNF605',
 'ZNF621',
 'ZNF623',
 'ZNF627',
 'ZNF628',
 'ZNF629',
 'ZNF639',
 'ZNF641',
 'ZNF646',
 'ZNF652',
 'ZNF653',
 'ZNF654',
 'ZNF655',
 'ZNF662',
 'ZNF664',
 'ZNF672',
 'ZNF680',
 'ZNF684',
 'ZNF687',
 'ZNF688',
 'ZNF689',
 'ZNF69',
 'ZNF691',
 'ZNF692',
 'ZNF697',
 'ZNF700',
 'ZNF706',
 'ZNF710',
 'ZNF721',
 'ZNF74',
 'ZNF740',
 'ZNF746',
 'ZNF75A',
 'ZNF75D',
 'ZNF76',
 'ZNF761',
 'ZNF764',
 'ZNF768',
 'ZNF770',
 'ZNF771',
 'ZNF776',
 'ZNF777',
 'ZNF780B',
 'ZNF784',
 'ZNF785',
 'ZNF786',
 'ZNF787',
 'ZNF789',
 'ZNF79',
 'ZNF791',
 'ZNF800',
 'ZNF814',
 'ZNF823',
 'ZNF830',
 'ZNF837',
 'ZNF84',
 'ZNF841',
 'ZNF844',
 'ZSCAN16',
 'ZSCAN18',
 'ZSCAN21',
 'ZSCAN25',
 'ZSCAN26',
 'ZSCAN29',
 'ZSCAN30',
 'ZSCAN32',
 'ZSCAN9',
 'ZXDB',
 'ZXDC',
 'ZZZ3']

# Convert to DataFrame and specify column names
df = pd.DataFrame(filtered_tf_list1, columns=['Name'])

# Save to CSV (index=False prevents writing the row numbers)
# df.to_csv('output.csv', index=False)
df.to_csv(base+"/liver_gtex_tf.csv",index=False)

Selecting Transcription factor sources

The list of transcription factors contains all types of genes including General TFs and housekeeping genes. These genes typically are highly expressed across samples with a lower variance. Since the tool relies of a random forest model which when taking a decision picks a feature which leads to the highest reduction in impurity, it tends to pick up these highly expressed low variance TFs. This leads to the generation of false positives.

from coregtor import PlotExprData
PlotExprData(df_matrix, tf_list, log_scale=True)

Based on the plot we can likely classify genes in 4 categories:

• low mean, low variance:

• high mean, low variable:

• low mean, high variance:

• high mean, high variance:

We can use this information to filter the TF list.

def filter_tfs_by_mean_and_variance(df_matrix, tf_list, min_mean=1.5, min_var=0.5, log_transformed=True):
    """
    Filters TFs using a joint Mean Expression and Variance cutoff on log2-transformed data.
    """
    # Convert to Log2(TPM + 1) if not already done
    data = df_matrix if log_transformed else np.log2(df_matrix + 1)
    
    # 1. Strictly intersect candidates with valid TF list
    tf_set = set(tf_list)
    valid_tfs = [gene for gene in data.columns if gene in tf_set]
    tf_data = data[valid_tfs]
    
    # 2. Calculate stats
    means = tf_data.mean(axis=0)
    variances = tf_data.var(axis=0)
    
    # 3. Apply dual filter
    pass_mean = means >= min_mean
    pass_var = variances >= min_var
    
    selected_tfs = means[pass_mean & pass_var].index.tolist()
    
    print(f"Total TFs evaluated: {len(valid_tfs)}")
    print(f"TFs passing filter (Mean >= {min_mean}, Variance >= {min_var}): {len(selected_tfs)}")
    
    return selected_tfs

filtered_tf_list = filter_tfs_by_mean_and_variance(df_matrix, tf_list, min_mean = 1.5, min_var = 0.75)

Total TFs evaluated: 1383
TFs passing filter (Mean >= 1.5, Variance >= 0.75): 602

Running CoRegTor the easy way

To run the tool on a single target, the easist way is to run RunCoRegTor

import importlib
import coregtor
from coregtor import RunCoRegTor

importlib.reload(coregtor)
from coregtor import RunCoRegTor

target_gene = "CYP3A4"
target_gene1 = "TTR"
results, details = RunCoRegTor(df_matrix,filtered_tf_list,target_gene1,ensemble_options= {"random_state":100,"n_estimators":1000})
results

	cluster_uid	target	sources	n_source	silhouette_score	fi_score	note	cluster_note
0	d7dd3c8533ba09	TTR	ZFP1;NFIA	2	0.04204	0.02800		leiden
1	d8959b05a42c0a	TTR	ARID3C;TEAD2;DBP	3	0.12160	0.13587		leiden
2	bb6ef7a3c93095	TTR	TEAD4;NR1I3;MLXIP;E2F4;ELF3;GLIS2;ZNF217;ESR1;...	11	0.39537	0.00610		leiden

coreg_list = results.iloc[2]["sources"]
coreg_list

'TEAD4;NR1I3;MLXIP;E2F4;ELF3;GLIS2;ZNF217;ESR1;HMGN3;STAT3;ONECUT2'

Manual validation of a biological case study

CYP3A4 is an important enzyme in humans, which is mainly found in the liver and interstine and is involved in drug metabolism. CYP3A4 changes the chemical structure of foreign molecule to help process them.

Key regulators / Biological pathways found :

• Master Poineer factors in liver: “HNF4A, FOXA2, FOXA3” are master liver poineer factors. Foxa1/2/3 act as pioneer factors that bind and open closed chromatin, then sequentially and cooperatively recruit Hnf4α to adjacent Hnf4α-binding sites to activate liver-specific genes. (https://doi.org/10.1016/j.molcel.2020.07.012)

• Drugs & Xenobiotics (Upregulate): When the liver detects foreign chemicals, nuclear receptors (like PXR, CAR, and NR3C1) bind to the CYP3A4 promoter to drastically increase its production. The goal is to metabolize and clear the invading chemical as fast as possible. (PMID: 12514743)

Hormonal Stress (Upregulates): Chronic psychological, metabolic, or starvation stress releases cortisol (glucocorticoids) into the bloodstream. This activates the Glucocorticoid Receptor (NR3C1), which synergizes with HNF4A to upregulate CYP3A4 expression. (PMID: 12815172)

• NR3C1 (the Glucocorticoid Receptor or GR) handles the drug and stress-induced upregulation of CYP3A4. However, GR cannot bind effectively to the CYP3A4 promoter on its own; it requires HNF4A to act as a physical and epigenetic anchor

liver_CYP3A4_evidence = [
    {"target":"CYP3A4","sources":["HNF4A","FOXA2","FOXA3"],"title":"PMID: 32755593"},
     {"target":"CYP3A4","sources":["HNF4A","NR3C1"],"title":"PMCID: PMC2981672"}
]

from coregtor import  PlotClusters
PlotClusters(results, details,liver_CYP3A4_evidence,{"figsize":(8,7),"font_size": 8,"title_prefix":"Liver (GTEx) "})

[<Figure size 800x700 with 1 Axes>]

Exploring the complete CoRegTor pipeline

Main steps:

• Generate model input

• Train RF model

• Extract trees from the forest and paths from the forest

• Find most predictive nodes and gets generate context for comparasion

• Transform context into gene frequency histogram and generate similarity matrix

• Clustering to generate the final coregulatory clusters

from coregtor import create_model_input,create_model,tree_paths,create_context,transform_context,compare_context,identify_coregulators

target_gene = "CTSD"
X, Y = create_model_input(df_matrix,target_gene, filtered_tf_list)

print("CTSD" in df_matrix.columns)
print("PPARG" in filtered_tf_list)
print("CEBPA" in filtered_tf_list)

True
True
True

ensemble_method = "rf"
ensemble_options = {"random_state":100,"n_estimators":1000,"max_depth":5,"n_jobs":-1}
model, fi = create_model( X, Y, method=ensemble_method, options=ensemble_options)    

paths = tree_paths(model, X, Y)

contexts = create_context(paths)

transformed = transform_context(contexts)

distance_measure = "canberra_distance"
matrix = compare_context(transformed, method=distance_measure, transformation_type="gene_frequency")

matrix

	YBX1	ZNF219	ATF6B	MBD2	SP2	TFE3	NR1H2	DNTTIP1	CUX1	HMG20A	...	ZFP91	CREBL2	TERF2	NFKB1	HBP1	MLX	ERG	TFCP2	DEAF1	ZNF664
YBX1	0.000000	332.181490	253.082356	419.666318	385.886185	460.700980	450.821560	233.378400	248.832909	417.947598	...	487.696746	536.874446	532.797166	406.980229	497.215448	450.377514	514.678629	532.537825	533.497858	548.982269
ZNF219	332.181490	0.000000	321.730051	377.738136	333.360181	380.248591	363.953705	328.310608	329.612886	360.515381	...	389.819154	415.990929	422.822410	338.679764	387.387984	356.637332	400.945510	410.901287	419.702015	428.358010
ATF6B	253.082356	321.730051	0.000000	437.016544	389.363873	465.445773	452.267204	242.679486	248.624925	437.375279	...	492.990379	541.760149	532.798987	409.270185	491.869520	450.086576	516.943465	536.482112	534.398107	552.539627
MBD2	419.666318	377.738136	437.016544	0.000000	309.700191	254.046801	290.254233	442.771311	437.735386	278.012232	...	266.531541	255.831958	260.568625	280.195244	282.354725	270.652173	275.988852	247.782355	273.769072	262.493758
SP2	385.886185	333.360181	389.363873	309.700191	0.000000	282.503211	299.432995	399.756665	388.864669	298.658323	...	301.910928	296.322058	298.458019	288.612948	314.540277	292.639244	317.273512	300.335269	306.443660	305.589781
TFE3	460.700980	380.248591	465.445773	254.046801	282.503211	0.000000	260.465135	480.262129	469.424627	270.837091	...	218.852457	201.382084	199.083829	292.140043	249.581763	252.068809	231.105999	209.970684	233.283792	205.234445
NR1H2	450.821560	363.953705	452.267204	290.254233	299.432995	260.465135	0.000000	453.477111	447.256167	284.587511	...	253.400125	247.596290	261.367023	297.315883	262.020763	280.711767	277.367102	251.936705	259.663210	251.194711
DNTTIP1	233.378400	328.310608	242.679486	442.771311	399.756665	480.262129	453.477111	0.000000	202.312622	442.524751	...	510.042041	564.937054	556.856879	418.486499	512.173643	463.918002	524.920771	557.651346	551.390276	574.468439
CUX1	248.832909	329.612886	248.624925	437.735386	388.864669	469.424627	447.256167	202.312622	0.000000	420.312506	...	501.189978	551.405553	543.709029	408.255767	504.495465	462.268505	518.610314	550.158766	540.732085	562.717838
HMG20A	417.947598	360.515381	437.375279	278.012232	298.658323	270.837091	284.587511	442.524751	420.312506	0.000000	...	258.472330	258.384624	249.389829	279.382909	280.712501	269.313605	276.598347	262.420697	277.747336	260.863778
SP100	360.031476	331.856229	375.612627	323.938448	317.883807	308.838101	332.234406	379.052495	371.995149	293.814130	...	316.667136	328.601043	319.156430	319.134461	339.689268	311.444002	324.987007	331.078797	344.646922	338.515377
ETV5	479.733393	383.149169	478.934318	275.313665	307.432726	228.582378	283.548691	498.582797	491.391856	275.379046	...	215.864594	172.575944	183.935435	320.327626	228.437753	240.949622	209.168383	185.573984	196.190522	171.185630
NFE2L1	497.335720	386.208086	493.096654	274.943100	289.368036	201.884760	264.956671	515.839685	510.878733	288.553716	...	202.338225	161.183041	158.410660	294.953229	220.685956	238.776009	192.258436	160.495116	170.254858	150.929689
STAT6	497.619660	390.539658	506.441649	272.853575	315.059662	231.529093	273.191776	521.794912	515.051166	276.685990	...	220.611503	157.812743	170.670263	308.757068	233.279210	251.396466	205.576381	165.618628	192.870305	152.433151
KDM5B	530.735279	418.898149	533.683651	256.166096	311.327891	216.122083	260.414641	557.201103	549.241993	259.826282	...	166.985638	87.486497	112.299850	304.299676	189.531997	229.086041	163.593114	113.758047	124.509229	84.402149
RELA	528.165403	420.535499	530.412299	251.191519	295.492520	207.288478	252.619045	550.114977	537.179666	270.966014	...	161.376955	114.506896	116.707019	291.185500	202.323254	235.871902	175.138634	115.780049	146.193291	105.780415
LYL1	523.375635	416.658857	529.957862	273.845633	316.253090	234.173578	281.639060	545.303932	536.571154	263.452511	...	193.092794	131.845726	151.963518	317.401869	203.197809	251.526342	186.374383	142.441993	151.235452	123.589955
ZFP91	487.696746	389.819154	492.990379	266.531541	301.910928	218.852457	253.400125	510.042041	501.189978	258.472330	...	0.000000	163.541497	164.304392	291.099036	206.690580	226.965816	198.827140	160.939662	185.227423	156.168043
CREBL2	536.874446	415.990929	541.760149	255.831958	296.322058	201.382084	247.596290	564.937054	551.405553	258.384624	...	163.541497	0.000000	85.234835	290.466234	178.554690	223.270094	148.747300	90.312799	107.821613	61.605128
TERF2	532.797166	422.822410	532.798987	260.568625	298.458019	199.083829	261.367023	556.856879	543.709029	249.389829	...	164.304392	85.234835	0.000000	289.675180	191.060533	218.771417	167.058986	109.135276	135.248617	83.543678
NFKB1	406.980229	338.679764	409.270185	280.195244	288.612948	292.140043	297.315883	418.486499	408.255767	279.382909	...	291.099036	290.466234	289.675180	0.000000	320.207267	289.254708	312.580124	296.628752	302.354292	298.764823
HBP1	497.215448	387.387984	491.869520	282.354725	314.540277	249.581763	262.020763	512.173643	504.495465	280.712501	...	206.690580	178.554690	191.060533	320.207267	0.000000	253.918682	211.573432	180.975049	199.640043	177.633680
MLX	450.377514	356.637332	450.086576	270.652173	292.639244	252.068809	280.711767	463.918002	462.268505	269.313605	...	226.965816	223.270094	218.771417	289.254708	253.918682	0.000000	232.010514	223.505170	223.390175	222.222270
ERG	514.678629	400.945510	516.943465	275.988852	317.273512	231.105999	277.367102	524.920771	518.610314	276.598347	...	198.827140	148.747300	167.058986	312.580124	211.573432	232.010514	0.000000	151.386516	163.258831	147.376966
TFCP2	532.537825	410.901287	536.482112	247.782355	300.335269	209.970684	251.936705	557.651346	550.158766	262.420697	...	160.939662	90.312799	109.135276	296.628752	180.975049	223.505170	151.386516	0.000000	118.787193	82.722700
DEAF1	533.497858	419.702015	534.398107	273.769072	306.443660	233.283792	259.663210	551.390276	540.732085	277.747336	...	185.227423	107.821613	135.248617	302.354292	199.640043	223.390175	163.258831	118.787193	0.000000	98.313349
ZNF664	548.982269	428.358010	552.539627	262.493758	305.589781	205.234445	251.194711	574.468439	562.717838	260.863778	...	156.168043	61.605128	83.543678	298.764823	177.633680	222.222270	147.376966	82.722700	98.313349	0.000000

27 rows × 27 columns

# print("NR1I2"  in filtered_tf_list)
# print("NR1I3" in filtered_tf_list)
print("PPARG" in filtered_tf_list)
print("CEBPA" in filtered_tf_list)

True
True

matrix.columns

Index(['YBX1', 'ZNF219', 'ATF6B', 'MBD2', 'SP2', 'TFE3', 'NR1H2', 'DNTTIP1',
       'CUX1', 'HMG20A', 'SP100', 'ETV5', 'NFE2L1', 'STAT6', 'KDM5B', 'RELA',
       'LYL1', 'ZFP91', 'CREBL2', 'TERF2', 'NFKB1', 'HBP1', 'MLX', 'ERG',
       'TFCP2', 'DEAF1', 'ZNF664'],
      dtype='object')

paths

	tree	source	target	path_length	node1	node2	node3	node4	node5	node6	node7	node8	node9
0	0	NFE2L2	CYP3A4	6	DDIT3	TEAD2	HOMEZ	ZNF581	NFE2L2	None	None	None	None
1	0	NFE2L2	CYP3A4	10	DDIT3	TEAD2	HLF	ZNF74	MEF2A	DBP	HAND2	HHEX	RXRA
2	0	NFE2L2	CYP3A4	9	DDIT3	TEAD2	HOMEZ	GTF2IRD1	NFIX	HES6	ZNF584	ZNF710	None
3	0	NFE2L2	CYP3A4	10	DDIT3	TEAD2	HOMEZ	GTF2IRD1	NFIX	HES6	ZNF655	HSF4	FOSL1
4	0	NFE2L2	CYP3A4	10	DDIT3	TEAD2	HOMEZ	GTF2IRD1	NFIX	MYC	THAP5	KLF2	TEF
...	...	...	...	...	...	...	...	...	...	...	...	...	...
49097	999	LTF	CYP3A4	10	XPA	TEAD2	SNAPC5	TCF7	ZNF274	FOXN3	RELB	HOXB4	CREB3L2
49098	999	LTF	CYP3A4	9	XPA	TEAD2	SNAPC5	TCF7	ZNF274	FOXN3	ZXDB	HES4	None
49099	999	LTF	CYP3A4	7	XPA	TEAD2	SNAPC5	TCF7	ZNF529	ZNF224	None	None	None
49100	999	LTF	CYP3A4	10	XPA	TEAD2	SNAPC5	TCF7	ZNF529	ZNF334	NFIC	ZNF202	RARG
49101	999	LTF	CYP3A4	8	XPA	TEAD2	ZNF513	KLF15	ZNF205	ZNF628	ZNF688	None	None

49102 rows × 13 columns

for col in paths.columns:
    print(col)
    print("PPARG" in paths[col])

tree
False
source
False
target
False
path_length
False
node1
False
node2
False
node3
False
node4
False
node5
False
node6
False
node7
False
node8
False
node9
False

cluster_method = "community_detection"
cluster_options =  {"edge_creation_method": "threshold_percentile","edge_creation_value": 0.25, "resolution": 0.5}
results = identify_coregulators(matrix, target_gene, cluster_method, cluster_options, feature_importance=fi)

results

	cluster_uid	target	sources	n_source	silhouette_score	fi_score	note	cluster_note
0	d1c8a342cb550d	CTSD	DNTTIP1;CUX1;ATF6B;YBX1;SP100;ZNF219;NFKB1;SP2...	24	0.27569	0.02816		leiden

results.iloc[-1]["sources"]

'DNTTIP1;CUX1;ATF6B;YBX1;SP100;ZNF219;NFKB1;SP2;HMG20A;NR1H2;MBD2;TFE3;HBP1;MLX;LYL1;ETV5;ERG;NFE2L1;STAT6;ZFP91;DEAF1;RELA;TERF2;KDM5B'