def getTicks(self, idx):
"""
This is essentially a wrapper around getProfileTicks to accomdate the fact that each column has its own ticks.
"""
xticks, xtickslabel = getProfileTicks(self.hm,
self.reference_point_label[idx],
self.start_label[idx],
self.end_label[idx], idx)
return xticks, xtickslabel
def __init__(self,
hm,
out_file_name,
plot_title='',
y_axis_label='',
y_min=None,
y_max=None,
averagetype='median',
reference_point_label='TSS',
start_label='TSS',
end_label="TES",
plot_height=7,
plot_width=11,
per_group=False,
plot_type='simple',
image_format=None,
color_list=None,
legend_location='auto',
plots_per_row=8):
"""
Using the hm matrix, makes a line plot
either per group or per sample
using the specified parameters.
Args:
hm: heatmapper object
out_file_name: string
plot_title: string
y_axis_label: list
y_min: int
y_max: int
averagetype: mean, sum, median
reference_point_label: string
start_label: string
end_label: string
plot_height: in cm
plot_width: in cm
per_group: bool
plot_type: string
image_format: string
color_list: list
legend_location:
plots_per_row: int
Returns:
"""
self.hm = hm
self.out_file_name = out_file_name
self.plot_title = plot_title
self.y_axis_label = y_axis_label
self.y_min = y_min
self.y_max = y_max
self.averagetype = averagetype
self.reference_point_label = reference_point_label
self.start_label = start_label
self.end_label = end_label
self.plot_height = plot_height
self.plot_width = plot_width
self.per_group = per_group
self.plot_type = plot_type
self.image_format = image_format
self.color_list = color_list
self.legend_location = legend_location
self.plots_per_row = plots_per_row
# decide how many plots are needed
if self.per_group:
self.numplots = self.hm.matrix.get_num_groups()
self.numlines = self.hm.matrix.get_num_samples()
else:
self.numplots = self.hm.matrix.get_num_samples()
self.numlines = self.hm.matrix.get_num_groups()
if self.numplots > self.plots_per_row:
rows = np.ceil(self.numplots / self.plots_per_row).astype(int)
cols = self.plots_per_row
else:
rows = 1
cols = self.numplots
self.grids = gridspec.GridSpec(rows, cols)
plt.rcParams['font.size'] = 10.0
self.font_p = FontProperties()
self.font_p.set_size('small')
# convert cm values to inches
plot_height_inches = rows * self.cm2inch(self.plot_height)[0]
self.fig = plt.figure(figsize=self.cm2inch(cols *
self.plot_width, rows *
self.plot_height))
self.fig.suptitle(self.plot_title, y=(1 - (0.06 / plot_height_inches)))
self.xticks, self.xtickslabel = getProfileTicks(
self.hm, self.reference_point_label, self.start_label,
self.end_label)
def plotMatrix(hm, outFileName,
colorMapDict={'colorMap': 'binary', 'missingDataColor': 'black'},
plotTitle='',
xAxisLabel='', yAxisLabel='', regionsLabel='',
zMin=None, zMax=None,
yMin=None, yMax=None,
averageType='median',
reference_point_label='TSS',
startLabel='TSS', endLabel="TES",
heatmapHeight=25,
heatmapWidth=7.5,
perGroup=False, whatToShow='plot, heatmap and colorbar',
plotType='simple',
image_format=None,
legend_location='upper-left'):
matrix_flatten = None
if zMin is None:
matrix_flatten = hm.matrix.flatten()
# try to avoid outliers by using np.percentile
zMin = np.percentile(matrix_flatten, 1.0)
if np.isnan(zMin):
zMin = None
if zMax is None:
if matrix_flatten is None:
matrix_flatten = hm.matrix.flatten()
# try to avoid outliers by using np.percentile
zMax = np.percentile(matrix_flatten, 98.0)
if np.isnan(zMax):
zMax = None
plt.rcParams['font.size'] = 8.0
fontP = FontProperties()
# fontP.set_size('small')
showSummaryPlot = False
showColorbar = False
if whatToShow == 'plot and heatmap':
showSummaryPlot = True
elif whatToShow == 'heatmap and colorbar':
showColorbar = True
elif whatToShow == 'plot, heatmap and colorbar':
showSummaryPlot = True
showColorbar = True
grids = prepare_layout(hm.matrix, (heatmapWidth, heatmapHeight),
showSummaryPlot, showColorbar, perGroup)
# figsize: w,h tuple in inches
figwidth = heatmapWidth / 2.54
figheight = heatmapHeight / 2.54
if showSummaryPlot:
# the summary plot ocupies a height
# equal to the fig width
figheight += figwidth
numsamples = hm.matrix.get_num_samples()
total_figwidth = figwidth * numsamples
if showColorbar:
total_figwidth += 1 / 2.54
fig = plt.figure(figsize=(total_figwidth, figheight))
hm.parameters['upstream']
hm.parameters['downstream']
hm.parameters['body']
hm.parameters['bin size']
xticks, xtickslabel = getProfileTicks(hm, reference_point_label, startLabel, endLabel)
xticks_heat, xtickslabel_heat = get_heatmap_ticks(hm, reference_point_label, startLabel, endLabel)
fig.suptitle(plotTitle, y=1 - (0.06 / figheight))
# colormap for the heatmap
if colorMapDict['colorMap']:
cmap = plt.get_cmap(colorMapDict['colorMap'])
if colorMapDict['colorList'] and len(colorMapDict['colorList']) > 0:
cmap = matplotlib.colors.LinearSegmentedColormap.from_list(
'my_cmap', colorMapDict['colorList'], N=colorMapDict['colorNumber'])
# color map for the summary plot (profile) on top of the heatmap
cmap_plot = plt.get_cmap('jet')
numgroups = hm.matrix.get_num_groups()
if perGroup:
color_list = cmap_plot(np.arange(hm.matrix.get_num_samples()) / hm.matrix.get_num_samples())
else:
color_list = cmap_plot(np.arange(numgroups) / numgroups)
cmap.set_bad(colorMapDict['missingDataColor']) # nans are printed using this color
# check if matrix is reference-point based using the upstream >0 value
# and is sorted by region length. If this is
# the case, prepare the data to plot a border at the regions end
if hm.parameters['upstream'] > 0 and \
hm.matrix.sort_using == 'region_length' and \
hm.matrix.sort_method != 'no':
_regions = hm.matrix.get_regions()
regions_length_in_bins = []
for _group in _regions:
_reg_len = []
for ind_reg in _group:
_len = ind_reg['end'] - ind_reg['start']
_reg_len.append((hm.parameters['upstream'] + _len) / hm.parameters['bin size'])
# print hm.parameters['upstream'] + (_len / hm.parameters['bin size'])
regions_length_in_bins.append(_reg_len)
else:
regions_length_in_bins = None
first_group = 0 # helper variable to place the title per sample/group
for sample in range(hm.matrix.get_num_samples()):
sample_idx = sample
for group in range(numgroups):
group_idx = group
# add the respective profile to the
# summary plot
sub_matrix = hm.matrix.get_matrix(group, sample)
if showSummaryPlot:
if perGroup:
sample_idx = sample + 2 # plot + spacer
else:
group += 2 # plot + spacer
first_group = 1
if perGroup:
ax = fig.add_subplot(grids[sample_idx, group])
else:
ax = fig.add_subplot(grids[group, sample])
if group == first_group and not showSummaryPlot and not perGroup:
title = hm.matrix.sample_labels[sample]
ax.set_title(title)
rows, cols = sub_matrix['matrix'].shape
interpolation_type = None if rows >= 1000 and cols >= 200 else 'nearest'
img = ax.imshow(sub_matrix['matrix'],
aspect='auto',
interpolation=interpolation_type,
origin='upper',
vmin=zMin,
vmax=zMax,
cmap=cmap,
extent=[0, cols, rows, 0])
# plot border at the end of the regions
# if ordered by length
if regions_length_in_bins is not None:
x_lim = ax.get_xlim()
y_lim = ax.get_ylim()
ax.plot(regions_length_in_bins[group_idx],
np.arange(len(regions_length_in_bins[group_idx])),
'--', color='black', linewidth=0.5, dashes=(3, 2))
ax.set_xlim(x_lim)
ax.set_ylim(y_lim)
if perGroup:
ax.axes.set_xlabel(sub_matrix['group'])
if sample < hm.matrix.get_num_samples() - 1:
ax.axes.get_xaxis().set_visible(False)
else:
ax.axes.get_xaxis().set_visible(False)
ax.axes.set_xlabel(xAxisLabel)
ax.axes.set_yticks([])
if perGroup and group == 0:
ax.axes.set_ylabel(sub_matrix['sample'])
elif not perGroup and sample == 0:
ax.axes.set_ylabel(sub_matrix['group'])
# add xticks to the bottom heatmap (last group)
ax.axes.get_xaxis().set_visible(True)
ax.axes.set_xticks(xticks_heat)
ax.axes.set_xticklabels(xtickslabel_heat, size=8)
# align the first and last label
# such that they don't fall off
# the heatmap sides
ticks = ax.xaxis.get_major_ticks()
ticks[0].label1.set_horizontalalignment('left')
ticks[-1].label1.set_horizontalalignment('right')
ax.get_xaxis().set_tick_params(
which='both',
top='off',
direction='out')
# plot the profiles on top of the heatmaps
if showSummaryPlot:
ax_list = []
if perGroup:
iterNum = numgroups
else:
iterNum = hm.matrix.get_num_samples()
# plot each of the profiles
for sample_id in range(iterNum):
if perGroup:
title = hm.matrix.group_labels[sample_id]
else:
title = hm.matrix.sample_labels[sample_id]
if sample_id > 0:
ax_profile = fig.add_subplot(grids[0, sample_id],
sharey=ax_list[0])
else:
ax_profile = fig.add_subplot(grids[0, sample_id])
ax_profile.set_title(title)
if perGroup:
iterNum2 = hm.matrix.get_num_samples()
else:
iterNum2 = numgroups
for group in range(iterNum2):
if perGroup:
sub_matrix = hm.matrix.get_matrix(sample_id, group)
line_label = sub_matrix['sample']
else:
sub_matrix = hm.matrix.get_matrix(group, sample_id)
line_label = sub_matrix['group']
plot_single(ax_profile, sub_matrix['matrix'],
averageType,
color_list[group],
line_label,
plot_type='simple')
if sample_id > 0:
plt.setp(ax_profile.get_yticklabels(), visible=False)
if sample_id == 0 and yAxisLabel != '':
ax_profile.set_ylabel(yAxisLabel)
ax_profile.axes.set_xticks(xticks)
ax_profile.axes.set_xticklabels(xtickslabel)
ax_list.append(ax_profile)
# align the first and last label
# such that they don't fall off
# the heatmap sides
ticks = ax_profile.xaxis.get_major_ticks()
ticks[0].label1.set_horizontalalignment('left')
ticks[-1].label1.set_horizontalalignment('right')
# reduce the number of yticks by half
ax_list[0].set_ylim(yMin, yMax)
num_ticks = len(ax_list[0].get_yticks())
yticks = [ax_list[0].get_yticks()[i] for i in range(1, num_ticks, 2)]
ax_list[0].set_yticks(yticks)
if legend_location != 'none':
ax_list[-1].legend(loc=legend_location.replace('-', ' '), ncol=1, prop=fontP,
frameon=False, markerscale=0.5)
if showColorbar:
if showSummaryPlot:
# we dont want to colorbar to extend
# over the profiles row
grid_start = 2
else:
grid_start = 0
ax = fig.add_subplot(grids[grid_start:, -1])
fig.colorbar(img, cax=ax)
plt.subplots_adjust(wspace=0.10, hspace=0.025, top=0.85, bottom=0, left=0.04, right=0.96)
plt.savefig(outFileName, bbox_inches='tight', pdd_inches=0, dpi=200,
format=image_format)
def plotMatrix(hm,
outFileName,
colorMapDict={
'colorMap': ['binary'],
'missingDataColor': 'black',
'alpha': 1.0
},
plotTitle='',
xAxisLabel='',
yAxisLabel='',
regionsLabel='',
zMin=None,
zMax=None,
yMin=None,
yMax=None,
averageType='median',
reference_point_label='TSS',
startLabel='TSS',
endLabel="TES",
heatmapHeight=25,
heatmapWidth=7.5,
perGroup=False,
whatToShow='plot, heatmap and colorbar',
image_format=None,
legend_location='upper-left',
box_around_heatmaps=True,
dpi=200):
matrix_flatten = None
if zMin is None:
matrix_flatten = hm.matrix.flatten()
# try to avoid outliers by using np.percentile
zMin = np.percentile(matrix_flatten, 1.0)
if np.isnan(zMin):
zMin = [None]
else:
zMin = [zMin] # convert to list to support multiple entries
if zMax is None:
if matrix_flatten is None:
matrix_flatten = hm.matrix.flatten()
# try to avoid outliers by using np.percentile
zMax = np.percentile(matrix_flatten, 98.0)
if np.isnan(zMax):
zMax = [None]
else:
zMax = [zMax]
plt.rcParams['font.size'] = 8.0
fontP = FontProperties()
showSummaryPlot = False
showColorbar = False
if whatToShow == 'plot and heatmap':
showSummaryPlot = True
elif whatToShow == 'heatmap and colorbar':
showColorbar = True
elif whatToShow == 'plot, heatmap and colorbar':
showSummaryPlot = True
showColorbar = True
# colormap for the heatmap
if colorMapDict['colorMap']:
cmap = []
for color_map in colorMapDict['colorMap']:
cmap.append(plt.get_cmap(color_map))
cmap[-1].set_bad(colorMapDict['missingDataColor']
) # nans are printed using this color
if colorMapDict['colorList'] and len(colorMapDict['colorList']) > 0:
# make a cmap for each color list given
cmap = []
for color_list in colorMapDict['colorList']:
cmap.append(
matplotlib.colors.LinearSegmentedColormap.from_list(
'my_cmap',
color_list.replace(' ', '').split(","),
N=colorMapDict['colorNumber']))
cmap[-1].set_bad(colorMapDict['missingDataColor']
) # nans are printed using this color
if len(cmap) > 1 or len(zMin) > 1 or len(zMax) > 1:
# position color bar below heatmap when more than one
# heatmap color is given
colorbar_position = 'below'
else:
colorbar_position = 'side'
grids = prepare_layout(hm.matrix, (heatmapWidth, heatmapHeight),
showSummaryPlot, showColorbar, perGroup,
colorbar_position)
# figsize: w,h tuple in inches
figwidth = heatmapWidth / 2.54
figheight = heatmapHeight / 2.54
if showSummaryPlot:
# the summary plot ocupies a height
# equal to the fig width
figheight += figwidth
numsamples = hm.matrix.get_num_samples()
if perGroup:
num_cols = hm.matrix.get_num_groups()
else:
num_cols = numsamples
total_figwidth = figwidth * num_cols
if showColorbar:
if colorbar_position == 'below':
figheight += 1 / 2.54
else:
total_figwidth += 1 / 2.54
fig = plt.figure(figsize=(total_figwidth, figheight))
xticks, xtickslabel = getProfileTicks(hm, reference_point_label,
startLabel, endLabel)
xticks_heat, xtickslabel_heat = get_heatmap_ticks(hm,
reference_point_label,
startLabel, endLabel)
fig.suptitle(plotTitle, y=1 - (0.06 / figheight))
# color map for the summary plot (profile) on top of the heatmap
cmap_plot = plt.get_cmap('jet')
numgroups = hm.matrix.get_num_groups()
if perGroup:
color_list = cmap_plot(
np.arange(hm.matrix.get_num_samples()) /
hm.matrix.get_num_samples())
else:
color_list = cmap_plot(np.arange(numgroups) / numgroups)
alpha = colorMapDict['alpha']
# check if matrix is reference-point based using the upstream >0 value
# and is sorted by region length. If this is
# the case, prepare the data to plot a border at the regions end
if hm.parameters['upstream'] > 0 and \
hm.matrix.sort_using == 'region_length' and \
hm.matrix.sort_method != 'no':
_regions = hm.matrix.get_regions()
regions_length_in_bins = []
for _group in _regions:
_reg_len = []
for ind_reg in _group:
_len = ind_reg['end'] - ind_reg['start']
_reg_len.append((hm.parameters['upstream'] + _len) /
hm.parameters['bin size'])
regions_length_in_bins.append(_reg_len)
else:
regions_length_in_bins = None
first_group = 0 # helper variable to place the title per sample/group
for sample in range(hm.matrix.get_num_samples()):
sample_idx = sample
for group in range(numgroups):
group_idx = group
# add the respective profile to the
# summary plot
sub_matrix = hm.matrix.get_matrix(group, sample)
if showSummaryPlot:
if perGroup:
sample_idx = sample + 2 # plot + spacer
else:
group += 2 # plot + spacer
first_group = 1
if perGroup:
ax = fig.add_subplot(grids[sample_idx, group])
# the remainder (%) is used to iterate
# over the available color maps (cmap).
# if the user only provided, lets say two
# and there are 10 groups, colormaps they are reused every
# two groups.
cmap_idx = group_idx % len(cmap)
zmin_idx = group_idx % len(zMin)
zmax_idx = group_idx % len(zMax)
else:
ax = fig.add_subplot(grids[group, sample])
# see above for the use of '%'
cmap_idx = sample % len(cmap)
zmin_idx = sample % len(zMin)
zmax_idx = sample % len(zMax)
if group == first_group and not showSummaryPlot and not perGroup:
title = hm.matrix.sample_labels[sample]
ax.set_title(title)
if box_around_heatmaps is False:
# Turn off the boxes around the individual heatmaps
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.spines['bottom'].set_visible(False)
ax.spines['left'].set_visible(False)
rows, cols = sub_matrix['matrix'].shape
interpolation_type = None if rows >= 1000 and cols >= 200 else 'nearest'
img = ax.imshow(sub_matrix['matrix'],
aspect='auto',
interpolation=interpolation_type,
origin='upper',
vmin=zMin[zmin_idx],
vmax=zMax[zmax_idx],
cmap=cmap[cmap_idx],
alpha=alpha,
extent=[0, cols, rows, 0])
img.set_rasterized(True)
# plot border at the end of the regions
# if ordered by length
if regions_length_in_bins is not None:
x_lim = ax.get_xlim()
y_lim = ax.get_ylim()
ax.plot(regions_length_in_bins[group_idx],
np.arange(len(regions_length_in_bins[group_idx])),
'--',
color='black',
linewidth=0.5,
dashes=(3, 2))
ax.set_xlim(x_lim)
ax.set_ylim(y_lim)
if perGroup:
ax.axes.set_xlabel(sub_matrix['group'])
if sample < hm.matrix.get_num_samples() - 1:
ax.axes.get_xaxis().set_visible(False)
else:
ax.axes.get_xaxis().set_visible(False)
ax.axes.set_xlabel(xAxisLabel)
ax.axes.set_yticks([])
if perGroup and group == 0:
ax.axes.set_ylabel(sub_matrix['sample'])
elif not perGroup and sample == 0:
ax.axes.set_ylabel(sub_matrix['group'])
# add labels to last block in a column
if (perGroup and sample == numsamples - 1) or \
(not perGroup and group_idx == numgroups - 1):
# add xticks to the bottom heatmap (last group)
ax.axes.get_xaxis().set_visible(True)
ax.axes.set_xticks(xticks_heat)
ax.axes.set_xticklabels(xtickslabel_heat, size=8)
# align the first and last label
# such that they don't fall off
# the heatmap sides
ticks = ax.xaxis.get_major_ticks()
ticks[0].label1.set_horizontalalignment('left')
ticks[-1].label1.set_horizontalalignment('right')
ax.get_xaxis().set_tick_params(which='both',
top='off',
direction='out')
if showColorbar and colorbar_position == 'below':
# draw a colormap per each heatmap below the last block
if perGroup:
col = group_idx
else:
col = sample
ax = fig.add_subplot(grids[-1, col])
tick_locator = ticker.MaxNLocator(nbins=3)
cbar = fig.colorbar(img,
cax=ax,
alpha=alpha,
orientation='horizontal',
ticks=tick_locator)
labels = cbar.ax.get_xticklabels()
ticks = cbar.ax.get_xticks()
if ticks[0] == 0:
# if the label is at the start of the colobar
# move it a bit inside to avoid overlapping
# with other labels
labels[0].set_horizontalalignment('left')
if ticks[-1] == 1:
# if the label is at the end of the colobar
# move it a bit inside to avoid overlapping
# with other labels
labels[-1].set_horizontalalignment('right')
# cbar.ax.set_xticklabels(labels, rotation=90)
# plot the profiles on top of the heatmaps
if showSummaryPlot:
ax_list = []
if perGroup:
iterNum = numgroups
else:
iterNum = hm.matrix.get_num_samples()
# plot each of the profiles
for sample_id in range(iterNum):
if perGroup:
title = hm.matrix.group_labels[sample_id]
else:
title = hm.matrix.sample_labels[sample_id]
if sample_id > 0:
ax_profile = fig.add_subplot(grids[0, sample_id],
sharey=ax_list[0])
else:
ax_profile = fig.add_subplot(grids[0, sample_id])
ax_profile.set_title(title)
if perGroup:
iterNum2 = hm.matrix.get_num_samples()
else:
iterNum2 = numgroups
for group in range(iterNum2):
if perGroup:
sub_matrix = hm.matrix.get_matrix(sample_id, group)
line_label = sub_matrix['sample']
else:
sub_matrix = hm.matrix.get_matrix(group, sample_id)
line_label = sub_matrix['group']
plot_single(ax_profile,
sub_matrix['matrix'],
averageType,
color_list[group],
line_label,
plot_type='simple')
if sample_id > 0:
plt.setp(ax_profile.get_yticklabels(), visible=False)
if sample_id == 0 and yAxisLabel != '':
ax_profile.set_ylabel(yAxisLabel)
ax_profile.axes.set_xticks(xticks)
ax_profile.axes.set_xticklabels(xtickslabel)
ax_list.append(ax_profile)
# align the first and last label
# such that they don't fall off
# the heatmap sides
ticks = ax_profile.xaxis.get_major_ticks()
ticks[0].label1.set_horizontalalignment('left')
ticks[-1].label1.set_horizontalalignment('right')
# It turns out that set_ylim only takes np.float64s
if yMin:
yMin = np.float64(yMin)
if yMax:
yMax = np.float64(yMax)
ax_list[0].set_ylim(yMin, yMax)
# reduce the number of yticks by half
num_ticks = len(ax_list[0].get_yticks())
yticks = [ax_list[0].get_yticks()[i] for i in range(1, num_ticks, 2)]
ax_list[0].set_yticks(yticks)
if legend_location != 'none':
ax_list[-1].legend(loc=legend_location.replace('-', ' '),
ncol=1,
prop=fontP,
frameon=False,
markerscale=0.5)
if showColorbar and colorbar_position != 'below':
if showSummaryPlot:
# we don't want to colorbar to extend
# over the profiles and spacer top rows
grid_start = 2
else:
grid_start = 0
ax = fig.add_subplot(grids[grid_start:, -1])
fig.colorbar(img, cax=ax, alpha=alpha)
if box_around_heatmaps:
plt.subplots_adjust(wspace=0.10,
hspace=0.025,
top=0.85,
bottom=0,
left=0.04,
right=0.96)
else:
# When no box is plotted the space between heatmaps is reduced
plt.subplots_adjust(wspace=0.05,
hspace=0.01,
top=0.85,
bottom=0,
left=0.04,
right=0.96)
plt.savefig(outFileName,
bbox_inches='tight',
pdd_inches=0,
dpi=dpi,
format=image_format)
plt.close()
def __init__(self, hm, out_file_name,
plot_title='', y_axis_label='',
y_min=None, y_max=None,
averagetype='median',
reference_point_label='TSS',
start_label='TSS', end_label="TES",
plot_height=7,
plot_width=11,
per_group=False,
plot_type='simple',
image_format=None,
color_list=None,
legend_location='auto',
plots_per_row=8):
"""
Using the hm matrix, makes a line plot
either per group or per sample
using the specified parameters.
Args:
hm: heatmapper object
out_file_name: string
plot_title: string
y_axis_label: list
y_min: int
y_max: int
averagetype: mean, sum, median
reference_point_label: string
start_label: string
end_label: string
plot_height: in cm
plot_width: in cm
per_group: bool
plot_type: string
image_format: string
color_list: list
legend_location:
plots_per_row: int
Returns:
"""
self.hm = hm
self.out_file_name = out_file_name
self.plot_title = plot_title
self.y_axis_label = y_axis_label
self.y_min = y_min
self.y_max = y_max
self.averagetype = averagetype
self.reference_point_label = reference_point_label
self.start_label = start_label
self.end_label = end_label
self.plot_height = plot_height
self.plot_width = plot_width
self.per_group = per_group
self.plot_type = plot_type
self.image_format = image_format
self.color_list = color_list
self.legend_location = legend_location
self.plots_per_row = plots_per_row
# decide how many plots are needed
if self.per_group:
self.numplots = self.hm.matrix.get_num_groups()
self.numlines = self.hm.matrix.get_num_samples()
else:
self.numplots = self.hm.matrix.get_num_samples()
self.numlines = self.hm.matrix.get_num_groups()
if self.numplots > self.plots_per_row:
rows = np.ceil(self.numplots / self.plots_per_row).astype(int)
cols = self.plots_per_row
else:
rows = 1
cols = self.numplots
self.grids = gridspec.GridSpec(rows, cols)
plt.rcParams['font.size'] = 10.0
self.font_p = FontProperties()
self.font_p.set_size('small')
# convert cm values to inches
plot_height_inches = rows * self.cm2inch(self.plot_height)[0]
self.fig = plt.figure(figsize=self.cm2inch(cols * self.plot_width, rows * self.plot_height))
self.fig.suptitle(self.plot_title, y=(1 - (0.06 / plot_height_inches)))
self.xticks, self.xtickslabel = getProfileTicks(self.hm, self.reference_point_label,
self.start_label, self.end_label)
def plotMatrix(hm, outFileName,
colorMapDict={'colorMap': ['binary'], 'missingDataColor': 'black', 'alpha': 1.0},
plotTitle='',
xAxisLabel='', yAxisLabel='', regionsLabel='',
zMin=None, zMax=None,
yMin=None, yMax=None,
averageType='median',
reference_point_label='TSS',
startLabel='TSS', endLabel="TES",
heatmapHeight=25,
heatmapWidth=7.5,
perGroup=False, whatToShow='plot, heatmap and colorbar',
image_format=None,
legend_location='upper-left',
box_around_heatmaps=True,
dpi=200):
matrix_flatten = None
if zMin is None:
matrix_flatten = hm.matrix.flatten()
# try to avoid outliers by using np.percentile
zMin = np.percentile(matrix_flatten, 1.0)
if np.isnan(zMin):
zMin = [None]
else:
zMin = [zMin] # convert to list to support multiple entries
if zMax is None:
if matrix_flatten is None:
matrix_flatten = hm.matrix.flatten()
# try to avoid outliers by using np.percentile
zMax = np.percentile(matrix_flatten, 98.0)
if np.isnan(zMax) or zMax <= zMin[0]:
zMax = [None]
else:
zMax = [zMax]
if yMin is None:
yMin = [None]
if yMax is None:
yMax = [None]
if not isinstance(yMin, list):
yMin = [yMin]
if not isinstance(yMax, list):
yMax = [yMax]
plt.rcParams['font.size'] = 8.0
fontP = FontProperties()
showSummaryPlot = False
showColorbar = False
if whatToShow == 'plot and heatmap':
showSummaryPlot = True
elif whatToShow == 'heatmap and colorbar':
showColorbar = True
elif whatToShow == 'plot, heatmap and colorbar':
showSummaryPlot = True
showColorbar = True
# colormap for the heatmap
if colorMapDict['colorMap']:
cmap = []
for color_map in colorMapDict['colorMap']:
cmap.append(plt.get_cmap(color_map))
cmap[-1].set_bad(colorMapDict['missingDataColor']) # nans are printed using this color
if colorMapDict['colorList'] and len(colorMapDict['colorList']) > 0:
# make a cmap for each color list given
cmap = []
for color_list in colorMapDict['colorList']:
cmap.append(matplotlib.colors.LinearSegmentedColormap.from_list(
'my_cmap', color_list.replace(' ', '').split(","), N=colorMapDict['colorNumber']))
cmap[-1].set_bad(colorMapDict['missingDataColor']) # nans are printed using this color
if len(cmap) > 1 or len(zMin) > 1 or len(zMax) > 1:
# position color bar below heatmap when more than one
# heatmap color is given
colorbar_position = 'below'
else:
colorbar_position = 'side'
grids = prepare_layout(hm.matrix, (heatmapWidth, heatmapHeight),
showSummaryPlot, showColorbar, perGroup, colorbar_position)
# figsize: w,h tuple in inches
figwidth = heatmapWidth / 2.54
figheight = heatmapHeight / 2.54
if showSummaryPlot:
# the summary plot ocupies a height
# equal to the fig width
figheight += figwidth
numsamples = hm.matrix.get_num_samples()
if perGroup:
num_cols = hm.matrix.get_num_groups()
else:
num_cols = numsamples
total_figwidth = figwidth * num_cols
if showColorbar:
if colorbar_position == 'below':
figheight += 1 / 2.54
else:
total_figwidth += 1 / 2.54
fig = plt.figure(figsize=(total_figwidth, figheight))
xticks, xtickslabel = getProfileTicks(hm, reference_point_label, startLabel, endLabel)
xticks_heat, xtickslabel_heat = get_heatmap_ticks(hm, reference_point_label, startLabel, endLabel)
fig.suptitle(plotTitle, y=1 - (0.06 / figheight))
# color map for the summary plot (profile) on top of the heatmap
cmap_plot = plt.get_cmap('jet')
numgroups = hm.matrix.get_num_groups()
if perGroup:
color_list = cmap_plot(np.arange(hm.matrix.get_num_samples()) / hm.matrix.get_num_samples())
else:
color_list = cmap_plot(np.arange(numgroups) / numgroups)
alpha = colorMapDict['alpha']
# check if matrix is reference-point based using the upstream >0 value
# and is sorted by region length. If this is
# the case, prepare the data to plot a border at the regions end
if hm.parameters['upstream'] > 0 and \
hm.matrix.sort_using == 'region_length' and \
hm.matrix.sort_method != 'no':
_regions = hm.matrix.get_regions()
regions_length_in_bins = []
for _group in _regions:
_reg_len = []
for ind_reg in _group:
if isinstance(ind_reg, dict):
_len = ind_reg['end'] - ind_reg['start']
else:
_len = sum([x[1] - x[0] for x in ind_reg[1]])
_reg_len.append((hm.parameters['upstream'] + _len) / hm.parameters['bin size'])
regions_length_in_bins.append(_reg_len)
else:
regions_length_in_bins = None
# plot the profiles on top of the heatmaps
if showSummaryPlot:
if perGroup:
iterNum = numgroups
iterNum2 = hm.matrix.get_num_samples()
else:
iterNum = hm.matrix.get_num_samples()
iterNum2 = numgroups
ax_list = addProfilePlot(hm, plt, fig, grids, iterNum, iterNum2, perGroup, averageType, xticks, xtickslabel, yAxisLabel, color_list, yMin, yMax, None, None, colorbar_position)
if len(yMin) > 1 or len(yMax) > 1:
# replot with a tight layout
import matplotlib.tight_layout as tl
specList = tl.get_subplotspec_list(fig.axes, grid_spec=grids)
renderer = tl.get_renderer(fig)
kwargs = tl.get_tight_layout_figure(fig, fig.axes, specList, renderer, pad=1.08)
for ax in ax_list:
fig.delaxes(ax)
ax_list = addProfilePlot(hm, plt, fig, grids, iterNum, iterNum2, perGroup, averageType, xticks, xtickslabel, yAxisLabel, color_list, yMin, yMax, kwargs['wspace'], kwargs['hspace'], colorbar_position)
# reduce the number of yticks by half
num_ticks = len(ax_list[0].get_yticks())
yticks = [ax_list[0].get_yticks()[i] for i in range(1, num_ticks, 2)]
ax_list[0].set_yticks(yticks)
if legend_location != 'none':
ax_list[-1].legend(loc=legend_location.replace('-', ' '), ncol=1, prop=fontP,
frameon=False, markerscale=0.5)
first_group = 0 # helper variable to place the title per sample/group
for sample in range(hm.matrix.get_num_samples()):
sample_idx = sample
for group in range(numgroups):
group_idx = group
# add the respective profile to the
# summary plot
sub_matrix = hm.matrix.get_matrix(group, sample)
if showSummaryPlot:
if perGroup:
sample_idx = sample + 2 # plot + spacer
else:
group += 2 # plot + spacer
first_group = 1
if perGroup:
ax = fig.add_subplot(grids[sample_idx, group])
# the remainder (%) is used to iterate
# over the available color maps (cmap).
# if the user only provided, lets say two
# and there are 10 groups, colormaps they are reused every
# two groups.
cmap_idx = group_idx % len(cmap)
zmin_idx = group_idx % len(zMin)
zmax_idx = group_idx % len(zMax)
else:
ax = fig.add_subplot(grids[group, sample])
# see above for the use of '%'
cmap_idx = sample % len(cmap)
zmin_idx = sample % len(zMin)
zmax_idx = sample % len(zMax)
if group == first_group and not showSummaryPlot and not perGroup:
title = hm.matrix.sample_labels[sample]
ax.set_title(title)
if box_around_heatmaps is False:
# Turn off the boxes around the individual heatmaps
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.spines['bottom'].set_visible(False)
ax.spines['left'].set_visible(False)
rows, cols = sub_matrix['matrix'].shape
interpolation_type = None if rows >= 1000 and cols >= 200 else 'nearest'
img = ax.imshow(sub_matrix['matrix'],
aspect='auto',
interpolation=interpolation_type,
origin='upper',
vmin=zMin[zmin_idx],
vmax=zMax[zmax_idx],
cmap=cmap[cmap_idx],
alpha=alpha,
extent=[0, cols, rows, 0])
img.set_rasterized(True)
# plot border at the end of the regions
# if ordered by length
if regions_length_in_bins is not None:
x_lim = ax.get_xlim()
y_lim = ax.get_ylim()
ax.plot(regions_length_in_bins[group_idx],
np.arange(len(regions_length_in_bins[group_idx])),
'--', color='black', linewidth=0.5, dashes=(3, 2))
ax.set_xlim(x_lim)
ax.set_ylim(y_lim)
if perGroup:
ax.axes.set_xlabel(sub_matrix['group'])
if sample < hm.matrix.get_num_samples() - 1:
ax.axes.get_xaxis().set_visible(False)
else:
ax.axes.get_xaxis().set_visible(False)
ax.axes.set_xlabel(xAxisLabel)
ax.axes.set_yticks([])
if perGroup and group == 0:
ax.axes.set_ylabel(sub_matrix['sample'])
elif not perGroup and sample == 0:
ax.axes.set_ylabel(sub_matrix['group'])
# add labels to last block in a column
if (perGroup and sample == numsamples - 1) or \
(not perGroup and group_idx == numgroups - 1):
# add xticks to the bottom heatmap (last group)
ax.axes.get_xaxis().set_visible(True)
if np.ceil(max(xticks_heat)) != float(sub_matrix['matrix'].shape[1]):
tickscale = float(sub_matrix['matrix'].shape[1]) / max(xticks_heat)
xticks_heat_use = [x * tickscale for x in xticks_heat]
ax.axes.set_xticks(xticks_heat_use)
else:
ax.axes.set_xticks(xticks_heat)
ax.axes.set_xticklabels(xtickslabel_heat, size=8)
# align the first and last label
# such that they don't fall off
# the heatmap sides
ticks = ax.xaxis.get_major_ticks()
ticks[0].label1.set_horizontalalignment('left')
ticks[-1].label1.set_horizontalalignment('right')
ax.get_xaxis().set_tick_params(
which='both',
top='off',
direction='out')
if showColorbar and colorbar_position == 'below':
# draw a colormap per each heatmap below the last block
if perGroup:
col = group_idx
else:
col = sample
ax = fig.add_subplot(grids[-1, col])
tick_locator = ticker.MaxNLocator(nbins=3)
cbar = fig.colorbar(img, cax=ax, alpha=alpha, orientation='horizontal', ticks=tick_locator)
labels = cbar.ax.get_xticklabels()
ticks = cbar.ax.get_xticks()
if ticks[0] == 0:
# if the label is at the start of the colobar
# move it a bit inside to avoid overlapping
# with other labels
labels[0].set_horizontalalignment('left')
if ticks[-1] == 1:
# if the label is at the end of the colobar
# move it a bit inside to avoid overlapping
# with other labels
labels[-1].set_horizontalalignment('right')
# cbar.ax.set_xticklabels(labels, rotation=90)
if showColorbar and colorbar_position != 'below':
if showSummaryPlot:
# we don't want to colorbar to extend
# over the profiles and spacer top rows
grid_start = 2
else:
grid_start = 0
ax = fig.add_subplot(grids[grid_start:, -1])
fig.colorbar(img, cax=ax, alpha=alpha)
if box_around_heatmaps:
plt.subplots_adjust(wspace=0.10, hspace=0.025, top=0.85, bottom=0, left=0.04, right=0.96)
else:
# When no box is plotted the space between heatmaps is reduced
plt.subplots_adjust(wspace=0.05, hspace=0.01, top=0.85, bottom=0, left=0.04, right=0.96)
plt.savefig(outFileName, bbox_inches='tight', pdd_inches=0, dpi=dpi, format=image_format)
plt.close()
def getTicks(self, idx):
"""
This is essentially a wrapper around getProfileTicks to accomdate the fact that each column has its own ticks.
"""
xticks, xtickslabel = getProfileTicks(self.hm, self.reference_point_label[idx], self.start_label[idx], self.end_label[idx], idx)
return xticks, xtickslabel
def plotMatrix(hm,
outFileName,
colorMapDict={
'colorMap': 'binary',
'missingDataColor': 'black'
},
plotTitle='',
xAxisLabel='',
yAxisLabel='',
regionsLabel='',
zMin=None,
zMax=None,
yMin=None,
yMax=None,
averageType='median',
reference_point_label='TSS',
startLabel='TSS',
endLabel="TES",
heatmapHeight=25,
heatmapWidth=7.5,
perGroup=False,
whatToShow='plot, heatmap and colorbar',
plotType='simple',
image_format=None,
legend_location='upper-left'):
matrix_flatten = None
if zMin is None:
matrix_flatten = hm.matrix.flatten()
# try to avoid outliers by using np.percentile
zMin = np.percentile(matrix_flatten, 1.0)
if np.isnan(zMin):
zMin = None
if zMax is None:
if matrix_flatten is None:
matrix_flatten = hm.matrix.flatten()
# try to avoid outliers by using np.percentile
zMax = np.percentile(matrix_flatten, 98.0)
if np.isnan(zMax):
zMax = None
plt.rcParams['font.size'] = 8.0
fontP = FontProperties()
# fontP.set_size('small')
showSummaryPlot = False
showColorbar = False
if whatToShow == 'plot and heatmap':
showSummaryPlot = True
elif whatToShow == 'heatmap and colorbar':
showColorbar = True
else:
showSummaryPlot = True
showColorbar = True
grids = prepare_layout(hm.matrix, (heatmapWidth, heatmapHeight),
showSummaryPlot, showColorbar, perGroup)
# figsize: w,h tuple in inches
figwidth = heatmapWidth / 2.54
figheight = heatmapHeight / 2.54
if showSummaryPlot:
# the summary plot ocupies a height
# equal to the fig width
figheight += figwidth
numsamples = hm.matrix.get_num_samples()
total_figwidth = figwidth * numsamples
if showColorbar:
total_figwidth += 1 / 2.54
fig = plt.figure(figsize=(total_figwidth, figheight))
hm.parameters['upstream']
hm.parameters['downstream']
hm.parameters['body']
hm.parameters['bin size']
xticks, xtickslabel = getProfileTicks(hm, reference_point_label,
startLabel, endLabel)
xticks_heat, xtickslabel_heat = get_heatmap_ticks(hm,
reference_point_label,
startLabel, endLabel)
fig.suptitle(plotTitle, y=1 - (0.06 / figheight))
# colormap for the heatmap
if colorMapDict['colorMap']:
cmap = plt.get_cmap(colorMapDict['colorMap'])
if colorMapDict['colorList'] and len(colorMapDict['colorList']) > 0:
cmap = matplotlib.colors.LinearSegmentedColormap.from_list(
'my_cmap',
colorMapDict['colorList'],
N=colorMapDict['colorNumber'])
# color map for the summary plot (profile) on top of the heatmap
cmap_plot = plt.get_cmap('jet')
numgroups = hm.matrix.get_num_groups()
if perGroup:
color_list = cmap_plot(
np.arange(hm.matrix.get_num_samples()) /
hm.matrix.get_num_samples())
else:
color_list = cmap_plot(np.arange(numgroups) / numgroups)
cmap.set_bad(
colorMapDict['missingDataColor']) # nans are printed using this color
# check if matrix is reference-point based using the upstream >0 value
# and is sorted by region length. If this is
# the case, prepare the data to plot a border at the regions end
if hm.parameters['upstream'] > 0 and \
hm.matrix.sort_using == 'region_length' and \
hm.matrix.sort_method != 'no':
_regions = hm.matrix.get_regions()
regions_length_in_bins = []
for _group in _regions:
_reg_len = []
for ind_reg in _group:
_len = ind_reg['end'] - ind_reg['start']
_reg_len.append((hm.parameters['upstream'] + _len) /
hm.parameters['bin size'])
# print hm.parameters['upstream'] + (_len / hm.parameters['bin size'])
regions_length_in_bins.append(_reg_len)
else:
regions_length_in_bins = None
first_group = 0 # helper variable to place the title per sample/group
for sample in range(hm.matrix.get_num_samples()):
sample_idx = sample
for group in range(numgroups):
group_idx = group
# add the respective profile to the
# summary plot
sub_matrix = hm.matrix.get_matrix(group, sample)
if showSummaryPlot:
if perGroup:
sample_idx = sample + 2 # plot + spacer
else:
group += 2 # plot + spacer
first_group = 1
if perGroup:
ax = fig.add_subplot(grids[sample_idx, group])
else:
ax = fig.add_subplot(grids[group, sample])
if group == first_group and not showSummaryPlot and not perGroup:
title = hm.matrix.sample_labels[sample]
ax.set_title(title)
rows, cols = sub_matrix['matrix'].shape
interpolation_type = 'bicubic' if rows > 200 and \
cols > 1000 else 'nearest'
img = ax.imshow(sub_matrix['matrix'],
aspect='auto',
interpolation=interpolation_type,
origin='upper',
vmin=zMin,
vmax=zMax,
cmap=cmap,
extent=[0, cols, rows, 0])
# plot border at the end of the regions
# if ordered by length
if regions_length_in_bins is not None:
x_lim = ax.get_xlim()
y_lim = ax.get_ylim()
ax.plot(regions_length_in_bins[group_idx],
np.arange(len(regions_length_in_bins[group_idx])),
'--',
color='black',
linewidth=0.5,
dashes=(3, 2))
ax.set_xlim(x_lim)
ax.set_ylim(y_lim)
if perGroup:
ax.axes.set_xlabel(sub_matrix['group'])
if sample < hm.matrix.get_num_samples() - 1:
ax.axes.get_xaxis().set_visible(False)
else:
ax.axes.get_xaxis().set_visible(False)
ax.axes.set_xlabel(xAxisLabel)
ax.axes.set_yticks([])
if perGroup and group == 0:
ax.axes.set_ylabel(sub_matrix['sample'])
elif not perGroup and sample == 0:
ax.axes.set_ylabel(sub_matrix['group'])
# add xticks to the bottom heatmap (last group)
ax.axes.get_xaxis().set_visible(True)
ax.axes.set_xticks(xticks_heat)
ax.axes.set_xticklabels(xtickslabel_heat, size=8)
# align the first and last label
# such that they don't fall off
# the heatmap sides
ticks = ax.xaxis.get_major_ticks()
ticks[0].label1.set_horizontalalignment('left')
ticks[-1].label1.set_horizontalalignment('right')
ax.get_xaxis().set_tick_params(which='both',
top='off',
direction='out')
# plot the profiles on top of the heatmaps
if showSummaryPlot:
ax_list = []
if perGroup:
iterNum = numgroups
else:
iterNum = hm.matrix.get_num_samples()
# plot each of the profiles
for sample_id in range(iterNum):
if perGroup:
title = hm.matrix.group_labels[sample_id]
else:
title = hm.matrix.sample_labels[sample_id]
if sample_id > 0:
ax_profile = fig.add_subplot(grids[0, sample_id],
sharey=ax_list[0])
else:
ax_profile = fig.add_subplot(grids[0, sample_id])
ax_profile.set_title(title)
if perGroup:
iterNum2 = hm.matrix.get_num_samples()
else:
iterNum2 = numgroups
for group in range(iterNum2):
if perGroup:
sub_matrix = hm.matrix.get_matrix(sample_id, group)
line_label = sub_matrix['sample']
else:
sub_matrix = hm.matrix.get_matrix(group, sample_id)
line_label = sub_matrix['group']
plot_single(ax_profile,
sub_matrix['matrix'],
averageType,
color_list[group],
line_label,
plot_type='simple')
if sample_id > 0:
plt.setp(ax_profile.get_yticklabels(), visible=False)
if sample_id == 0 and yAxisLabel != '':
ax_profile.set_ylabel(yAxisLabel)
ax_profile.axes.set_xticks(xticks)
ax_profile.axes.set_xticklabels(xtickslabel)
ax_list.append(ax_profile)
# align the first and last label
# such that they don't fall off
# the heatmap sides
ticks = ax_profile.xaxis.get_major_ticks()
ticks[0].label1.set_horizontalalignment('left')
ticks[-1].label1.set_horizontalalignment('right')
# reduce the number of yticks by half
num_ticks = len(ax_list[0].get_yticks())
yticks = [ax_list[0].get_yticks()[i] for i in range(1, num_ticks, 2)]
ax_list[0].set_yticks(yticks)
ax_list[0].set_ylim(yMin, yMax)
if legend_location != 'none':
ax_list[-1].legend(loc=legend_location.replace('-', ' '),
ncol=1,
prop=fontP,
frameon=False,
markerscale=0.5)
if showColorbar:
if showSummaryPlot:
# we dont want to colorbar to extend
# over the profiles row
grid_start = 2
else:
grid_start = 0
ax = fig.add_subplot(grids[grid_start:, -1])
fig.colorbar(img, cax=ax)
plt.subplots_adjust(wspace=0.05,
hspace=0.025,
top=0.85,
bottom=0,
left=0.04,
right=0.96)
plt.savefig(outFileName,
bbox_inches='tight',
pdd_inches=0,
dpi=200,
format=image_format)
