def plot_consensus_boundary_properties(consensus_bounds):
figsize = get_default_figsize()
fig, axes = plt.subplots(2,
1,
sharex=True,
tight_layout=True,
figsize=(figsize[0] * 1.5, figsize[1] * 2))
min_n, max_n = consensus_bounds.n_cell_types.min(
), consensus_bounds.n_cell_types.max()
sns.countplot(consensus_bounds.n_cell_types,
ax=axes[0],
order=range(min_n, max_n + 1),
palette="Reds")
axes[0].set_xlabel("")
axes[0].set_ylabel("Boundaries")
axes[0].grid(axis='y')
axes[0].ticklabel_format(axis='y', scilimits=(0, 0), useMathText=True)
sns.boxplot(data=consensus_bounds,
x='n_cell_types',
y="length",
order=range(min_n, max_n + 1),
ax=axes[1],
palette="Reds")
axes[1].set_xlabel("Conservation score (s)")
axes[1].set_ylabel("Boundary size")
axes[1].grid(axis='y')
axes[1].ticklabel_format(axis='y', scilimits=(0, 0), useMathText=True)
fig.savefig(figures_path / "consensus_boundaries_stats.pdf",
bbox_inches='tight',
transparent=True)
def plot_cons_bounds_vs_grbs(grb_vs_windows, length_vs_windows, grbs):
X = grb_vs_windows.reindex(index=grbs.sort_values(
'length', ascending=False).grb_uid.tolist()).fillna(0).values
min_cons = 2
X[X < min_cons] = 0
X[X >= min_cons] = 1
L = length_vs_windows.reindex(index=grbs.sort_values(
'length', ascending=False).grb_uid.tolist()).fillna(0).values
L[L > 0] = 1
L = L[L.sum(1) != 0]
figsize = get_default_figsize()
fig, axes = plt.subplots(1,
2,
sharey=True,
figsize=(figsize[0] * 2, figsize[1]))
axes[0].imshow(X, aspect="auto", interpolation='bilinear', cmap='hot')
axes[0].set_yticks([])
axes[0].set_xticks([0, int(X.shape[1] / 2), X.shape[1] - 1])
axes[0].set_xticklabels(['-2.5Mb', '0', "+2.5Mb"])
axes[0].set_xlabel("Distance from GRB center")
axes[0].set_title("Conserved boundaries ($s \geq {}$)".format(min_cons),
fontsize='large')
axes[1].imshow(L, aspect="auto", interpolation='bilinear', cmap='Blues')
axes[1].set_yticks([])
axes[1].set_xticks([0, int(L.shape[1] / 2), L.shape[1] - 1])
axes[1].set_xticklabels(['-2.5Mb', '0', "+2.5Mb"])
axes[1].set_xlabel("Distance from GRB center")
axes[1].set_title("GRB length", fontsize='large')
fig.savefig(figures_path / "cons_bounds_vs_GRBs.pdf",
bbox_inches='tight',
transparent=True)
def plot_n_ctcf_sites_in_boundaries(b):
def __to_bin(x):
if x in [0, 1, 2]:
return str(x)
elif x in [3, 4]:
return "3-4"
elif x in range(5, 9):
return "5-8"
elif x in range(9, 12):
return "9-11"
bounds_with_ctcf_count = b.copy()
bounds_with_ctcf_count['cat'] = bounds_with_ctcf_count.n_ctcfs.map(
__to_bin)
bounds_with_ctcf_count = pd.concat(
(bounds_with_ctcf_count, pd.get_dummies(bounds_with_ctcf_count.cat)),
axis=1)
groups = bounds_with_ctcf_count.groupby('n_cell_types')[[
'0', '1', '2', '3-4', '5-8', '9-11'
]].sum()
cmap = plt.cm.get_cmap('Blues', 8)
newcolors = cmap(np.linspace(0, 1, 8))
blue = np.array([1, 1, 1, 1])
newcolors[0, :] = blue
newcmp = ListedColormap(newcolors)
figsize = get_default_figsize()
fig, axes = plt.subplots(1, 7, figsize=(figsize[0] * 7, figsize[1] * 2))
for i, g in enumerate(range(1, 8)):
xg = groups.loc[[g], ['1', '2', '3-4', '5-8', '9-11']]
xg.loc['0', '0'] = groups.loc[g, '0']
xg = xg.fillna(0)
xg.loc[g, 'not_intersect'] = 0
xg = xg.loc[['0', g], ['0', '1', '2', '3-4', '5-8', '9-11']]
xg.plot.bar(stacked=True, cmap=newcmp, ax=axes[i], edgecolor='black')
axes[i].legend().set_visible(False)
axes[i].set_xticklabels([])
axes[i].grid(axis="y")
axes[i].tick_params(axis='y')
axes[i].set_xlabel("")
axes[i].set_title("$s = {}$".format(g),
fontweight="bold",
fontsize='xx-large')
axes[0].set_ylabel("Boundaries", fontsize='xx-large')
handles, labels = axes[i].get_legend_handles_labels()
fig.legend(handles,
labels,
loc='lower center',
ncol=8,
bbox_to_anchor=(0.43, 1.05),
title='CTCF binding sites')
fig.text(x=0.53,
y=-0.01,
horizontalalignment='center',
fontsize='xx-large',
s="Boundary conservation score ($s$)")
fig.savefig(figures_path /
"n_ctcf_sites_in_boundaries_by_conservation.pdf",
bbox_inches='tight',
transparent=True)
def plot_conservation_CTCF_sites(n_overlaps):
figsize = get_default_figsize()
fig = plt.figure(figsize=(figsize[0]*2.5, figsize[1]*0.5))
plt.hist(n_overlaps, bins=np.arange(max(n_overlaps) + 2), rwidth=0.9)
plt.grid(axis="y")
plt.xlabel("# of samples in which the CTCF site is conserved")
plt.ylabel("# CTCF sites")
plt.xticks(np.arange(max(n_overlaps) + 1) + 0.5, labels=np.arange(max(n_overlaps) + 1))
fig.savefig(figures_path / "CTCF_sites_number_overlaps_with_chipseqs.pdf", bbox_inches='tight', transparent=True)
def plot_triangle_enrichment_horizontal(pos_triangles_vs_ctcfs,
pos_triangles,
neg_triangles_vs_ctcfs,
neg_triangles,
path=figures_path /
"triangle_enrichment.pdf",
cmap='Greens'):
figsize = get_default_figsize()
fig, axes = plt.subplots(1, 2, figsize=(figsize[0] * 2, figsize[1] * 1))
X_pos = pos_triangles_vs_ctcfs.reindex(index = pos_triangles\
.sort_values('length', ascending=False)\
.triangle_uid.tolist()).dropna().values
X_pos[X_pos < 0] = -1
X_pos[X_pos > 0] = 1
axes[0].matshow(X_pos, aspect='auto', cmap=cmap, interpolation='bilinear')
axes[0].set_yticklabels([])
axes[0].set_xticks([])
axes[0].set_xticks([
0, X_pos.shape[1] // 4, X_pos.shape[1] // 2, X_pos.shape[1] * 3 // 4,
X_pos.shape[1]
])
axes[0].set_xticklabels(['-1 Mb', '-500 Kb', '0', '+500 Kb', '+1 Mb'])
axes[0].xaxis.set_ticks_position('bottom')
axes[0].set_xlabel("Distance from\nnegative inversion point")
axes[0].set_yticks([])
axes[0].set_ylabel(
"GM12878 TADs ordered by\ntheir Positive DI region length")
X_neg = neg_triangles_vs_ctcfs.reindex(index = neg_triangles\
.sort_values('length', ascending=False)\
.triangle_uid.tolist())\
.dropna().values
X_neg[X_neg < 0] = -1
X_neg[X_neg > 0] = 1
axes[1].matshow(X_neg, aspect='auto', cmap=cmap, interpolation='bilinear')
axes[1].set_yticklabels([])
axes[1].set_yticks([])
axes[1].set_xticks([
0, X_pos.shape[1] // 4, X_pos.shape[1] // 2, X_pos.shape[1] * 3 // 4,
X_pos.shape[1]
])
axes[1].set_xticklabels(['-1 Mb', '-500 Kb', '0', '+500 Kb', '+1 Mb'])
axes[1].set_xlabel("Distance from\nnegative inversion point")
axes[1].xaxis.set_ticks_position('bottom')
axes[1].set_yticks([])
axes[1].set_ylabel(
"GM12878 TADs ordered by\ntheir Negative DI region length",
rotation=270,
labelpad=27)
axes[1].yaxis.set_label_position("right")
fig.savefig(path, bbox_inches='tight', transparent=True)
def plot_aggregations_by_tad(aggregations_by_tad,
tad_start_window,
tad_end_window,
path=figures_path / "aggregations_by_tad.pdf"):
figsize = get_default_figsize()
fig, axes = plt.subplots(3,
1,
sharex=True,
figsize=(figsize[0] * 1.5, figsize[1] * 3))
axes[0].plot(aggregations_by_tad['ctcf_id'].mean(0),
label='both',
color=ctcf_colors['all'])
axes[0].set_ylabel("Avg. CTCF sites\nper % of TAD")
axes[0].axvline(tad_start_window, color='black', linestyle=':')
axes[0].axvline(tad_end_window, color='black', linestyle=':')
axes[0].grid()
axes[0].legend(loc='upper center')
axes[1].plot(aggregations_by_tad['forward'].mean(0),
label='Forward',
color=ctcf_colors['forward'])
axes[1].plot(aggregations_by_tad['reverse'].mean(0),
label='Reverse',
color=ctcf_colors['reverse'])
axes[1].set_ylabel("Avg. CTCF sites\nper % of TAD")
axes[1].axvline(tad_start_window, color='black', linestyle=':')
axes[1].axvline(tad_end_window, color='black', linestyle=':')
axes[1].grid()
axes[1].legend(loc='upper center')
axes[2].plot(aggregations_by_tad['S'].mean(0),
label='Same',
color=ctcf_colors['S'])
axes[2].plot(aggregations_by_tad['C'].mean(0),
label='Convergent',
color=ctcf_colors['C'])
axes[2].plot(aggregations_by_tad['D'].mean(0),
label='Divergent',
color=ctcf_colors['D'])
axes[2].plot(aggregations_by_tad['CD'].mean(0),
label='Convergent-Divergent',
color=ctcf_colors['CD'])
axes[2].set_ylabel("Avg. CTCF sites\nper % of TAD")
axes[2].grid()
axes[2].legend(loc='upper center')
axes[2].axvline(tad_start_window, color='black', linestyle=':')
axes[2].axvline(tad_end_window, color='black', linestyle=':')
axes[2].set_xticks([
tad_start_window,
int(tad_start_window + tad_end_window) / 2, tad_end_window
])
axes[2].set_xticklabels(
['TAD start\n(0%)', 'TAD center\n(50%)', 'TAD end\n(100%)'])
axes[2].set_xlabel("Position on TAD")
fig.savefig(path, bbox_inches='tight', transparent=True)
def plot_aggregations_by_boundary(aggregations_by_bound_tot, extended,
window_size):
figsize = get_default_figsize()
fig, axes = plt.subplots(3,
1,
sharex=True,
figsize=(figsize[0] * 1.5, figsize[1] * 3))
axes[0].plot(aggregations_by_bound_tot['ctcf_id'].mean(0),
label='both',
color=ctcf_colors['all'])
axes[0].set_ylabel("Avg. CTCF sites per 5kb")
axes[0].axvline(extended / window_size, color='black')
axes[0].grid()
axes[0].legend(loc='upper left')
axes[1].plot(aggregations_by_bound_tot['forward'].mean(0),
label='Forward',
color=ctcf_colors['forward'])
axes[1].plot(aggregations_by_bound_tot['reverse'].mean(0),
label='Reverse',
color=ctcf_colors['reverse'])
axes[1].set_ylabel("Avg. CTCF sites per 5kb")
axes[1].axvline(extended / window_size, color='black')
axes[1].grid()
axes[1].legend(loc='upper left')
axes[2].plot(aggregations_by_bound_tot['S'].mean(0),
label='Same',
color=ctcf_colors['S'])
axes[2].plot(aggregations_by_bound_tot['C'].mean(0),
label='Convergent',
color=ctcf_colors['C'])
axes[2].plot(aggregations_by_bound_tot['D'].mean(0),
label='Divergent',
color=ctcf_colors['D'])
axes[2].plot(aggregations_by_bound_tot['CD'].mean(0),
label='Convergent-\nDivergent',
color=ctcf_colors['CD'])
axes[2].set_ylabel("Avg. CTCF sites per 5kb")
axes[2].axvline(extended / window_size, color='black')
axes[2].grid()
axes[2].legend(loc='upper left')
axes[2].set_xticks([0, extended / window_size, extended * 2 / window_size])
axes[2].set_xticklabels(['-250kb', '0', '+250kb'])
axes[2].set_xlabel("Distance from boundary center")
fig.savefig(figures_path / "aggregations_by_boundary.pdf",
bbox_inches='tight',
transparent=True)
def plot_triangle_enrichment_halves(pos_triangles_vs_ctcfs,
pos_triangles,
neg_triangles_vs_ctcfs,
neg_triangles,
path=figures_path /
"triangle_enrichment.pdf",
cmap='Greens'):
figsize = get_default_figsize()
fig, axes = plt.subplots(1, 2, figsize=(figsize[0] * 2, figsize[1] * 1))
X_pos = pos_triangles_vs_ctcfs.reindex(index = pos_triangles\
.sort_values('length', ascending=False)\
.triangle_uid.tolist()).dropna().values
X_pos[X_pos < 0] = -1
X_pos[X_pos > 0] = 1
half = X_pos.shape[1] // 2
X_pos = X_pos[:, :half - 1]
axes[0].matshow(X_pos, aspect='auto', cmap=cmap, interpolation='bilinear')
axes[0].set_yticklabels([])
axes[0].set_xticks([])
axes[0].set_xticks([0, X_pos.shape[1] // 2, X_pos.shape[1] - 1])
axes[0].set_xticklabels(['-1Mb', '-500Kb', '0'])
axes[0].xaxis.set_ticks_position('bottom')
axes[0].set_xlabel("Distance from\nnegative inversion point")
axes[0].set_title("Positive DI regions")
X_neg = neg_triangles_vs_ctcfs.reindex(index = neg_triangles\
.sort_values('length', ascending=False)\
.triangle_uid.tolist())\
.dropna().values
X_neg[X_neg < 0] = -1
X_neg[X_neg > 0] = 1
half = X_neg.shape[1] // 2
X_neg = X_neg[:, half - 2:]
axes[1].matshow(X_neg, aspect='auto', cmap=cmap, interpolation='bilinear')
axes[1].set_yticklabels([])
axes[1].set_yticks([])
axes[1].set_xticks([0, X_neg.shape[1] // 2, X_neg.shape[1] - 1])
axes[1].set_xticklabels(['0', '+500Kb', '+1Mb'])
axes[1].set_xlabel("Distance from\nnegative inversion point")
axes[1].set_title("Negative DI regions")
axes[1].xaxis.set_ticks_position('bottom')
fig.savefig(path, bbox_inches='tight', transparent=True)
def plot_ctcf_patterns_vs_size(all_patterns):
figsize = get_default_figsize()
fig = plt.figure(figsize=(figsize[0]*2, figsize[1]*1.5))
ax = sns.boxplot(data=all_patterns,
x='n_ctcf_sites', y='size',
hue_order=['Same',
'Convergent',
'Divergent',
"Convergent-Divergent"],
hue='pattern_class',
showfliers=False,
order=[2, 3, 4],
palette=ctcf_colors, medianprops={'color':'red'})
# add_stat_annotation(ax, data=all_patterns, x='n_ctcf_sites', y='size',
# order=[2, 3, 4],
# hue_order=['Same',
# 'Convergent',
# 'Divergent',
# "Convergent-Divergent"],
# hue='pattern_class',
# box_pairs=[
# ( (2, "Same"), (2, "Convergent") ),
# ( (2, "Same"), (2, "Divergent") ),
# ( (2, "Divergent"), (2, "Convergent") ),
# ( (3, "Same"), (3, "Convergent") ),
# ( (3, "Same"), (3, "Divergent") ),
# ( (3, "Same"), (3, "Convergent-Divergent") ),
# ( (3, "Divergent"), (3, "Convergent") ),
# # ( (3, "Convergent"), (3, "Convergent-Divergent") ),
# # ( (3, "Divergent"), (3, "Convergent-Divergent") ),
# # ( (4, "Same"), (4, "Convergent") ),
# # ( (4, "Same"), (4, "Divergent") ),
# # ( (4, "Same"), (4, "Convergent-Divergent") ),
# # ( (4, "Divergent"), (4, "Convergent") ),
# # ( (4, "Convergent"), (4, "Convergent-Divergent") ),
# # ( (4, "Divergent"), (4, "Convergent-Divergent") ),
# ],
# test='t-test_ind',
# text_format='full', loc='outside',
# verbose=2)
plt.legend(title="Pattern class")
plt.xlabel("N. CTCF sites composing the pattern")
plt.ylabel("Pattern size (bp)")
plt.ticklabel_format(axis='y', style='sci', scilimits=(0,0), useMathText=True)
fig.savefig(figures_path / "CTCF_patterns_vs_size.pdf", bbox_inches='tight', transparent=True)
def plot_perc_regions_bearing_CTCF_site_cat(
pois_ext_ctcf,
catname='point_of_interest',
path=figures_path / "perc_regions_bearing_CTCF_site_cat.pdf"):
figsize = get_default_figsize()
fig = plt.figure(figsize=(figsize[0] * 1.5, figsize[1]))
sns.barplot(data=pois_ext_ctcf.groupby(catname)[['all', 'S', 'CD', 'D', 'C']]\
.mean().reset_index()\
.melt(id_vars=catname),
x=catname, y='value', hue='variable', palette=ctcf_colors)
plt.grid(axis='y')
plt.ylim(0, 1)
yticks, _ = plt.yticks()
plt.yticks(yticks, ["{:.0f}%".format(yi * 100) for yi in yticks],
rotation=0)
plt.xlabel("TAD position")
plt.ylabel("Percentage of regions\nbearing a CTCF site")
plt.legend(loc='center left', bbox_to_anchor=(1, 0.5))
fig.savefig(path, bbox_inches='tight', transparent=True)
def plot_cons_bounds_vs_gm12878_DI(
aggregations_by_bound,
extended,
window_size,
measure,
ylabel="Average Directionality\nIndex on GM12878"):
figsize = get_default_figsize()
fig, axes = plt.subplots(1,
len(aggregations_by_bound.keys()),
sharex='col',
sharey=True,
tight_layout=True,
figsize=(figsize[0] * 7, figsize[1]))
for level in sorted(aggregations_by_bound.keys()):
y = aggregations_by_bound[level].mean(0)
axes[level - 1].plot(y, linewidth=1, color='black')
axes[level - 1].fill_between(np.arange(y.shape[0]),
y,
where=y < 0,
facecolor='red')
axes[level - 1].fill_between(np.arange(y.shape[0]),
y,
where=y >= 0,
facecolor='blue')
if level == 1:
axes[level - 1].set_ylabel(ylabel, fontsize="xx-large")
axes[level - 1].tick_params(axis='y', which='major')
axes[level - 1].grid()
axes[level - 1].set_title("$s = {}$".format(level), fontweight="bold")
axes[level - 1].set_xticks(
[0, extended / window_size, extended * 2 / window_size])
yticklabels = axes[level - 1].get_yticklabels()
length_n = "{}".format(extended // 1000)
axes[level - 1].set_xticklabels(
['-{}kb'.format(length_n), '0', '+{}kb'.format(length_n)])
axes[level - 1].set_xlabel("Distance from\nboundary center",
fontsize='xx-large')
fig.savefig(figures_path / "cons_bounds_vs_gm12878_{}.pdf".format(measure),
bbox_inches='tight',
transparent=True)
def plot_triangle_silouette(pos_triangles_vs_ctcfs, pos_triangles,
neg_triangles_vs_ctcfs, neg_triangles, path):
figsize = get_default_figsize()
fig, axes = plt.subplots(2, 1, figsize=(figsize[0] * 1.5, figsize[1] * 2))
X_pos = pos_triangles_vs_ctcfs.reindex(index = pos_triangles\
.sort_values('length', ascending=False)\
.triangle_uid.tolist()).dropna().values
X_pos[X_pos < 0] = -1
X_pos[X_pos > 0] = 1
axes[0].matshow(X_pos,
aspect='auto',
cmap='seismic_r',
interpolation='bilinear')
axes[0].set_yticklabels([])
axes[0].set_xticks([])
axes[0].set_yticks([])
axes[0].set_xticklabels([])
X_neg = neg_triangles_vs_ctcfs.reindex(index = neg_triangles\
.sort_values('length', ascending=False)\
.triangle_uid.tolist())\
.dropna().values
X_neg[X_neg < 0] = -1
X_neg[X_neg > 0] = 1
axes[1].matshow(X_neg,
aspect='auto',
cmap='seismic_r',
interpolation='bilinear')
axes[1].set_yticklabels([])
axes[1].set_yticks([])
axes[1].set_xticks([0, X_pos.shape[1] / 2, X_pos.shape[1]])
axes[1].set_xticklabels(['-1Mb', '0', '+1Mb'])
plt.xlabel("Distance from (+) to (-) inversion point")
axes[1].xaxis.set_ticks_position('bottom')
fig.savefig(figures_path / "triangle_enrichment.pdf",
bbox_inches='tight',
transparent=True)
def plot_conservation_shift_test(regions_n_cross):
figsize = get_default_figsize()
fig = plt.figure(figsize=(figsize[0] * 2, figsize[1] * 1.2))
sns.lineplot(data=regions_n_cross.assign(
conservation=lambda x: x.conservation.astype(str) + " times"),
x='shift',
y='count',
hue='conservation',
hue_order=["{} times".format(x) for x in range(1, 8)],
palette='Reds',
linewidth=2,
legend='full')
plt.xlabel("Shift")
plt.ticklabel_format(axis='x',
style='sci',
scilimits=(0, 0),
useMathText=True)
plt.ylabel("Average n. overlapping\nPC-HiC interactions")
plt.legend(loc='center left', bbox_to_anchor=(1, 0.5), ncol=1)
fig.savefig(figures_path / "cons_tads_vs_PC-HIC.pdf",
bbox_inches='tight',
transparent=True)
def plot_consensus_boundaries_intersection_with_GM12878(x):
consensus_boundaries_fixed_gm12878 = x.assign(
has_gm12878_bound=lambda x: x.n_gm12878_bounds.map(lambda y: 'Intersects GM12878 boundary' if y > 0 \
else 'Does not intersect GM12878 boundary'))
figsize = get_default_figsize()
fig, axes = plt.subplots(2,
1,
sharex=True,
figsize=(figsize[0] * 1.5, figsize[1] * 2))
sns.countplot(data=consensus_boundaries_fixed_gm12878,
x='n_cell_types',
hue='has_gm12878_bound',
ax=axes[0],
hue_order=[
'Intersects GM12878 boundary',
'Does not intersect GM12878 boundary'
])
axes[0].legend()
axes[0].grid(axis='y')
axes[0].set_ylabel("# boundaries")
axes[0].legend()
x = consensus_boundaries_fixed_gm12878.groupby(
['n_cell_types', 'has_gm12878_bound']).size().unstack()
x = x.div(x.sum(1), axis=0)
x['Intersects GM12878 boundary'].plot.bar(ax=axes[1])
plt.ylim(0, 0.8)
yticks, _ = plt.yticks()
plt.yticks(yticks, ["{:.0f}%".format(yi * 100) for yi in yticks],
rotation=0)
plt.xticks(rotation=0)
plt.xlabel("Boundary conservation ($s$)")
plt.ylabel("% boundaries overlapping\nwith a GM12878 boundary")
plt.grid(axis='y')
fig.savefig(figures_path / "cons_bounds_vs_gm12878_bounds.pdf",
bbox_inches='tight',
transparent=True)
def plot_simulation_stats(simulation_metrics, name_to_full_name, name_to_color,
name_to_style):
figsize = get_default_figsize()
fig, axes = plt.subplots(1,
2,
figsize=(figsize[0] * 2, figsize[1]),
tight_layout=True)
for name, stats in simulation_metrics.items():
axes[0].plot(stats[0],
stats[1],
label=name_to_full_name[name],
color=name_to_color[name],
style=name_to_style[name])
axes[1].plot(*order_for_ROC(stats[1], stats[2]),
label=name_to_full_name[name],
color=name_to_color[name],
style=name_to_style[name])
yticks = axes[0].get_yticks()
axes[0].set_yticklabels(["{:.0f}%".format(yi * 100) for yi in yticks])
axes[0].legend(title='Simulation')
axes[0].set_xlabel(
"Minimum threshold\n(Loop occurences in the simulation)")
axes[0].set_ylabel("RECALL (% HiCCUP loops\nrecovered by simulation)")
axes[0].grid()
yticks = axes[1].get_yticks()
axes[1].set_yticklabels(["{:.0f}%".format(yi * 100) for yi in yticks])
xticks = axes[1].get_xticks()
axes[1].set_xticklabels(["{:.0f}%".format(yi * 100) for yi in xticks])
axes[1].grid()
axes[1].legend(title='Simulation')
axes[1].set_xlabel("RECALL (% HiCCUP loops\nrecovered by simulation)")
axes[1].set_ylabel("PRECISION (% simulated\nloops confirmed by HiCCUP)")
handles, labels = axes[1].get_legend_handles_labels()
fig.savefig(figures_path / "simulation_stats.pdf",
bbox_inches='tight',
transparent=True)
def plot_patterns_by_window_and_class(pattern_enrichment_by_distance,
path=figures_path / "CTCF_patterns_vs_size.pdf",
vmin=-0.06, vmax=0.06):
figsize=get_default_figsize()
height_ratios=[
len(patternclass_to_pattern['Same']),
len(patternclass_to_pattern['Convergent']),
len(patternclass_to_pattern['Divergent']),
len(patternclass_to_pattern['Convergent-Divergent'])
]
fig, axes = plt.subplots(len(patternclass_to_pattern.keys()),
1, sharex=True,
figsize=(figsize[0]*1.6,
figsize[1]*len(patternclass_to_pattern.keys())*0.5),
gridspec_kw={'height_ratios': height_ratios},
tight_layout=True)
cbar_ax = fig.add_axes([1, .3, .03, .4])
pebd_names = pattern_enrichment_by_distance\
.assign(index=lambda x: x.index.map(window_string_converter))\
.set_index('index')
for i, p in enumerate(patternclass_to_pattern.items()):
pclass, pnames = p
sns.heatmap(pebd_names.T.loc[pnames]\
.applymap(lambda x: np.log10(x)), cbar= i == 0,
cmap='bwr', linewidth=0.01, vmin=vmin, vmax=vmax, ax=axes[i],
linecolor='black', cbar_ax= None if i != 0 else cbar_ax,
cbar_kws={'label': "$log_{10}(enrichment)$"})
for tick in axes[i].get_yticklabels():
tick.set_rotation(0)
if i != len(patternclass_to_pattern.keys()) - 1:
axes[i].set_xlabel("")
axes[i].tick_params(axis='x', bottom=False)
else:
axes[i].set_xlabel("Clustering window")
axes[i].text(x=-0.12, y=0.5, s=pclass, horizontalalignment='center',
verticalalignment='center',
rotation=90, transform=axes[i].transAxes)
fig.savefig(path, bbox_inches='tight', transparent=True)
pos_triangles_vs_ctcfs,
pos_triangles,
neg_triangles_vs_ctcfs,
neg_triangles,
figures_path / "triangle_enrichment_CTCF_complete_square.pdf",
cmap="seismic_r")
plot_triangle_enrichment_halves(pos_triangles_vs_ctcfs,
pos_triangles,
neg_triangles_vs_ctcfs,
neg_triangles,
figures_path /
"triangle_enrichment_CTCF_square.pdf",
cmap="seismic_r")
figsize = get_default_figsize()
fig, ax = plt.subplots(1, 2, figsize=(figsize[0] * 2, figsize[1]))
tbyleft = tads.sort_values("left_size", ascending=False)
ax[0].barh(np.arange(tbyleft.shape[0]),
-tbyleft.left_size,
0.5,
label='Positive DI region')
ax[0].barh(np.arange(tbyleft.shape[0]),
tbyleft.right_size,
0.5,
label='Negative DI region')
ax[0].invert_yaxis()
ax[0].set_xlim(-1e6, +1e6)
ax[0].set_xticks([-1e6, -5e5, 0, 5e5, 1e6])
ax[0].set_xticklabels(['-1 Mb', '-500 Kb', '0', '+500 Kb', '1 Mb'])
ax[0].legend(bbox_to_anchor=(0.5, 1.21), loc='upper center')
def plot_aggregations_by_conservation(aggregations_by_bound,
windows_with_ctcf):
figsize = get_default_figsize()
n_figs = len(windows_with_ctcf.n_cell_types.unique())
fig, axes = plt.subplots(3,
n_figs,
sharex='col',
sharey='row',
tight_layout=False,
figsize=(figsize[0] * n_figs, figsize[1] * 3))
for level in sorted(windows_with_ctcf.n_cell_types.unique()):
axes[0,
level - 1].plot(aggregations_by_bound[(level, 'ctcf_id')].mean(0),
label='both',
color=ctcf_colors['all'])
if level == 1:
axes[0, level - 1].legend(loc='upper left')
axes[0, level - 1].set_ylabel("Avg. CTCF sites per 5kb")
else:
axes[0, level - 1].legend().set_visible(False)
axes[0, level - 1].grid()
axes[0, level - 1].set_title("$s = {}$".format(level),
fontweight="bold",
fontsize='xx-large')
axes[0, level - 1].tick_params(axis='y')
axes[1,
level - 1].plot(aggregations_by_bound[(level, 'forward')].mean(0),
label='Forward',
color=ctcf_colors['forward'])
axes[1,
level - 1].plot(aggregations_by_bound[(level, 'reverse')].mean(0),
label='Reverse',
color=ctcf_colors['reverse'])
axes[1, level - 1].tick_params(axis='y')
if level == 1:
axes[1, level - 1].legend(loc='upper left')
axes[1, level - 1].set_ylabel("Avg. CTCF sites per 5kb")
else:
axes[1, level - 1].legend().set_visible(False)
axes[1, level - 1].grid()
axes[2, level - 1].plot(aggregations_by_bound[(level, 'S')].mean(0),
label='Same',
color=ctcf_colors['S'])
axes[2, level - 1].plot(aggregations_by_bound[(level, 'C')].mean(0),
label='Convergent',
color=ctcf_colors['C'])
axes[2, level - 1].plot(aggregations_by_bound[(level, 'D')].mean(0),
label='Divergent',
color=ctcf_colors['D'])
axes[2, level - 1].plot(aggregations_by_bound[(level, 'CD')].mean(0),
label='Convergent-Divergent',
color=ctcf_colors['CD'])
if level == 1:
axes[2, level - 1].legend(loc='upper left')
axes[2, level - 1].set_ylabel("Avg. CTCF sites per 5kb")
else:
axes[2, level - 1].legend().set_visible(False)
axes[2, level - 1].grid()
axes[2, level - 1].tick_params(axis='y')
axes[2, level - 1].set_xticks(
[0, extended / window_size, extended * 2 / window_size])
axes[2, level - 1].set_xticklabels(['-250kb', '0', '+250kb'])
axes[2, level - 1].set_xlabel("Distance from\nboundary center (kb)")
fig.savefig(figures_path / "aggregations_by_conservation.pdf",
bbox_inches='tight',
transparent=True)
