kla_only_no_p65 = sum((gr_group['gr_dex_tag_count'] <= min_tags) &
(gr_group['gr_kla_dex_tag_count'] > min_tags) &
(gr_group['p65_kla_tag_count'] + gr_group['p65_kla_dex_tag_count'] <= min_tags)
)
counts = [tethered, direct_comp_gr, indirect_comp_gr,
direct_comp_p65, cobound,
direct_novel, indirect_novel,
in_dex_no_p65, kla_only_no_p65]
else: counts = [0]*9
stats = stats + counts
all_stats.append(dict(zip(labels, stats))), index.append(group.name)
grouped.apply(count_enhancers)
# There must be a better way to do this group-apply, but I can't make it turn back into a DF...
enhancer_counts = DataFrame(all_stats, index=index)
spaced_labels = ['\n'.join(map(' '.join,
[l.split()[i:i+2] for i in xrange(0,len(l.split()),2)] ))
for l in labels]
erna_title = 'Enhancers per Gene by enhancer subtype {0}'.format(name)
ax = yzer.boxplot([enhancer_counts[col] for col in labels], spaced_labels,
title=erna_title,
xlabel='Subset',
ylabel='Count',
show_outliers=False, show_plot=False, wide=True
)
yzer.ylim(ax, -1, 2)
pyplot.setp(ax.get_xticklabels(), fontsize=10)
yzer.save_plot(yzer.get_filename(img_dirpath, erna_title + '.png'))
yzer.show_plot()
'{}_{}'.format(peptide, ab))
filename = yzer.get_filename(
pep_dirpath, '{}_{}_enhancers_batf.txt'.format(peptide, ab))
data = yzer.import_file(filename)
data = data.fillna(0)
subset = data[data['no_pep_tag_count'] == 0]
#subset = data[data['tag_count(2)'] == 0]
#subset = subset[subset['tag_count(3)'] == 0]
all_tag_counts.append(data['tag_count'])
de_novo_tag_counts.append(subset['tag_count'])
# Plot as boxplot
ax = yzer.boxplot(all_tag_counts,
conditions,
title='Tag Counts in {} Enhancers'.format(
ab.title()),
xlabel='Condition',
ylabel='Normalized tag count',
save_dir=savepath,
show_plot=True)
ax = yzer.boxplot(de_novo_tag_counts,
conditions,
title='Tag Counts in de novo {} Enhancers'.format(
ab.title()),
xlabel='Condition',
ylabel='Normalized tag count',
save_dir=savepath,
show_plot=True)
fold >= wt_data['tag_count'])
& (wt_data['tag_count'] *
fold >= wt_data['foxo1_ko_naive_atac_tag_count'])]
ko_only = ko_data[ko_data['naive_atac_tag_count'] < min_thresh]
save_path = yzer.get_and_create_path(dirpath, 'Figures',
'Foxo1_group_overlaps')
groups = [wt_only, both, ko_only]
labels = ['WT only', 'WT and KO', 'Foxo1 KO only']
if True:
yzer.boxplot([gp['naive_foxo1_tag_count'] for gp in groups],
labels,
title='Foxo1 tags in ATAC-seq regions by group',
ylabel='Foxo1 peak tag count',
save_dir=save_path,
show_plot=False)
yzer.boxplot([gp['lcmv_d12_foxo1_tag_count'] for gp in groups],
labels,
title='LCMV d12 Foxo1 tags in ATAC-seq regions by group',
ylabel='Foxo1 peak tag count',
save_dir=save_path,
show_plot=False)
yzer.boxplot([gp['naive_h3k4me2_tag_count'] for gp in groups],
labels,
title='H3K4me2 tags in ATAC-seq regions by group',
ylabel='H3K4me2 region count',
save_dir=save_path,
show_plot=False)
yzer.boxplot(
groups = [data[none], data[kla_gt], data[nc], data[kla_dex_gt]]
# We want to randomly sample to get equi-sized groups
desired = len(nearby)
for i, g in enumerate(groups):
rows = random.sample(g.index, desired)
groups[i] = g.ix[rows]
to_plot = [g[colname] for g in (groups + [nearby])]
title = 'LFC in KLA + Dex over KLA by change in p65:' \
+ '\nRefSeq, randomly sampled to {0} transcripts'.format(desired)
if pausing:
title = 'Pausing Ratio Ratio by change in p65:' \
+ '\nRefSeq, randomly sampled to {0} transcripts'.format(desired)
ax = yzer.boxplot(to_plot,
names,
title=title,
xlabel='Transcript Status',
ylabel=(pausing and 'PausingRatio(KLA+Dex)/PausingRatio(KLA)')\
or 'log2(KLA+Dex GRO-seq/KLA GRO-seq)',
show_outliers=False, show_plot=False)
yzer.save_plot(
yzer.get_filename(
img_dirpath,
'{2}_with_nearby_unique_{0}x_{3}_sampled_{1}.png'.format(
ratio, random.randint(0, 9999), colname, change_type)))
yzer.show_plot()
kla_4h_with_me2_counts = data['interacting_in_kla_4h_with_me2'][
'count'].values.tolist() + [0] * zero_intxns_in_kla_4h_with_me2
labels = [
'Less than 1/4\navg H3K4me2 in notx,\ninteractions in notx',
'At least avg H3K4me2 in notx\ninteractions in notx',
'Less than 1/4\navg H3K4me2 in notx,\ninteractions in KLA 30m',
'At least avg H3K4me2 in notx\ninteractions in KLA 30m',
'Less than 1/4\navg H3K4me2 in notx,\ninteractions in KLA 4h',
'At least avg H3K4me2 in notx\ninteractions in KLA 4h',
]
vals = [
notx_with_less_me2_counts, notx_with_me2_counts,
kla_30m_with_less_me2_counts, kla_30m_with_me2_counts,
kla_4h_with_less_me2_counts, kla_4h_with_me2_counts
]
labels = [
l + '\n(count: {0})'.format(len(v)) for l, v in zip(labels, vals)
]
title = 'Number of interactions with "enhancers" by H3K4me2 state in notx'
ax = yzer.boxplot(vals,
labels,
title=title,
xlabel='Enhancer subset',
ylabel='Number of interactions with other transcripts',
show_outliers=False,
show_plot=True,
wide=True,
save_dir=img_dirpath)
for group in (me2_only, ctcf_only, k27_only, ctcf_me2, k27_me2, nothing):
print len(group), len(group)/len(data)
if True:
ax = yzer.boxplot([data['rpkm'], me2_tf['rpkm'], ctcf_tf['rpkm'],
k27_tf['rpkm'], ctcf_me2_tf['rpkm'], k27_me2_tf['rpkm'], tf['rpkm'],
me2_only['rpkm'], ctcf_only['rpkm'],
k27_only['rpkm'], ctcf_me2['rpkm'], k27_me2['rpkm'], nothing['rpkm']],
bar_names=['All Potential\nEnhancers',
'me2 + TF', 'CTCF + TF',
'K27 + TF',
'CTCF + me2\n+ TF', 'K27 + me2\n+ TF', 'TF',
'me2 only', 'CTCF only',
'K27 only',
'CTCF + me2', 'K27 + me2',
'No peaks',],
title='GRO-seq RPKM at non-genic H3K4me2 regions',
xlabel='', ylabel='Tags per 1000bp in GRO-seq transcript overlapping H3K4me2 peak',
show_outliers=False, show_plot=False)
yzer.save_plot(os.path.join(dirpath, 'groseq_rpkm_at_h3k4me2_peaks.png'))
yzer.show_plot()
'''
-- With H3K27me3
select distinct on (e.id) e.*, e.id as me2, reg.id as refseq, p1.id as pu_1, p2.id as cebpa,
grapher.show_plot()
if True:
# Boxplots: avg PU.1 in Bl6 for whole set; avg PU.1 in BALB for whole set;
# avg PU.1 for NOD in whole set; avg PU.1 in NOD set with Bl6; avg PU.1 in NOD set with BALB
ax = grapher.boxplot(
[
data['wt_pu_1_tag_count'], data['balb_pu_1_tag_count_norm'],
data['nod_pu_1_tag_count_norm'],
nod_with_bl6['nod_pu_1_tag_count_norm'],
nod_with_balb['nod_pu_1_tag_count_norm']
],
bar_names=[
'C57Bl6 Peaks',
'BALBc Peaks',
'NOD Peaks',
'NOD Peaks\nwhere\nNOD == C57Bl6',
'NOD Peaks\nwhere\nNOD == BALBc',
],
title=
'PU.1 peak tags where BALBc has a SNP that ruins its PU.1 Motif',
xlabel='',
ylabel='Tags per PU.1 peak',
show_outliers=False,
show_plot=False)
grapher.save_plot(
os.path.join(dirpath,
'peak_boxplots_no_balb_motif_filter_low_peaks.png'))
grapher.show_plot()
print 'p-val that BALBc is different than C57Bl6: %g' % ttest_ind(
save_path = yzer.get_and_create_path(
dirpath, 'Figures', 'me2_atac_overlaps')
yzer.piechart([len(atac_only), len(atac_me2)],
['ATAC only', 'ATAC with H3K4me2'],
title='ATAC-seq region overlaps',
save_dir=save_path)
yzer.piechart([len(me2_only), len(me2_atac)],
['H3K4me2 only', 'H3K4me2 with ATAC'],
title='H3K4me2 overlaps',
save_dir=save_path)
yzer.boxplot([atac_only['tag_count'], atac_me2['tag_count']],
['ATAC only', 'ATAC with H3K4me2'],
title='ATAC-seq tag counts by H3K4me2 overlap',
xlabel='Group', ylabel='Peak tag count',
save_dir=save_path)
yzer.boxplot([me2_only['tag_count'], me2_atac['tag_count']],
['H3K4me2 only', 'H3K4me2 with ATAC'],
title='H3K4me2 tag counts by ATAC-seq overlap',
xlabel='Group', ylabel='Peak tag count',
save_dir=save_path)
yzer.histogram(atac_only['tag_count'].tolist(), bins=20,
title='ATAC-seq-only peak tag count distribution',
xlabel='Tag count in peak', ylabel='Number of peaks',
save_dir=save_path)
yzer.histogram(me2_only['tag_count'].tolist(), bins=20,
title='H3K4me2-only peak tag count distribution',
xlabel='Tag count in peak', ylabel='Number of peaks',
save_dir=save_path)
dirpath,
'balbc_vs_nod_pu_1_peak_tag_counts_bl6_gt_balb_unique.png'))
grapher.show_plot()
if True:
# Boxplots
ax = grapher.boxplot(
[
data['wt_tag_count'], data['balb_tag_count_norm'],
data['nod_tag_count_norm'], nod_with_bl6['nod_tag_count_norm'],
nod_with_balb['nod_tag_count_norm']
],
bar_names=[
'C57Bl6 Tags',
'BALBc Tags',
'NOD Tags',
'NOD Tags\nwhere\nNOD == C57Bl6',
'NOD Tags\nwhere\nNOD == BALBc',
],
title='GRO-seq tags where BALBc has a SNP and half H3K4me2',
xlabel='',
ylabel='Tags in transcript at H3K4me2 peak',
show_outliers=False,
show_plot=False)
grapher.save_plot(
os.path.join(dirpath, 'peak_boxplots_all_h3k4me2_collapsed.png'))
grapher.show_plot()
if True:
print 'p-val that BALBc is different than C57Bl6: %g' % ttest_ind(
data['wt_tag_count'], data['balb_tag_count_norm'])[1]
