def makePureSignal(row):
signal = row[0]
control = row[1]
norm_factor = row[2]
bam = BedTool('../H3K27me3/bw/' + signal + '.bam')
bam.genome_coverage(bg=True, output="../H3K27me3/bw/" + signal + ".bdg")
bam_control = BedTool('../H3K27me3/control/' + control + '.bam')
bam_control.genome_coverage(bg=True,
scale=norm_factor,
output="../H3K27me3/control/" + control +
".bdg")
s = BedTool('../H3K27me3/bw/' + signal + '.bdg')
c = BedTool('../H3K27me3/control/' + control + '.bdg')
BedTool().union_bedgraphs(i=[s.fn, c.fn],
output="../H3K27me3/bw/" + signal + "_" +
control + ".bdg")
union = pd.read_csv("../H3K27me3/bw/" + signal + "_" + control + ".bdg",
sep='\t',
index_col=False,
header=None)
diff = union[3] - union[4]
diff = diff.apply(lambda x: 0 if x < 0 else x)
union["diff"] = diff
union.drop([3, 4], axis=1, inplace=True)
without_zero = union[union['diff'] != 0]
os.remove("../H3K27me3/bw/" + signal + "_" + control + ".bdg")
without_zero.to_csv("../H3K27me3/bw/pure_" + signal + ".bdg", sep='\t')
def cov_at_loci(bam, cov_fn):
bed = BedTool(bam)
cov_df = bed.genome_coverage(bg=True).to_dataframe()
# get coverage
inter = {}
for idx, row in cov_df.iterrows():
chrom,s,e,n = row
s = int(s); e = int(e)
if chrom not in inter:
inter[chrom] = [[s,e]]
else:
if s - inter[chrom][-1][1] < 1000:
inter[chrom][-1][1] = e
else:
inter[chrom].append([s,e])
# prepare coverage list
with open(cov_fn,'w') as out_f:
for k,v in inter.items(): # k is chrom, v is list of region pos
for c in v:
s,e = c
cov = cov_df.query('chrom==@k and start >=@s and end <=@e')['name'].mean()
out_f.write('\t'.join([k,str(s),str(e),str(cov)])+'\n')
cov_df = pd.read_csv(cov_fn,sep='\t',header=0,names=['chr','s','e','cov'])
cov_df = cov_df.sort_values('cov',ascending=False)
cov_df.to_csv(cov_fn,sep='\t',index=False)
def make_bg(files):
tfp = files[1]+str(uuid.uuid4())
bamf = BedTool(files[0])
bgf = bamf.genome_coverage(bg=True, strand='+')
bgf.saveas(tfp)
#os.system("cat "+tfp+" | awk '$4 > 3' > "+tfp) # Filter low cov regions
df = pd.read_csv(tfp, sep='\t', header=None, names=['Chr', 'Start', 'End', 'Strand'], index_col=None)
df['Strand'] = '+'
df.to_csv(tfp, sep='\t', index=False, header=None)
tfn = files[1]+str(uuid.uuid4())
bgf = bamf.genome_coverage(bg=True, strand='-')
bgf.saveas(tfn)
#os.system("cat "+tfn+" | awk '$4 > 3' > "+tfn) # Filter low cov regions
df = pd.read_csv(tfn, sep='\t', header=None, names=['Chr', 'Start', 'End', 'Strand'], index_col=None)
df['Strand'] = '-'
df.to_csv(tfn, sep='\t', index=False, header=None)
os.system('cat ' + tfp + ' ' + tfn + ' > ' + files[1]+files[0].split("/")[-1].replace('.bam', '.bg'))
temp = BedTool(files[1]+files[0].split("/")[-1].replace('.bam', '.bg'))
temp = temp.sort()
temp.saveas(files[1]+files[0].split("/")[-1].replace('.bam', '.bg'))
os.system('rm ' + tfp + ' ' + tfn)
return(1)
def plot_coverage(in_bam,
out_plot_file,
plot_format,
plot_data_style,
plot_style,
plot_width,
plot_height,
plot_dpi,
plot_title,
plot_x_limits,
plot_y_limits,
base_q_threshold,
mapping_q_threshold,
max_coverage_depth,
read_length_threshold,
plot_only_non_duplicates=False,
bin_large_plots=False,
binning_summary_statistic="max",
out_summary=None):
'''
Generate a coverage plot from an aligned bam file
'''
samtools = tools.samtools.SamtoolsTool()
# check if in_bam is aligned, if not raise an error
num_mapped_reads = samtools.count(in_bam, opts=["-F", "4"])
if num_mapped_reads == 0:
raise Exception(
"""The bam file specified appears to have zero mapped reads. 'plot_coverage' requires an aligned bam file. You can try 'align_and_plot_coverage' if the plot input bam file contains reads and you don't mind a simple bwa alignment. \n File: %s"""
% in_bam)
if out_summary is None:
coverage_tsv_file = util.file.mkstempfname('.summary.tsv')
else:
coverage_tsv_file = out_summary
bam_dupe_processed = util.file.mkstempfname('.dupe_processed.bam')
if plot_only_non_duplicates:
# TODO: this is probably not necessary since "samtools depth" does not count marked duplicates
# write a new bam file; exclude reads with the 1024 flag set (PCR or optical duplicates)
samtools.view(["-F", "1024", '-@', '3'], in_bam, bam_dupe_processed)
else:
bam_dupe_processed = in_bam
# only sort if not sorted
bam_sorted = util.file.mkstempfname('.sorted.bam')
should_remove_sorted = True
if not util.file.bam_is_sorted(bam_dupe_processed):
samtools.sort(bam_dupe_processed, bam_sorted, args=["-O", "bam"])
if plot_only_non_duplicates:
os.unlink(bam_dupe_processed)
else:
bam_sorted = bam_dupe_processed
if not plot_only_non_duplicates:
# in this case we are passing through the original in_bam directly
should_remove_sorted = False
# call samtools index
samtools.index(bam_sorted)
# call samtools depth
opts = []
opts += ['-aa'] # report coverate at "absolutely all" positions
if base_q_threshold:
if not plot_only_non_duplicates:
# Note: "bedtools genomecov" will count depth including duplicates, but does
# not expose options for filtering by quality. When duplicates
# are excluded, "samtools depth" is used which does support quality filtering
# We use either samtools or bedtools, because the former ignores marked duplicates
# from its depth count while bedtools includes them.
log.warning("'-q' ignored since --plotOnlyNonDuplicates is absent")
opts += ["-q", str(base_q_threshold)]
if mapping_q_threshold:
if not plot_only_non_duplicates:
log.warning("'-Q' ignored since --plotOnlyNonDuplicates is absent")
opts += ["-Q", str(mapping_q_threshold)]
if max_coverage_depth:
if not plot_only_non_duplicates:
log.warning("'-m' ignored since --plotOnlyNonDuplicates is absent")
opts += ["-m", str(max_coverage_depth)]
if read_length_threshold:
if not plot_only_non_duplicates:
log.warning("'-l' ignored since --plotOnlyNonDuplicates is absent")
opts += ["-l", str(read_length_threshold)]
# add option here for bedtools to report coverage w/ duplicates
# (and then samtools for no-dups)
#
# Ex.
# samtools depth -aa mapped-to-ref.with-dups.tmp.bam
# bedtools genomecov -ibam mapped-to-ref.with-dups.tmp.bam -d
if not plot_only_non_duplicates:
bt = BedTool(bam_sorted)
# "d=True" is the equivalent of passing "-d" to the bedtools CLI
bt.genome_coverage(d=True).saveas(coverage_tsv_file)
else:
samtools.depth(bam_sorted, coverage_tsv_file, opts)
# only remove the sorted bam if it is not the original input bam
# which we use directly in some casess
if should_remove_sorted:
os.unlink(bam_sorted)
# ---- create plot based on coverage_tsv_file ----
segment_depths = OrderedDict()
domain_max = 0
with open(coverage_tsv_file, "r") as tabfile:
for row in csv.reader(tabfile, delimiter='\t'):
segment_depths.setdefault(row[0], []).append(float(row[2]))
domain_max += 1
with matplotlib.pyplot.style.context(plot_style):
fig = matplotlib.pyplot.gcf()
DPI = plot_dpi or fig.get_dpi()
fig.set_size_inches(
float(plot_width) / float(DPI),
float(plot_height) / float(DPI))
font_size = (2.5 * plot_height) / float(DPI)
ax = matplotlib.pyplot.subplot(
) # Defines ax variable by creating an empty plot
# Set the tick labels font
for label in (ax.get_xticklabels() + ax.get_yticklabels()):
label.set_fontsize(font_size)
# Binning
bin_size = 1
if bin_large_plots:
# Bin locations and take summary value (maximum or minimum) in each bin
binning_fn = {
"min": min,
"max": max,
"mean": mean,
"median": median
}
binning_action = binning_fn.get(binning_summary_statistic, "max")
inner_plot_width_inches = ax.get_window_extent().transformed(
fig.dpi_scale_trans.inverted()).width
inner_plot_width_px = inner_plot_width_inches * fig.dpi # width of actual plot (sans whitespace and y axis text)
bins_per_pixel = 1 # increase to make smaller (but less visible) bins
bin_size = 1 + int(domain_max /
(inner_plot_width_px * bins_per_pixel))
binned_segment_depths = OrderedDict()
for segment_num, (segment_name, position_depths) in enumerate(
segment_depths.items()):
summary_depths_in_bins = [
binning_action(position_depths[i:i + bin_size])
for i in range(0, len(position_depths), bin_size)
]
binned_segment_depths[segment_name] = summary_depths_in_bins
segment_depths = binned_segment_depths
# Plotting
domain_max = 0
for segment_num, (segment_name, position_depths) in enumerate(
segment_depths.items()):
prior_domain_max = domain_max
domain_max += len(position_depths)
colors = list(
matplotlib.pyplot.rcParams['axes.prop_cycle'].by_key()
['color']) # get the colors for this style
segment_color = colors[
segment_num %
len(colors)] # pick a color, offset by the segment index
x_values = range(prior_domain_max, domain_max)
x_values = [x * bin_size for x in x_values]
if plot_data_style == "filled":
matplotlib.pyplot.fill_between(x_values,
position_depths,
[0] * len(position_depths),
linewidth=0,
antialiased=True,
color=segment_color)
elif plot_data_style == "line":
matplotlib.pyplot.plot(x_values,
position_depths,
antialiased=True,
color=segment_color)
elif plot_data_style == "dots":
matplotlib.pyplot.plot(x_values,
position_depths,
'ro',
antialiased=True,
color=segment_color)
matplotlib.pyplot.title(plot_title, fontsize=font_size * 1.2)
matplotlib.pyplot.xlabel("bp", fontsize=font_size * 1.1)
ylabel = "read depth"
if (bin_size > 1):
ylabel = "read depth ({summary} in {size}-bp bin)".format(
size=bin_size, summary=binning_summary_statistic)
matplotlib.pyplot.ylabel(ylabel, fontsize=font_size * 1.1)
if plot_x_limits is not None:
x_min, x_max = plot_x_limits
matplotlib.pyplot.xlim(x_min, x_max)
if plot_y_limits is not None:
y_min, y_max = plot_y_limits
matplotlib.pyplot.ylim(y_min, y_max)
# to squash a backend renderer error on OSX related to tight layout
if matplotlib.pyplot.get_backend().lower() in ['agg', 'macosx']:
fig.set_tight_layout(True)
else:
fig.tight_layout()
matplotlib.pyplot.savefig(out_plot_file, format=plot_format,
dpi=DPI) #, bbox_inches='tight')
log.info("Coverage plot saved to: " + out_plot_file)
if not out_summary:
os.unlink(coverage_tsv_file)
def plot_coverage(in_bam,
out_plot_file,
plot_format,
plot_data_style,
plot_style,
plot_width,
plot_height,
plot_dpi,
plot_title,
base_q_threshold,
mapping_q_threshold,
max_coverage_depth,
read_length_threshold,
plot_only_non_duplicates=False,
out_summary=None):
'''
Generate a coverage plot from an aligned bam file
'''
# TODO: remove this:
#coverage_tsv_file = "/Users/tomkinsc/Downloads/plottest/test_multisegment.tsv"
samtools = tools.samtools.SamtoolsTool()
# check if in_bam is aligned, if not raise an error
num_mapped_reads = samtools.count(in_bam, opts=["-F", "4"])
if num_mapped_reads == 0:
raise Exception(
"""The bam file specified appears to have zero mapped reads. 'plot_coverage' requires an aligned bam file. You can try 'align_and_plot_coverage' if you don't mind a simple bwa alignment. \n File: %s"""
% in_bam)
if out_summary is None:
coverage_tsv_file = util.file.mkstempfname('.summary.tsv')
else:
coverage_tsv_file = out_summary
bam_dupe_processed = util.file.mkstempfname('.dupe_processed.bam')
if plot_only_non_duplicates:
# TODO: this is probably not necessary since "samtools depth" does not count marked duplicates
# write a new bam file; exclude reads with the 1024 flag set (PCR or optical duplicates)
samtools.view(["-F", "1024"], in_bam, bam_dupe_processed)
else:
bam_dupe_processed = in_bam
# call samtools sort
bam_sorted = util.file.mkstempfname('.sorted.bam')
samtools.sort(bam_dupe_processed, bam_sorted, args=["-O", "bam"])
if plot_only_non_duplicates:
os.unlink(bam_dupe_processed)
# call samtools index
samtools.index(bam_sorted)
# call samtools depth
opts = []
opts += ['-aa'] # report coverate at "absolutely all" positions
if base_q_threshold:
if not plot_only_non_duplicates:
# Note: "bedtools genomecov" will count depth including duplicates, but does
# not expose options for filtering by quality. When duplicates
# are excluded, "samtools depth" is used which does support quality filtering
# We use either samtools or bedtools, because the former ignores marked duplicates
# from its depth count while bedtools includes them.
log.warning("'-q' ignored since --plotOnlyNonDuplicates is absent")
opts += ["-q", str(base_q_threshold)]
if mapping_q_threshold:
if not plot_only_non_duplicates:
log.warning("'-Q' ignored since --plotOnlyNonDuplicates is absent")
opts += ["-Q", str(mapping_q_threshold)]
if max_coverage_depth:
if not plot_only_non_duplicates:
log.warning("'-m' ignored since --plotOnlyNonDuplicates is absent")
opts += ["-m", str(max_coverage_depth)]
if read_length_threshold:
if not plot_only_non_duplicates:
log.warning("'-l' ignored since --plotOnlyNonDuplicates is absent")
opts += ["-l", str(read_length_threshold)]
# add option here for bedtools to report coverage w/ duplicates
# (and then samtools for no-dups)
#
# Ex.
# samtools depth -aa mapped-to-ref.with-dups.tmp.bam
# bedtools genomecov -ibam mapped-to-ref.with-dups.tmp.bam -d
if not plot_only_non_duplicates:
bt = BedTool(bam_sorted)
# "d=True" is the equivalent of passing "-d" to the bedtools CLI
bt.genome_coverage(d=True).saveas(coverage_tsv_file)
else:
samtools.depth(bam_sorted, coverage_tsv_file, opts)
os.unlink(bam_sorted)
# ---- create plot based on coverage_tsv_file ----
segment_depths = OrderedDict()
domain_max = 0
with open(coverage_tsv_file, "r") as tabfile:
for row in csv.reader(tabfile, delimiter='\t'):
segment_depths.setdefault(row[0], []).append(int(row[2]))
domain_max += 1
domain_max = 0
with plt.style.context(plot_style):
fig = plt.gcf()
DPI = plot_dpi or fig.get_dpi()
fig.set_size_inches(
float(plot_width) / float(DPI),
float(plot_height) / float(DPI))
font_size = (2.5 * plot_height) / float(DPI)
ax = plt.subplot() # Defines ax variable by creating an empty plot
# Set the tick labels font
for label in (ax.get_xticklabels() + ax.get_yticklabels()):
label.set_fontsize(font_size)
for segment_num, (segment_name, position_depths) in enumerate(
segment_depths.items()):
prior_domain_max = domain_max
domain_max += len(position_depths)
colors = list(plt.rcParams['axes.prop_cycle'].by_key()
['color']) # get the colors for this style
segment_color = colors[
segment_num %
len(colors)] # pick a color, offset by the segment index
if plot_data_style == "filled":
plt.fill_between(range(prior_domain_max, domain_max),
position_depths, [0] * len(position_depths),
linewidth=0,
antialiased=True,
color=segment_color)
elif plot_data_style == "line":
plt.plot(range(prior_domain_max, domain_max),
position_depths,
antialiased=True,
color=segment_color)
elif plot_data_style == "dots":
plt.plot(range(prior_domain_max, domain_max),
position_depths,
'ro',
antialiased=True,
color=segment_color)
plt.title(plot_title, fontsize=font_size * 1.2)
plt.xlabel("bp", fontsize=font_size * 1.1)
plt.ylabel("read depth", fontsize=font_size * 1.1)
# to squash a backend renderer error on OSX related to tight layout
if plt.get_backend().lower() in ['agg', 'macosx']:
fig.set_tight_layout(True)
else:
fig.tight_layout()
plt.savefig(out_plot_file, format=plot_format,
dpi=DPI) #, bbox_inches='tight')
log.info("Coverage plot saved to: " + out_plot_file)
if not out_summary:
os.unlink(coverage_tsv_file)
