def get_bampaths_4_sample(self):
'''
This will extract tag count matrix from bam files for all samples.
:param bamnames:
:param rna_bam:
:return:
'''
bam_paths = {}
# Check if all the bam file exist
for bam in self.bam_name:
if type(bam) is str:
try:
bam_path = differential_binding.getBam(bam)
bam_paths[bam] = bam_path
print('Found bam path:', bam_path)
except ValueError:
raise (
'Error: Bam file not found in the default locations:',
bam)
else:
for ba in bam[1]:
try:
bam_path = differential_binding.getBam(ba)
bam_paths[ba] = bam_path
print('Found bam path:', bam_path)
except ValueError:
raise (
'Error: Bam file not found in the default locations:',
ba)
return bam_paths
def create_heatmap(self):
column = ['chr', 'start', 'stop', 'GenomicPosition TSS=1250 bp, upstream=5000 bp', 'Next transcript gene name',
'Next transcript strand', 'Next Transcript tss distance', 'summit', 'Next Transcript stable_id']
peakdf = self.filter_peaks()
peakdf.loc[:, 'chr'] = peakdf.loc[:, 'chr'].astype(str)
bam_id = self.sample_id
bam_name = self.name
bam_path = differential_binding.getBam(bam_id, path='/ps/imt/e/Encode_data_all/ENCODE_HL60')
sample_bam = pysam.Samfile(bam_path, "rb")
total_mapped = sample_bam.mapped
print(bam_name+'\t'+str(total_mapped)+'\n')
distribution_df_norm = pd.DataFrame()
#print(peakdf.head())
# check if genome of alignment (UCSC or ENSEMBL) bam
try:
sample_bam.count('9', 99181564, 99181974)
except ValueError:
print('Bam file is UCSC aligned converting coordinates accordingly...')
peakdf.loc[: 'chr'] = 'chr' + peakdf.loc[: 'chr']
pass
for ind, row in peakdf.iterrows(): # reading peaksd
sys.stdout.write("\r%d%%" % ind)
sys.stdout.flush()
strand = row['Next transcript strand']
Chr = str(row['chr'])
start_pos = row['start']
next_tss_distance = row['Next Transcript tss distance']
interval = 100
list_sample_norm = []
start = (start_pos + next_tss_distance) - self.distance
stop = start + interval
if start > 0:
for i in range(0, int((2*self.distance)/100)): # Please set based on distance on one side = s*distance/50
seqcount = sample_bam.count(Chr, start, stop)
list_sample_norm.append((seqcount*(5.*10**6)/total_mapped)) # Normalized count per million
start = stop
stop = start + interval # divide peaks into length of 50 bp
distribution_df_norm = distribution_df_norm.append(pd.Series(list_sample_norm), ignore_index=True)
#if (strand == 1) or (strand == '+'):
# distribution_df_norm = distribution_df_norm.append(pd.Series(list_sample_norm), ignore_index=True)
#elif (strand == -1) or (strand == '-'):
# distribution_df_norm = distribution_df_norm.append(pd.Series(list_sample_norm[::-1]), ignore_index=True)
else:
print('Problem with gene strand information:', row['chr'], '-', row['start'])
sample_bam.close() # closing bam file
print(distribution_df_norm.head())
distribution_df_norm = pd.concat([peakdf[column], distribution_df_norm], axis=1)
distribution_df_norm.to_csv(os.path.join(self.outpath, 'HmapWRTTss_'+bam_name+'_norm.tsv'), header=True, index=True, sep='\t')
return distribution_df_norm
def totaltagCountinPeak(peakscorDF, sampleBam):
'''
This will insert a tagcount column in dataframe for first bam file in the list.
:param peakscorDF:
:param sampleBam:
:return:
'''
bam_path = differential_binding.getBam(sampleBam[0])
sample_bam = pysam.Samfile(bam_path, "rb")
countList = []
#print(peakscorDF.head())
for ind, row in peakscorDF.iterrows():
chr = str(row['chr'])
if 'chr' in chr:
chr = row[3:]
start = row['start']
stop = row['stop']
seqcount = sample_bam.count(chr, start, stop)
countList.append(seqcount)
peakscorDF.insert(5, 'tagcount', pd.Series(countList))
return peakscorDF
def GR_heatmaps_DF_for_peaks(bam_name_list,
peak_df,
region=None,
sort=False,
sort_column=None,
scale_df=True,
sample_name=None,
strength_divide=False,
normFact={}):
'''
Suggestion: Please do not use more then 3 samples at a time.
This function will take a list of bam files path and create a heatmap of genomic region for the provided peak dataset.
:param bam_name_list: A list cantaining names of bam files to be vizualized eg. bam_name_list = ['PRMT6_2_seq6', 'H3K4me3_seq2', 'Sample_K9me3']
:param peak_df: A datframe containing peak information
:param region: Which region to plot (eg. tss, intron, exon, intergenic or None)
:param sort: True if dataframe should be sorted
:param sort_column: If sorted=True; give the column name to be sorted
:return: A dataframe; Column: additive length of genomic regions for all the bam files, Rows: Number of peaks defined in the peak dataframe
'''
region = region.strip()
peak_df = peak_df
if region != 'all':
peak_df = peak_df[
peak_df['GenomicPosition TSS=1250 bp, upstream=5000 bp'] == region]
if region == 'tss': # Reduce peaks based on their distance from TSS
print('Selecting peaks only within +-300bp')
peak_df = peak_df[peak_df['Next Transcript tss distance'] < 300]
peak_df = peak_df[peak_df['Next Transcript tss distance'] > -300]
if len(peak_df) == 0:
raise ValueError('selected region does not contain any peaks')
print(region + ' found in dataframe: ', len(peak_df))
# print peak_df.head()
peak_df.index = range(0, len(peak_df))
if sort:
print('Dataframe is being sort...')
colnames = peak_df.columns.tolist()
indices1 = [i for i, s in enumerate(colnames) if sort_column in s]
#print peak_df.head()
for i in indices1:
if "RA" not in colnames[
i] and "RA" not in sort_column and "norm" not in colnames[
i]:
condition = colnames[i]
print('Sorted on column: ' + condition)
peak_df = peak_df.sort(condition, ascending=False)
sort_column = condition
break
elif "RA" in colnames[
i] and "RA" in sort_column and "norm" not in colnames[i]:
condition = colnames[i]
print('Sorted on column: ' + condition)
peak_df = peak_df.sort(condition, ascending=False)
sort_column = condition
break
#print peak_df.head()
bam_order = ','.join(bam_name_list)
if not sample_name is None:
path = make_dir(bam_order,
region + str(len(peak_df)) + '_' + sample_name)
else:
path = make_dir(bam_order, region + str(len(peak_df)))
# print peak_df.head()
big_df = pd.DataFrame()
big_df_raw = pd.DataFrame()
## Find if all bam names are found in database
bam_path = get_bampaths_4_sample(bam_name_list)
for v in bam_name_list:
print('Sample:' + v)
bam_path = differential_binding.getBam(v)
df, df1 = overlapping_peaks_distribution(v, bam_path, peak_df, path)
if v in normFact.keys():
print('Multiply with external norm fact.', v, normFact.get(v))
df1 = df1.multiply(normFact.get(v))
if scale_df:
df = scale_dataframe(df) # scaling of dataframe
print('scaled df')
big_df = pd.concat([big_df, df], axis=1)
big_df_raw = pd.concat([big_df_raw, df1], axis=1)
big_df.columns = range(0, big_df.shape[1])
big_df_raw.columns = range(0, big_df_raw.shape[1])
#print(big_df.head())
# Plot all sample in one line plot
plot_all_peaks_4_multiple_samples(big_df, bam_order, path, 'norm')
plot_all_peaks_4_multiple_samples(big_df_raw, bam_order, path, 'raw')
# Plot peaks after dividing them into strength basis
if strength_divide:
DividePeaksInStrength(big_df_raw, bam_order,
path).divide_peaks_in_strength()
DividePeaksInStrength(big_df, bam_order,
path).divide_peaks_in_strength()
# Plot peaks based on K-means clustering
else:
#try:
big_df = kmeans_clustering(big_df, 9,
1000) # performing k-means clustering
big_df_raw = kmeans_clustering(big_df_raw, 9,
1000) # performing k-means clustering
dict_of_df = differential_binding.group_DF(
big_df, 'cluster') # divide df in smaller dfs basis in clustering
dict_of_df_raw = differential_binding.group_DF(big_df_raw, 'cluster')
print(len(dict_of_df))
line_plot_peak_distribution(dict_of_df, bam_order, path,
'norm') # plotting individual clusters
line_plot_peak_distribution(dict_of_df_raw, bam_order, path, 'raw')
print('No. of sample to plot:', len(bam_name_list))
plot_clustered_peaks_4_multiple_samples(
dict_of_df, bam_order, path,
'norm') # plotting cluster for different bam in overlapping plot
plot_clustered_peaks_4_multiple_samples(dict_of_df_raw, bam_order,
path, 'raw')
#except:
#print('Dataframe can not be clustered, scipy error.')
### adding columns to heatmap df
try:
colList = commons.peakdf_columns()[::-1]
for col in colList:
#print(col, peak_df[col])
big_df.insert(0, col, peak_df[col])
big_df_raw.insert(0, col, peak_df[col])
if sort:
big_df.insert(0, sort_column, peak_df[sort_column])
big_df_raw.insert(0, sort_column, peak_df[sort_column])
except:
raise ValueError('Needed columns for peak profile are missing')
#print (big_df.head())
#print (big_df_raw.head())
# adding tagcount column for first bam file
big_df = totaltagCountinPeak(big_df, bam_name_list)
big_df_raw = totaltagCountinPeak(big_df_raw, bam_name_list)
big_df.to_csv(os.path.join(path, 'norm',
'tagcountDF_' + region + '_norm.tsv'),
sep="\t",
encoding='utf-8') # , ignore_index=True
big_df_raw.to_csv(os.path.join(path, 'raw',
'tagcountDF_' + region + '_raw.tsv'),
sep="\t",
encoding='utf-8') # , ignore_index=True
gc.collect()