%load_ext autoreload
%autoreload 2

import numpy as np
import pandas as pd
import seaborn as sns
from matplotlib import pyplot as plt
from collections import OrderedDict

from matlas.pwms import ic_scale, adjust_for_ic_scale
from matlas.matches import vdom_pssm
from modisco.visualization import viz_sequence
from matlas.matches import vdom_pssm

motif_name = 'CEBPB'

top_k_filters = OrderedDict()
filt_infl_from_loss = OrderedDict()
for fold_no in range(1,11):
    aitacdir = "/mnt/lab_data2/msharmin/oc-atlas/DanSkinData/alex_bpnets/{0}/{1}".format(motif_name, fold_no)
    filt_infl_from_loss[fold_no] = np.load(aitacdir+"/filt_infl_from_loss.npy")
    top_k_filters[fold_no] = np.flip(np.argsort(filt_infl_from_loss[fold_no]))[:15]
    print(fold_no, top_k_filters[fold_no])

1 [47 41 60 33 56 38  8  0 53  5 43 63 13  2 48]
2 [14 29  1  5  9 40 58 63 46 60 22 55 39 25 24]
3 [10 32 51 54 15 29  6 39  7 11 38 35 46 18 55]
4 [47  3 43 21 44 42 11  4 52  6  9 50 20 39 31]
5 [36  3  2 53 14 49 47 27 38 11 26 17 30 42 50]
6 [49 35 32 11 34 58 46 24  3  1 57 54 38 26 12]
7 [54  8 32  0  7 45 16 57  4 60  3 15 50 40 43]
8 [17 31 56 11  4 26 63 62  7 50  0 14 19 18 44]
9 [25 30 21 12  4 55 61  7 36 40 23  8 11 37 51]
10 [25 26  6  2 48 51 11 43 63  7 28 54 42 47 18]

import numpy as np
import seaborn as sns

motif_name = 'CEBPB'
filter_weights = OrderedDict()
for fold_no in range(1,11):
    filter_weights[fold_no] = np.load("/mnt/lab_data2/msharmin/oc-atlas/DanSkinData/alex_bpnets/filters/{0}/{0}_fold{1}_filters.npy".format(
    motif_name, fold_no))

from matplotlib import pyplot as plt

from matlas.pwms import ic_scale, adjust_for_ic_scale
from matlas.matches import vdom_pssm
from modisco.visualization import viz_sequence
from matlas.matches import vdom_pssm


def draw_filter_weights(start_filter, end_filter):
    
    alphabets = np.array(['A', 'C', 'G', 'T'])

    fig, axes = plt.subplots(4, 4, figsize=(30,10))

    for i in range(start_filter, end_filter):
        a = np.copy(filter_weights[:,:,i].T)
        r,c = (i-start_filter)//4, i%4
        #print(r, c)
        ax = sns.heatmap(a, yticklabels=alphabets,
                    xticklabels=alphabets[np.argmax(a, axis=0)], ax = axes[r, c]
                   )
        ax.set_ylabel("filter{}".format(i))
        
    return None

Raw and mean normalized filter weights¶

from modisco.visualization import viz_sequence

for fold_no in range(1,11):
    for i in top_k_filters[fold_no]:
        print('fold_no', fold_no, 'filter', i)
        mean_norm_weights = np.copy(filter_weights[fold_no][:,:,i])
        mean_norm_weights = mean_norm_weights - mean_norm_weights.mean(1, keepdims=True)
        print('raw weights')
        viz_sequence.plot_weights(filter_weights[fold_no][:,:,i])
        print('mean normalized weights')
        viz_sequence.plot_weights(mean_norm_weights)
        break

fold_no 1 filter 47
raw weights

mean normalized weights

fold_no 2 filter 14
raw weights

mean normalized weights

fold_no 3 filter 10
raw weights

mean normalized weights

fold_no 4 filter 47
raw weights

mean normalized weights

fold_no 5 filter 36
raw weights

mean normalized weights

from matlas.matches import vdom_pssm
from IPython.display import display

from vdom.helpers import (h1, p, li, img, div, b, br, ul, img, a, 
                          details, summary,
                          table, thead, th, tr, tbody, td, ol)

for fold_no in range(1,11):
    items_to_display = []
    for i in top_k_filters[fold_no]:
        filter_name = 'filter{}'.format(i)
        mean_norm_weights = np.copy(filter_weights[fold_no][:,:,i])
        mean_norm_weights = mean_norm_weights - mean_norm_weights.mean(1, keepdims=True)
        un_query = vdom_pssm(filter_weights[fold_no][:,:,i])
        n_query = vdom_pssm(mean_norm_weights)
        items_to_display.append(summary(b(filter_name), p('raw weights'), un_query, 'mean normalized', n_query))
        #break
    display(details(summary(b("weights of selected filters in fold {}".format(fold_no))), summary(items_to_display)))

from matlas.pwms import ic_scale, adjust_for_ic_scale
from matlas.matches import vdom_pssm
from sklearn.preprocessing import MinMaxScaler
from collections import OrderedDict

def ic_scale_filter_weights(filter_weights, filter_count=64):
    norm_weights = np.copy(filter_weights)
    norm_weights = (norm_weights-np.mean(norm_weights))/np.std(norm_weights)
    norm_weights[norm_weights>2] = 2.0
    norm_weights[norm_weights<-2] = -2.0

    ic_scaled_weights = OrderedDict()
    for i in range(filter_count):
        weights = np.copy(norm_weights[:,:,i])
        scaler = MinMaxScaler()
        arr = scaler.fit_transform(weights)

        #convert to probability
        arr = arr / arr.sum(1, keepdims=True) #divide by colsum
        arr = arr + 0.01  # add pseudo-counts
        probs = arr / arr.sum(1, keepdims=True)
        #probs = adjust_for_ic_scale(probs)
        try:
            ic_scaled_weights['filter{0}'.format(i)] = ic_scale(probs)
        except AssertionError:
            print('filter{0}'.format(i))
            return probs

    return ic_scaled_weights


ic_scaled_weights = OrderedDict()
for fold_no in range(1,11):
    ic_scaled_weights[fold_no] = ic_scale_filter_weights(filter_weights[fold_no])

filter weights in IC scale¶

All filter weights are standardised and each filter weights are scaled to 0-1 and converted to probability score before IC-scaling

# from IPython.display import display
# from matlas.pwms import ic_scale

# from vdom.helpers import (h1, p, li, img, div, b, br, ul, img, a, 
#                           details, summary,
#                           table, thead, th, tr, tbody, td, ol)

# items_to_display = []
# for i in range(64):
#     filter_name = 'filter{}'.format(i)
#     query = vdom_pssm(ic_scaled_weights[filter_name])
#     items_to_display.append(summary(filter_name, query))

# display(details(summary(b("filter weights in IC scale")), summary(items_to_display)))

#4,7,57

for fold_no in range(1,11):
    selected_items_to_display = []
    for i in top_k_filters[fold_no]:
        filter_name = 'filter{}'.format(i)
        query = vdom_pssm(ic_scaled_weights[fold_no][filter_name])
        selected_items_to_display.append(summary(filter_name, query))

    display(details(summary(b("Click here to see the filter weights in IC scale for fold {}".format(fold_no))), 
                    summary(selected_items_to_display)))

importance score of the filtered sequences¶

imp_scores = np.load(aitacdir+"/imp_scores.npy")
rev_imp_scores = np.load(aitacdir+"/rev_imp_scores.npy")
OCR_matrix = np.load(aitacdir+"/OCR_matrix_0.npy")
activations2 = np.load(aitacdir+"/activations2_0.npy")
seq_indices_of_activation = np.load(aitacdir+"/seq_indices_of_activation_0.npy")

already_visited = []
for filter_no in top_k_filters:
    filter_name = "filter{}".format(filter_no)
    print(filter_name)
    activated_seqs = np.argwhere(OCR_matrix[filter_no,:]==1)[:,0]; print(activated_seqs.shape)
    activated_scores = imp_scores[activated_seqs]; print(activated_scores.shape)
    for i in range(20):
        if i in already_visited: continue
        print('seq', activated_seqs[i])
        viz_sequence.plot_weights(activated_scores[i, 600:746,:],
                                  subticks_frequency=10, figsize=(20,1))
        break
    break

filter54
(8464,)
(8464, 1346, 4)
seq 0

from matlas.sliding_similarities import aggregate_seqlets


seqlen = imp_scores.shape[1]
for filter_no in top_k_filters:
    activated_subseqs = np.argwhere(seq_indices_of_activation[filter_no]!=0); #print(activated_subseqs.shape)
    activated_subscores = []
    rev_activated_subscores = []
    sub_activations = []
    rev_sub_activations = []
    for i,j,k in zip(activated_subseqs.T[0], activated_subseqs.T[1], activated_subseqs.T[1]+21):
        activated_subscores.append(imp_scores[i, j:k, :])
        rev_activated_subscores.append(rev_imp_scores[i, (seqlen-k):(seqlen-j), :])
        sub_activations.append(activations2[i, filter_no, j])

    activated_subscores = np.array(activated_subscores); #print(activated_subscores.shape) 
    rev_activated_subscores = np.array(rev_activated_subscores); #print(rev_activated_subscores.shape)
    sub_activations = np.array(sub_activations)
    sub_act_ind = np.flip(np.argsort(sub_activations))
    activated_subscores = activated_subscores[sub_act_ind]
    rev_activated_subscores = rev_activated_subscores[sub_act_ind]
    
    avg_activated_subscores = aggregate_seqlets(activated_subscores[:100], 
                                                rev_activated_subscores[:100], 'sum'); #print(avg_activated_subscores.shape)
    print('Aggregated importance scores using direct sum for filter-{}'.format(filter_no))
    viz_sequence.plot_weights(avg_activated_subscores)
#     print("filter{}".format(filter_no))
#     print('running average')
#     viz_sequence.plot_weights(avg_activated_subscores)
#     viz_sequence.plot_weights(aggregate_seqlets(activated_subscores, rev_activated_subscores, 'old'))
    #break

Aggregated number of sequences is 68
Aggregated importance scores using direct sum for filter-54

Aggregated number of sequences is 70
Aggregated importance scores using direct sum for filter-8

Aggregated number of sequences is 68
Aggregated importance scores using direct sum for filter-32

Aggregated number of sequences is 75
Aggregated importance scores using direct sum for filter-0

Aggregated number of sequences is 82
Aggregated importance scores using direct sum for filter-7

Aggregated number of sequences is 66
Aggregated importance scores using direct sum for filter-45

Aggregated number of sequences is 66
Aggregated importance scores using direct sum for filter-16

Aggregated number of sequences is 67
Aggregated importance scores using direct sum for filter-57

Aggregated number of sequences is 77
Aggregated importance scores using direct sum for filter-4

Aggregated number of sequences is 72
Aggregated importance scores using direct sum for filter-60

aitac motifs¶

from matlas.matches import DenovoAitac
motif_name = 'CEBPB'
obs = OrderedDict()
for fold_no in range(1,11):
    aitacdir = "/mnt/lab_data2/msharmin/oc-atlas/DanSkinData/alex_bpnets/{0}/{1}".format(motif_name, fold_no)

    ob = DenovoAitac(aitacdir, filt_infl_from_loss[fold_no], 
                     top_k_filters[fold_no], len(top_k_filters[fold_no]))
    ob.fetch_tomtom_matches(
                meme_db="/mnt/lab_data/kundaje/users/msharmin/annotations/HOCOMOCOv11_core_pwms_HUMAN_mono.renamed.nonredundant.annotated.meme",
                database_name="HOCOMOCO.nonredundant.annotated",
                save_report=True, tomtom_dir= "{0}/{1}_tomtomout".format(aitacdir, "HOCOMOCO.nonredundant.annotated"))
# #     ob.fetch_tomtom_matches(
# #                 meme_db="/mnt/lab_data/kundaje/users/msharmin/annotations/HOCOMOCOv11_core_pwms_HUMAN_mono.renamed.nonredundant.annotated.meme",
# #                 database_name="HOCOMOCO.nonredundant.annotated",
# #                 save_report=True, tomtom_dir= "{0}/{1}_tomtomout".format(aitacdir, "HOCOMOCO.nonredundant.annotated"))


#     ob.load_matched_motifs(database_name="HOCOMOCO.nonredundant.annotated")
#     ob.get_motif_per_celltype(match_threshold=0.02, match_criteria='p-value', database_name="HOCOMOCO.nonredundant.annotated")
#     pattern_tab, pattern_dict = ob.visualize_pattern_table()
#     tf_tab, tf_dict = ob.visualize_tf_table("Aitac")
    
#     display(details(summary('Click here for ', b('Denovo Patterns'), ' by ', b('{}'.format('Aitac')),
#                             ' in ', b(motif_name), 'for fold-', b(str(fold_no)),
#                             ": #{}".format(len(pattern_dict)),
#                            ), pattern_tab))
    
#     display(details(summary('Click here for ', b('Motifs'), ' by ', b('{}'.format('Aitac')),
#                         ' in ', b(motif_name), 'for fold-', b(str(fold_no)),
#                         ": #{}".format(len(tf_dict)),
#                        ), tf_tab))
#     break
    
#     obs[fold_no] = ob

from vdom.helpers import (b, summary, details)
from IPython.display import display

ob_tofold = OrderedDict()
pattern_tab_tofold, pattern_dict_tofold = OrderedDict(), OrderedDict()
tf_tab_tofold, tf_dict_tofold = OrderedDict(), OrderedDict()

for fold_no in range(1,11):
    aitacdir = "/mnt/lab_data2/msharmin/oc-atlas/DanSkinData/alex_bpnets/{0}/{1}".format(motif_name, fold_no)

    ob = DenovoAitac(aitacdir, filt_infl_from_loss[fold_no], 
                     top_k_filters[fold_no], len(top_k_filters[fold_no]))
    ob.load_matched_motifs(database_name="HOCOMOCO.nonredundant.annotated")
    ob.get_motif_per_celltype(match_threshold=0.02, match_criteria='p-value', database_name="HOCOMOCO.nonredundant.annotated")
    pattern_tab, pattern_dict = ob.visualize_pattern_table()
    tf_tab, tf_dict = ob.visualize_tf_table("Aitac")
    
    ob_tofold[fold_no] = ob
    pattern_tab_tofold[fold_no], pattern_dict_tofold[fold_no] = pattern_tab, pattern_dict
    tf_tab_tofold[fold_no], tf_dict_tofold[fold_no] = tf_tab, tf_dict
    

    display(details(summary('Click here for ', b('Denovo Patterns'), ' by ', b('{}'.format('Aitac')),
                            ' in ', b(motif_name), ' for fold-', b(str(fold_no)),
                            ": #{}".format(len(pattern_dict)),
                           ), pattern_tab))
    #break

for fold_no in range(1,11):
    display(details(summary('Click here for ', b('Motifs'), ' by ', b('{}'.format('Aitac')),
                        ' in ', b(motif_name), ' for fold-', b(str(fold_no)),
                        ": #{}".format(len(tf_dict_tofold[fold_no])),
                       ), tf_tab_tofold[fold_no]))

from matlas.matches import vdom_pssm
from matlas.genome_data import *
from modisco.visualization import viz_sequence
from matlas.pwms import load_motifDB_id_maps

def display_redundancy():
    tf_to_fold_count = OrderedDict()
    for fold_no in range(1,11):
        for tfname in ob_tofold[fold_no].tf_to_pattern.keys():
            if tfname not in tf_to_fold_count:
                tf_to_fold_count[tfname] = 1
            else:
                tf_to_fold_count[tfname] += 1

    sorted_tfs = sorted(tf_to_fold_count.items(), key=lambda x: x[1], reverse=True)
#     return sorted_tfs, 0
    motifDB_data = load_motifDB_id_maps(database_name='HOCOMOCO.nonredundant.annotated', method_to_load=None)

    redundant_tfs = OrderedDict()
    redundant_ppms = OrderedDict() 
    for tfname, count in sorted_tfs:
        if tf_to_fold_count[tfname] >= 5:
            pssm = viz_sequence.ic_scale(motifDB_data['maps'][tfname]['PPM'], np.array(BP_BACKGROUND))
            redundant_ppms[tfname] = vdom_pssm(pssm, letter_width=0.15, height=0.5)
            redundant_tfs[tfname] = count 
    
#     tab = table(thead(th('TF Name'), th('PPM'), th("Fold count")),
#                   tbody([tr(td(b(name)), td(redundant_ppms[name]), 
#                             td(b(str(value))))
#                              for name, value in redundant_tfs.items()]
#                        )
#                     )
#     display(tab)
# include median significance, median influence
    return redundant_tfs, redundant_ppms

redundant_tfs, redundant_ppms = display_redundancy()
from vdom.helpers import (h1, p, li, img, div, b, br, ul, img, a, 
                          details, summary,
                          table, thead, th, tr, tbody, td, ol)
tab = table(thead(th('TF Name'), th('PPM'), th("Fold count")),
                  tbody([tr(td(b(name)), td(redundant_ppms[name]), 
                            td(b(str(value))))
                             for name, value in redundant_tfs.items()]
                       )
                    )
# display(tab)

display(details(summary('Click here for ', b('HOCOMOCCO Merged Motifs'), ' sorted by ', b('Redundancy'),
                        ' in ', b(motif_name), ), 
                tab))

Rsat clustering on aitac motifs¶

motif_name = 'CEBPB'
aitacdir = "/mnt/lab_data2/msharmin/oc-atlas/DanSkinData/fold_a_alex/{}".format(motif_name)
from matlas.pwms import get_motifDB_id_maps, reduce_pwm_redundancy
from matlas.genome_data import *

memefile = "{0}/filter_motifs_pwm.meme".format(aitacdir)
motif_id_maps = get_motifDB_id_maps(database_name=memefile)

file_prefix = "/mnt/lab_data2/msharmin/oc-atlas/DanSkinData/fold_a_alex/{}/aitac".format(motif_name)
reduce_pwm_redundancy(
        motif_id_maps,
        out_pwm_file="{}.hclust_pwms.tmp".format(file_prefix),
        pwmtype="probabilities", #probabilities
        tmp_prefix=file_prefix,
        pseudocount=PSEUDOCOUNT,
        ic_thresh=IC_THRESHOLD,
        cor_thresh=0.6,
        ncor_thresh=0.4,
        num_threads=28)

filter4 2
filter8 0
filter15 0
filter16 0
filter32 0
filter40 1
filter54 2
/mnt/lab_data2/msharmin/oc-atlas/DanSkinData/fold_a_alex/CEBPB/aitac.cor.motifs.mat.txt /mnt/lab_data2/msharmin/oc-atlas/DanSkinData/fold_a_alex/CEBPB/aitac.ncor.motifs.mat.txt
0
4 filter5 (4, 12) 0.9999999999999996
7 filter12 (4, 21) 0.9999999999999998
(26, 26)
filter0;filter19 0.7078714398899135 1.0
filter2;filter41 0.7297810025907647 1.0
saving out filter34
saving out filter42
saving out filter57
filter43;filter61 0.6862553250824988 1.0
filter1;filter3 0.7906007147545048 0.8894258040988179
filter36;filter52 0.6600176379662774 1.0
saving out filter60
saving out filter11
filter44;filter50 0.6855557454701704 1.0
filter26;filter39 0.7108863479037282 1.0
saving out filter0;filter19
saving out filter35
saving out filter2;filter41
saving out filter45
saving out filter27
saving out filter5
saving out filter12
saving out filter7
saving out filter1;filter3
saving out filter26;filter39
saving out filter43;filter61
saving out filter36;filter52
saving out filter17
saving out filter44;filter50

/users/msharmin/code/mouse-atlas/matlas/pwms.py:449: FutureWarning:

read_table is deprecated, use read_csv instead, passing sep='\t'.

/users/msharmin/code/mouse-atlas/matlas/pwms.py:463: ClusterWarning:

scipy.cluster: The symmetric non-negative hollow observation matrix looks suspiciously like an uncondensed distance matrix

from matlas.reports import generate_merged_pwm_plots
from matlas.pwms import read_pwm_file, load_cisbp_maps

mergeddict = read_pwm_file("{}.hclust_pwms.tmp".format(file_prefix))
tab, tabledict = generate_merged_pwm_plots(mergeddict, motif_id_maps)

#display(tab)

Pattern Name	TF Name(s)	Influence
filter47		0.2141450319170016
filter41		0.026994974563086676
filter33	HCLUST-176_CBFB.UNK.0.A, HCLUST-127_FOXJ2.UNK.0.A, HCLUST-51_ZIM3.UNK.0.A, HCLUST-121_AR.UNK.0.A, HCLUST-161_GLI3.UNK.0.A	0.003863943019422465
filter56	HCLUST-185_EGR1.UNK.0.A, HCLUST-139_SREBF1.UNK.0.A, HCLUST-176_CBFB.UNK.0.A, HCLUST-8_TBX21.UNK.0.A, HCLUST-116_KLF8.UNK.0.A, HCLUST-161_GLI3.UNK.0.A, HCLUST-184_KLF12.UNK.0.A	0.002708672948704476
filter38	HCLUST-176_CBFB.UNK.0.A, HCLUST-51_ZIM3.UNK.0.A, HCLUST-121_AR.UNK.0.A, HCLUST-127_FOXJ2.UNK.0.A, HCLUST-108_FOXH1.UNK.0.A, HCLUST-14_HOXA13.UNK.0.A	0.0024906331847179687
filter8	HCLUST-16_RFX1.UNK.0.A	0.0016528585249671397
filter0	HCLUST-116_KLF8.UNK.0.A, HCLUST-119_AHR.UNK.0.A, HCLUST-102_NKX2-1.UNK.0.A, HCLUST-139_SREBF1.UNK.0.A	0.0014720499705049757
filter53	HCLUST-14_HOXA13.UNK.0.A, HCLUST-74_SRF.UNK.0.A, HCLUST-120_AIRE.UNK.0.A, HCLUST-127_FOXJ2.UNK.0.A, HCLUST-123_CDX2.UNK.0.A, HCLUST-76_TBP.UNK.0.A, HCLUST-121_AR.UNK.0.A, HCLUST-85_ZNF354A.UNK.0.A, HCLUST-13_CDX1.UNK.0.A	0.0009324956345416156
filter5	HCLUST-134_INSM1.UNK.0.A, HCLUST-185_EGR1.UNK.0.A, HCLUST-122_TFAP2B.UNK.0.A, HCLUST-116_KLF8.UNK.0.A, HCLUST-27_TAL1.UNK.0.A, HCLUST-63_ZNF320.UNK.0.A, HCLUST-149_CTCFL.UNK.0.A, HCLUST-40_ZNF816.UNK.0.A, HCLUST-183_E2F1.UNK.0.A, HCLUST-1_GATA1.UNK.0.A, HCLUST-68_ZNF341.UNK.0.A	0.0009066441587253029
filter43	HCLUST-34_ZNF586.UNK.0.A, HCLUST-182_NFATC1.UNK.0.A, HCLUST-107_FEZF1.UNK.0.A, HCLUST-127_FOXJ2.UNK.0.A, HCLUST-25_SOX17.UNK.0.A, HCLUST-121_AR.UNK.0.A, HCLUST-24_SOX5.UNK.0.A, HCLUST-79_ZFP82.UNK.0.A, HCLUST-177_MEF2A.UNK.0.A	0.000905770955833986
filter13	HCLUST-4_ATOH1.UNK.0.A	0.0006970092021308099
filter2	HCLUST-41_ZNF41.UNK.0.A, HCLUST-152_SOX10.UNK.0.A, HCLUST-107_FEZF1.UNK.0.A, HCLUST-159_EHF.UNK.0.A, HCLUST-25_SOX17.UNK.0.A, HCLUST-21_ZNF350.UNK.0.A, HCLUST-40_ZNF816.UNK.0.A, HCLUST-76_TBP.UNK.0.A, HCLUST-98_EBF1.UNK.0.A	0.00053401885547055
filter48	HCLUST-158_BCL11A.UNK.0.A, HCLUST-72_PAX6.UNK.0.A, HCLUST-88_ZFP28.UNK.0.A, HCLUST-165_LEF1.UNK.0.A, HCLUST-151_HSF1.UNK.0.A, HCLUST-29_ZNF547.UNK.0.A	0.00039986277388253343

Pattern Name	TF Name(s)	Influence
filter14		0.5303529192653758
filter29	HCLUST-118_MTF1.UNK.0.A, HCLUST-52_ZKSCAN1.UNK.0.A	0.04881478023585919
filter1	HCLUST-176_CBFB.UNK.0.A, HCLUST-31_ZNF449.UNK.0.A, HCLUST-102_NKX2-1.UNK.0.A, HCLUST-121_AR.UNK.0.A	0.0038786480474260663
filter5	HCLUST-176_CBFB.UNK.0.A, HCLUST-161_GLI3.UNK.0.A, HCLUST-185_EGR1.UNK.0.A, HCLUST-116_KLF8.UNK.0.A, HCLUST-51_ZIM3.UNK.0.A, HCLUST-139_SREBF1.UNK.0.A	0.00339283013944702
filter9	HCLUST-106_ATF6.UNK.0.A	0.002978532269916345
filter40	HCLUST-74_SRF.UNK.0.A, HCLUST-13_CDX1.UNK.0.A, HCLUST-123_CDX2.UNK.0.A, HCLUST-46_PRDM14.UNK.0.A, HCLUST-87_ZBTB48.UNK.0.A, HCLUST-14_HOXA13.UNK.0.A, HCLUST-177_MEF2A.UNK.0.A	0.0020161561346765744
filter58	HCLUST-47_REST.UNK.0.A, HCLUST-90_ZBTB6.UNK.0.A	0.0011998207688339757
filter63	HCLUST-153_CRX.UNK.0.A, HCLUST-21_ZNF350.UNK.0.A, HCLUST-182_NFATC1.UNK.0.A, HCLUST-85_ZNF354A.UNK.0.A, HCLUST-6_PPARA.UNK.0.A, HCLUST-25_SOX17.UNK.0.A	0.0009286532818935663
filter46	HCLUST-14_HOXA13.UNK.0.A, HCLUST-120_AIRE.UNK.0.A, HCLUST-93_NR2E3.UNK.0.A	0.0007539648678227104
filter60	HCLUST-106_ATF6.UNK.0.A	0.000728994850062196
filter39	HCLUST-31_ZNF449.UNK.0.A	0.0006705622938474953
filter25	HCLUST-40_ZNF816.UNK.0.A, HCLUST-185_EGR1.UNK.0.A, HCLUST-159_EHF.UNK.0.A, HCLUST-68_ZNF341.UNK.0.A, HCLUST-135_MBD2.UNK.0.A, HCLUST-39_ZNF770.UNK.0.A, HCLUST-184_KLF12.UNK.0.A	0.0006671001315200462
filter24	HCLUST-36_ZNF680.UNK.0.A, HCLUST-34_ZNF586.UNK.0.A, HCLUST-25_SOX17.UNK.0.A, HCLUST-42_ZNF85.UNK.0.A, HCLUST-182_NFATC1.UNK.0.A, HCLUST-24_SOX5.UNK.0.A, HCLUST-121_AR.UNK.0.A, HCLUST-127_FOXJ2.UNK.0.A, HCLUST-73_SMARCA1.UNK.0.A, HCLUST-107_FEZF1.UNK.0.A, HCLUST-177_MEF2A.UNK.0.A	0.0006218640672498633

Pattern Name	TF Name(s)	Influence
filter29	HCLUST-176_CBFB.UNK.0.A, HCLUST-121_AR.UNK.0.A, HCLUST-161_GLI3.UNK.0.A, HCLUST-116_KLF8.UNK.0.A	0.0015731521016156245
filter39	HCLUST-185_EGR1.UNK.0.A, HCLUST-176_CBFB.UNK.0.A, HCLUST-139_SREBF1.UNK.0.A, HCLUST-161_GLI3.UNK.0.A, HCLUST-51_ZIM3.UNK.0.A	0.0013639433753971403
filter18	HCLUST-161_GLI3.UNK.0.A, HCLUST-81_THAP1.UNK.0.A, HCLUST-27_TAL1.UNK.0.A	0.00022937509436457652
filter55	HCLUST-132_ALX1.UNK.0.A, HCLUST-42_ZNF85.UNK.0.A, HCLUST-85_ZNF354A.UNK.0.A	0.00020788976622628446

Pattern Name	TF Name(s)	Aitac	Influence
filter3	HCLUST-176_CBFB.UNK.0.A, HCLUST-161_GLI3.UNK.0.A, HCLUST-116_KLF8.UNK.0.A, HCLUST-127_FOXJ2.UNK.0.A, HCLUST-8_TBX21.UNK.0.A, HCLUST-185_EGR1.UNK.0.A		0.005059514071722837
filter9	HCLUST-151_HSF1.UNK.0.A		0.0005420086359834736

Pattern Name	TF Name(s)	Influence
filter36		0.16611183262263723
filter3	HCLUST-161_GLI3.UNK.0.A, HCLUST-176_CBFB.UNK.0.A, HCLUST-31_ZNF449.UNK.0.A, HCLUST-116_KLF8.UNK.0.A, HCLUST-58_ZNF250.UNK.0.A	0.004476138254140201
filter49	HCLUST-120_AIRE.UNK.0.A	0.0017054984991870974
filter47	HCLUST-176_CBFB.UNK.0.A, HCLUST-127_FOXJ2.UNK.0.A, HCLUST-51_ZIM3.UNK.0.A, HCLUST-121_AR.UNK.0.A, HCLUST-14_HOXA13.UNK.0.A	0.0009285687574980652
filter27		0.0007371537354956307
filter38		0.0006232905971403855
filter11	HCLUST-135_MBD2.UNK.0.A	0.0003010345693770689
filter26	HCLUST-117_LHX3.UNK.0.A	0.00022486129908857173
filter17		0.00019479307243913083
filter30	HCLUST-14_HOXA13.UNK.0.A, HCLUST-74_SRF.UNK.0.A, HCLUST-123_CDX2.UNK.0.A, HCLUST-182_NFATC1.UNK.0.A, HCLUST-34_ZNF586.UNK.0.A, HCLUST-121_AR.UNK.0.A, HCLUST-25_SOX17.UNK.0.A, HCLUST-148_ZNF146.UNK.0.A, HCLUST-120_AIRE.UNK.0.A	0.00018495083879069787
filter42	HCLUST-68_ZNF341.UNK.0.A, HCLUST-185_EGR1.UNK.0.A, HCLUST-40_ZNF816.UNK.0.A, HCLUST-134_INSM1.UNK.0.A, HCLUST-159_EHF.UNK.0.A, HCLUST-44_ZSCAN22.UNK.0.A, HCLUST-184_KLF12.UNK.0.A, HCLUST-39_ZNF770.UNK.0.A	0.00018264614041960616
filter50	HCLUST-122_TFAP2B.UNK.0.A, HCLUST-116_KLF8.UNK.0.A, HCLUST-134_INSM1.UNK.0.A, HCLUST-63_ZNF320.UNK.0.A	0.0001725042856237982

Pattern Name	TF Name(s)	Influence
filter49	HCLUST-106_ATF6.UNK.0.A, HCLUST-186_ARNTL.UNK.0.A, HCLUST-143_E2F2.UNK.0.A	0.04511756147483279
filter35		0.03418056875743683
filter32	HCLUST-174_CEBPA.UNK.0.A, HCLUST-105_ATF4.UNK.0.A, HCLUST-133_CEBPD.UNK.0.A	0.009460713455206328
filter11	HCLUST-121_AR.UNK.0.A, HCLUST-176_CBFB.UNK.0.A, HCLUST-31_ZNF449.UNK.0.A, HCLUST-51_ZIM3.UNK.0.A, HCLUST-127_FOXJ2.UNK.0.A, HCLUST-55_ZNF136.UNK.0.A, HCLUST-58_ZNF250.UNK.0.A	0.0031363830354860503
filter34	HCLUST-74_SRF.UNK.0.A	0.0021727132975229786
filter58	HCLUST-175_ATF1.UNK.0.A, HCLUST-142_TAF1.UNK.0.A	0.0021525233683859675
filter46	HCLUST-106_ATF6.UNK.0.A	0.0018044556986303638
filter24	HCLUST-185_EGR1.UNK.0.A, HCLUST-139_SREBF1.UNK.0.A, HCLUST-161_GLI3.UNK.0.A, HCLUST-8_TBX21.UNK.0.A, HCLUST-68_ZNF341.UNK.0.A, HCLUST-1_GATA1.UNK.0.A, HCLUST-116_KLF8.UNK.0.A, HCLUST-176_CBFB.UNK.0.A, HCLUST-31_ZNF449.UNK.0.A	0.0009971555231064425
filter3	HCLUST-8_TBX21.UNK.0.A, HCLUST-161_GLI3.UNK.0.A, HCLUST-27_TAL1.UNK.0.A, HCLUST-116_KLF8.UNK.0.A, HCLUST-108_FOXH1.UNK.0.A	0.0008638738471203829
filter1		0.0007134999859777418
filter57	HCLUST-8_TBX21.UNK.0.A, HCLUST-147_ESR1.UNK.0.A	0.0006584498306860246
filter38	HCLUST-149_CTCFL.UNK.0.A, HCLUST-122_TFAP2B.UNK.0.A, HCLUST-63_ZNF320.UNK.0.A	0.0004119439564808684
filter26	HCLUST-49_ZFX.UNK.0.A, HCLUST-122_TFAP2B.UNK.0.A, HCLUST-170_TP53.UNK.0.A, HCLUST-184_KLF12.UNK.0.A, HCLUST-87_ZBTB48.UNK.0.A, HCLUST-135_MBD2.UNK.0.A	0.00034410567512111004
filter12	HCLUST-36_ZNF680.UNK.0.A, HCLUST-121_AR.UNK.0.A	0.00029056874058164104

Pattern Name	TF Name(s)	Influence
filter0	HCLUST-176_CBFB.UNK.0.A, HCLUST-185_EGR1.UNK.0.A, HCLUST-51_ZIM3.UNK.0.A, HCLUST-161_GLI3.UNK.0.A	0.0042318836019674625
filter7		0.0037567814326760693
filter57		0.001043428027546221
filter3		0.0006908377866463216
filter50	HCLUST-182_NFATC1.UNK.0.A, HCLUST-82_TWIST1.UNK.0.A, HCLUST-177_MEF2A.UNK.0.A, HCLUST-34_ZNF586.UNK.0.A	0.0006470899264955723
filter43	HCLUST-46_PRDM14.UNK.0.A, HCLUST-44_ZSCAN22.UNK.0.A, HCLUST-142_TAF1.UNK.0.A, HCLUST-49_ZFX.UNK.0.A, HCLUST-162_ASCL1.UNK.0.A, HCLUST-87_ZBTB48.UNK.0.A	0.0004818343618733898

Pattern Name	TF Name(s)	Influence
filter31	HCLUST-174_CEBPA.UNK.0.A	0.008645216464248758
filter4	HCLUST-84_ZNF324.UNK.0.A, HCLUST-127_FOXJ2.UNK.0.A, HCLUST-176_CBFB.UNK.0.A, HCLUST-24_SOX5.UNK.0.A, HCLUST-116_KLF8.UNK.0.A	0.0028602785453494526
filter26	HCLUST-185_EGR1.UNK.0.A, HCLUST-68_ZNF341.UNK.0.A, HCLUST-161_GLI3.UNK.0.A, HCLUST-184_KLF12.UNK.0.A, HCLUST-27_TAL1.UNK.0.A, HCLUST-139_SREBF1.UNK.0.A, HCLUST-1_GATA1.UNK.0.A, HCLUST-116_KLF8.UNK.0.A	0.0014315341903029372
filter63	HCLUST-127_FOXJ2.UNK.0.A, HCLUST-121_AR.UNK.0.A, HCLUST-176_CBFB.UNK.0.A, HCLUST-181_HOXA9.UNK.0.A, HCLUST-14_HOXA13.UNK.0.A, HCLUST-51_ZIM3.UNK.0.A	0.0009373525827069896
filter62	HCLUST-60_ZNF264.UNK.0.A, HCLUST-153_CRX.UNK.0.A, HCLUST-108_FOXH1.UNK.0.A, HCLUST-145_RELA.UNK.0.A	0.0007538492697410265
filter50	HCLUST-158_BCL11A.UNK.0.A, HCLUST-146_NR1I2.UNK.0.A, HCLUST-135_MBD2.UNK.0.A, HCLUST-57_ZNF214.UNK.0.A, HCLUST-87_ZBTB48.UNK.0.A, HCLUST-129_ELF1.UNK.0.A, HCLUST-143_E2F2.UNK.0.A	0.000520312030659572
filter18	HCLUST-39_ZNF770.UNK.0.A, HCLUST-38_ZNF768.UNK.0.A, HCLUST-184_KLF12.UNK.0.A, HCLUST-1_GATA1.UNK.0.A, HCLUST-139_SREBF1.UNK.0.A	0.0003608926165120295

Pattern Name	TF Name(s)	Influence
filter12	HCLUST-176_CBFB.UNK.0.A, HCLUST-116_KLF8.UNK.0.A, HCLUST-161_GLI3.UNK.0.A, HCLUST-7_T.UNK.0.A, HCLUST-25_SOX17.UNK.0.A	0.002713735648911871
filter55	HCLUST-183_E2F1.UNK.0.A	0.0013512861455838097
filter7	HCLUST-14_HOXA13.UNK.0.A, HCLUST-76_TBP.UNK.0.A	0.000713767504577353
filter40	HCLUST-185_EGR1.UNK.0.A, HCLUST-161_GLI3.UNK.0.A, HCLUST-139_SREBF1.UNK.0.A, HCLUST-116_KLF8.UNK.0.A, HCLUST-68_ZNF341.UNK.0.A, HCLUST-176_CBFB.UNK.0.A, HCLUST-184_KLF12.UNK.0.A, HCLUST-27_TAL1.UNK.0.A, HCLUST-1_GATA1.UNK.0.A	0.0005779078660504286
filter23	HCLUST-51_ZIM3.UNK.0.A, HCLUST-13_CDX1.UNK.0.A, HCLUST-176_CBFB.UNK.0.A	0.0004649631931000118
filter8	HCLUST-81_THAP1.UNK.0.A, HCLUST-39_ZNF770.UNK.0.A, HCLUST-184_KLF12.UNK.0.A, HCLUST-170_TP53.UNK.0.A, HCLUST-59_ZNF257.UNK.0.A	0.0003463140917611424
filter51	HCLUST-51_ZIM3.UNK.0.A, HCLUST-53_ZNF121.UNK.0.A, HCLUST-3_ZBTB18.UNK.0.A	0.00029359926428542223

Pattern Name	TF Name(s)	Influence
filter25	HCLUST-118_MTF1.UNK.0.A, HCLUST-52_ZKSCAN1.UNK.0.A	0.531286771169387
filter26	HCLUST-31_ZNF449.UNK.0.A, HCLUST-161_GLI3.UNK.0.A	0.009207882687234754
filter2	HCLUST-176_CBFB.UNK.0.A, HCLUST-121_AR.UNK.0.A, HCLUST-51_ZIM3.UNK.0.A, HCLUST-127_FOXJ2.UNK.0.A, HCLUST-58_ZNF250.UNK.0.A	0.0033471098290529543
filter51	HCLUST-64_ZNF322.UNK.0.A	0.002090001760651158
filter11		0.0019314666263402188
filter63	HCLUST-92_NR2C2.UNK.0.A, HCLUST-183_E2F1.UNK.0.A, HCLUST-6_PPARA.UNK.0.A, HCLUST-38_ZNF768.UNK.0.A	0.0010632302137559041
filter7		0.0005846856482895691
filter54	HCLUST-176_CBFB.UNK.0.A, HCLUST-41_ZNF41.UNK.0.A, HCLUST-108_FOXH1.UNK.0.A, HCLUST-128_HNF4A.UNK.0.A, HCLUST-121_AR.UNK.0.A	0.0005678030107418309
filter42	HCLUST-177_MEF2A.UNK.0.A, HCLUST-121_AR.UNK.0.A, HCLUST-166_NANOG.UNK.0.A	0.0004439276465761844
filter47	HCLUST-127_FOXJ2.UNK.0.A, HCLUST-121_AR.UNK.0.A, HCLUST-176_CBFB.UNK.0.A, HCLUST-14_HOXA13.UNK.0.A, HCLUST-51_ZIM3.UNK.0.A, HCLUST-88_ZFP28.UNK.0.A, HCLUST-120_AIRE.UNK.0.A, HCLUST-24_SOX5.UNK.0.A, HCLUST-126_FOXA1.UNK.0.A	0.00039214048795743436
filter18	HCLUST-34_ZNF586.UNK.0.A, HCLUST-177_MEF2A.UNK.0.A, HCLUST-182_NFATC1.UNK.0.A, HCLUST-107_FEZF1.UNK.0.A, HCLUST-40_ZNF816.UNK.0.A, HCLUST-121_AR.UNK.0.A, HCLUST-76_TBP.UNK.0.A	0.00037618150853643834

TF Name	PPM	Fold count
HCLUST-176_CBFB.UNK.0.A		10
HCLUST-161_GLI3.UNK.0.A		10
HCLUST-51_ZIM3.UNK.0.A		9
HCLUST-185_EGR1.UNK.0.A		9
HCLUST-116_KLF8.UNK.0.A		8
HCLUST-127_FOXJ2.UNK.0.A		7
HCLUST-121_AR.UNK.0.A		7
HCLUST-139_SREBF1.UNK.0.A		6
HCLUST-184_KLF12.UNK.0.A		6
HCLUST-14_HOXA13.UNK.0.A		6
HCLUST-68_ZNF341.UNK.0.A		6
HCLUST-27_TAL1.UNK.0.A		5
HCLUST-34_ZNF586.UNK.0.A		5
HCLUST-182_NFATC1.UNK.0.A		5