Mercurial > repos > greg > vsnp_add_zero_coverage

comparison vsnp_add_zero_coverage.py @ 2:01312f8a6ca9 draft

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Uploaded
author greg
date Sun, 03 Jan 2021 16:29:00 +0000
parents 3cb0bf7e1b2d
children
comparison
equal deleted inserted replaced
1:eaf4c304fd22 2:01312f8a6ca9
1 #!/usr/bin/env python 1 #!/usr/bin/env python
2 2
3 import argparse 3 import argparse
4 import multiprocessing
5 import os 4 import os
5 import re
6 import shutil
7
6 import pandas 8 import pandas
7 import pysam 9 import pysam
8 import queue
9 import re
10 import shutil
11 from numpy import mean
12 from Bio import SeqIO 10 from Bio import SeqIO
13 11
14 INPUT_BAM_DIR = 'input_bam_dir'
15 INPUT_VCF_DIR = 'input_vcf_dir'
16 OUTPUT_VCF_DIR = 'output_vcf_dir'
17 OUTPUT_METRICS_DIR = 'output_metrics_dir'
18 12
19 13 def get_sample_name(file_path):
20 def get_base_file_name(file_path):
21 base_file_name = os.path.basename(file_path) 14 base_file_name = os.path.basename(file_path)
22 if base_file_name.find(".") > 0: 15 if base_file_name.find(".") > 0:
23 # Eliminate the extension. 16 # Eliminate the extension.
24 return os.path.splitext(base_file_name)[0] 17 return os.path.splitext(base_file_name)[0]
25 elif base_file_name.find("_") > 0: 18 return base_file_name
26 # The dot extension was likely changed to
27 # the " character.
28 items = base_file_name.split("_")
29 return "_".join(items[0:-1])
30 else:
31 return base_file_name
32 19
33 20
34 def get_coverage_and_snp_count(task_queue, reference, output_metrics, output_vcf, timeout): 21 def get_coverage_df(bam_file):
35 while True: 22 # Create a coverage dictionary.
36 try: 23 coverage_dict = {}
37 tup = task_queue.get(block=True, timeout=timeout) 24 coverage_list = pysam.depth(bam_file, split_lines=True)
38 except queue.Empty: 25 for line in coverage_list:
39 break 26 chrom, position, depth = line.split('\t')
40 bam_file, vcf_file = tup 27 coverage_dict["%s-%s" % (chrom, position)] = depth
41 # Create a coverage dictionary. 28 # Convert it to a data frame.
42 coverage_dict = {} 29 coverage_df = pandas.DataFrame.from_dict(coverage_dict, orient='index', columns=["depth"])
43 coverage_list = pysam.depth(bam_file, split_lines=True) 30 return coverage_df
44 for line in coverage_list:
45 chrom, position, depth = line.split('\t')
46 coverage_dict["%s-%s" % (chrom, position)] = depth
47 # Convert it to a data frame.
48 coverage_df = pandas.DataFrame.from_dict(coverage_dict, orient='index', columns=["depth"])
49 # Create a zero coverage dictionary.
50 zero_dict = {}
51 for record in SeqIO.parse(reference, "fasta"):
52 chrom = record.id
53 total_len = len(record.seq)
54 for pos in list(range(1, total_len + 1)):
55 zero_dict["%s-%s" % (str(chrom), str(pos))] = 0
56 # Convert it to a data frame with depth_x
57 # and depth_y columns - index is NaN.
58 zero_df = pandas.DataFrame.from_dict(zero_dict, orient='index', columns=["depth"])
59 coverage_df = zero_df.merge(coverage_df, left_index=True, right_index=True, how='outer')
60 # depth_x "0" column no longer needed.
61 coverage_df = coverage_df.drop(columns=['depth_x'])
62 coverage_df = coverage_df.rename(columns={'depth_y': 'depth'})
63 # Covert the NaN to 0 coverage and get some metrics.
64 coverage_df = coverage_df.fillna(0)
65 coverage_df['depth'] = coverage_df['depth'].apply(int)
66 total_length = len(coverage_df)
67 average_coverage = coverage_df['depth'].mean()
68 zero_df = coverage_df[coverage_df['depth'] == 0]
69 total_zero_coverage = len(zero_df)
70 total_coverage = total_length - total_zero_coverage
71 genome_coverage = "{:.2%}".format(total_coverage / total_length)
72 # Process the associated VCF input.
73 column_names = ["CHROM", "POS", "ID", "REF", "ALT", "QUAL", "FILTER", "INFO", "FORMAT", "Sample"]
74 vcf_df = pandas.read_csv(vcf_file, sep='\t', header=None, names=column_names, comment='#')
75 good_snp_count = len(vcf_df[(vcf_df['ALT'].str.len() == 1) & (vcf_df['REF'].str.len() == 1) & (vcf_df['QUAL'] > 150)])
76 base_file_name = get_base_file_name(vcf_file)
77 if total_zero_coverage > 0:
78 header_file = "%s_header.csv" % base_file_name
79 with open(header_file, 'w') as outfile:
80 with open(vcf_file) as infile:
81 for line in infile:
82 if re.search('^#', line):
83 outfile.write("%s" % line)
84 vcf_df_snp = vcf_df[vcf_df['REF'].str.len() == 1]
85 vcf_df_snp = vcf_df_snp[vcf_df_snp['ALT'].str.len() == 1]
86 vcf_df_snp['ABS_VALUE'] = vcf_df_snp['CHROM'].map(str) + "-" + vcf_df_snp['POS'].map(str)
87 vcf_df_snp = vcf_df_snp.set_index('ABS_VALUE')
88 cat_df = pandas.concat([vcf_df_snp, zero_df], axis=1, sort=False)
89 cat_df = cat_df.drop(columns=['CHROM', 'POS', 'depth'])
90 cat_df[['ID', 'ALT', 'QUAL', 'FILTER', 'INFO']] = cat_df[['ID', 'ALT', 'QUAL', 'FILTER', 'INFO']].fillna('.')
91 cat_df['REF'] = cat_df['REF'].fillna('N')
92 cat_df['FORMAT'] = cat_df['FORMAT'].fillna('GT')
93 cat_df['Sample'] = cat_df['Sample'].fillna('./.')
94 cat_df['temp'] = cat_df.index.str.rsplit('-', n=1)
95 cat_df[['CHROM', 'POS']] = pandas.DataFrame(cat_df.temp.values.tolist(), index=cat_df.index)
96 cat_df = cat_df[['CHROM', 'POS', 'ID', 'REF', 'ALT', 'QUAL', 'FILTER', 'INFO', 'FORMAT', 'Sample']]
97 cat_df['POS'] = cat_df['POS'].astype(int)
98 cat_df = cat_df.sort_values(['CHROM', 'POS'])
99 body_file = "%s_body.csv" % base_file_name
100 cat_df.to_csv(body_file, sep='\t', header=False, index=False)
101 if output_vcf is None:
102 output_vcf_file = os.path.join(OUTPUT_VCF_DIR, "%s.vcf" % base_file_name)
103 else:
104 output_vcf_file = output_vcf
105 with open(output_vcf_file, "w") as outfile:
106 for cf in [header_file, body_file]:
107 with open(cf, "r") as infile:
108 for line in infile:
109 outfile.write("%s" % line)
110 else:
111 if output_vcf is None:
112 output_vcf_file = os.path.join(OUTPUT_VCF_DIR, "%s.vcf" % base_file_name)
113 else:
114 output_vcf_file = output_vcf
115 shutil.copyfile(vcf_file, output_vcf_file)
116 bam_metrics = [base_file_name, "", "%4f" % average_coverage, genome_coverage]
117 vcf_metrics = [base_file_name, str(good_snp_count), "", ""]
118 if output_metrics is None:
119 output_metrics_file = os.path.join(OUTPUT_METRICS_DIR, "%s.tabular" % base_file_name)
120 else:
121 output_metrics_file = output_metrics
122 metrics_columns = ["File", "Number of Good SNPs", "Average Coverage", "Genome Coverage"]
123 with open(output_metrics_file, "w") as fh:
124 fh.write("# %s\n" % "\t".join(metrics_columns))
125 fh.write("%s\n" % "\t".join(bam_metrics))
126 fh.write("%s\n" % "\t".join(vcf_metrics))
127 task_queue.task_done()
128 31
129 32
130 def set_num_cpus(num_files, processes): 33 def get_zero_df(reference):
131 num_cpus = int(multiprocessing.cpu_count()) 34 # Create a zero coverage dictionary.
132 if num_files < num_cpus and num_files < processes: 35 zero_dict = {}
133 return num_files 36 for record in SeqIO.parse(reference, "fasta"):
134 if num_cpus < processes: 37 chrom = record.id
135 half_cpus = int(num_cpus / 2) 38 total_len = len(record.seq)
136 if num_files < half_cpus: 39 for pos in list(range(1, total_len + 1)):
137 return num_files 40 zero_dict["%s-%s" % (str(chrom), str(pos))] = 0
138 return half_cpus 41 # Convert it to a data frame with depth_x
139 return processes 42 # and depth_y columns - index is NaN.
43 zero_df = pandas.DataFrame.from_dict(zero_dict, orient='index', columns=["depth"])
44 return zero_df
45
46
47 def output_zc_vcf_file(base_file_name, vcf_file, zero_df, total_zero_coverage, output_vcf):
48 column_names = ["CHROM", "POS", "ID", "REF", "ALT", "QUAL", "FILTER", "INFO", "FORMAT", "Sample"]
49 vcf_df = pandas.read_csv(vcf_file, sep='\t', header=None, names=column_names, comment='#')
50 good_snp_count = len(vcf_df[(vcf_df['ALT'].str.len() == 1) & (vcf_df['REF'].str.len() == 1) & (vcf_df['QUAL'] > 150)])
51 if total_zero_coverage > 0:
52 header_file = "%s_header.csv" % base_file_name
53 with open(header_file, 'w') as outfile:
54 with open(vcf_file) as infile:
55 for line in infile:
56 if re.search('^#', line):
57 outfile.write("%s" % line)
58 vcf_df_snp = vcf_df[vcf_df['REF'].str.len() == 1]
59 vcf_df_snp = vcf_df_snp[vcf_df_snp['ALT'].str.len() == 1]
60 vcf_df_snp['ABS_VALUE'] = vcf_df_snp['CHROM'].map(str) + "-" + vcf_df_snp['POS'].map(str)
61 vcf_df_snp = vcf_df_snp.set_index('ABS_VALUE')
62 cat_df = pandas.concat([vcf_df_snp, zero_df], axis=1, sort=False)
63 cat_df = cat_df.drop(columns=['CHROM', 'POS', 'depth'])
64 cat_df[['ID', 'ALT', 'QUAL', 'FILTER', 'INFO']] = cat_df[['ID', 'ALT', 'QUAL', 'FILTER', 'INFO']].fillna('.')
65 cat_df['REF'] = cat_df['REF'].fillna('N')
66 cat_df['FORMAT'] = cat_df['FORMAT'].fillna('GT')
67 cat_df['Sample'] = cat_df['Sample'].fillna('./.')
68 cat_df['temp'] = cat_df.index.str.rsplit('-', n=1)
69 cat_df[['CHROM', 'POS']] = pandas.DataFrame(cat_df.temp.values.tolist(), index=cat_df.index)
70 cat_df = cat_df[['CHROM', 'POS', 'ID', 'REF', 'ALT', 'QUAL', 'FILTER', 'INFO', 'FORMAT', 'Sample']]
71 cat_df['POS'] = cat_df['POS'].astype(int)
72 cat_df = cat_df.sort_values(['CHROM', 'POS'])
73 body_file = "%s_body.csv" % base_file_name
74 cat_df.to_csv(body_file, sep='\t', header=False, index=False)
75 with open(output_vcf, "w") as outfile:
76 for cf in [header_file, body_file]:
77 with open(cf, "r") as infile:
78 for line in infile:
79 outfile.write("%s" % line)
80 else:
81 shutil.move(vcf_file, output_vcf)
82 return good_snp_count
83
84
85 def output_metrics_file(base_file_name, average_coverage, genome_coverage, good_snp_count, output_metrics):
86 bam_metrics = [base_file_name, "", "%4f" % average_coverage, genome_coverage]
87 vcf_metrics = [base_file_name, str(good_snp_count), "", ""]
88 metrics_columns = ["File", "Number of Good SNPs", "Average Coverage", "Genome Coverage"]
89 with open(output_metrics, "w") as fh:
90 fh.write("# %s\n" % "\t".join(metrics_columns))
91 fh.write("%s\n" % "\t".join(bam_metrics))
92 fh.write("%s\n" % "\t".join(vcf_metrics))
93
94
95 def output_files(vcf_file, total_zero_coverage, zero_df, output_vcf, average_coverage, genome_coverage, output_metrics):
96 base_file_name = get_sample_name(vcf_file)
97 good_snp_count = output_zc_vcf_file(base_file_name, vcf_file, zero_df, total_zero_coverage, output_vcf)
98 output_metrics_file(base_file_name, average_coverage, genome_coverage, good_snp_count, output_metrics)
99
100
101 def get_coverage_and_snp_count(bam_file, vcf_file, reference, output_metrics, output_vcf):
102 coverage_df = get_coverage_df(bam_file)
103 zero_df = get_zero_df(reference)
104 coverage_df = zero_df.merge(coverage_df, left_index=True, right_index=True, how='outer')
105 # depth_x "0" column no longer needed.
106 coverage_df = coverage_df.drop(columns=['depth_x'])
107 coverage_df = coverage_df.rename(columns={'depth_y': 'depth'})
108 # Covert the NaN to 0 coverage and get some metrics.
109 coverage_df = coverage_df.fillna(0)
110 coverage_df['depth'] = coverage_df['depth'].apply(int)
111 total_length = len(coverage_df)
112 average_coverage = coverage_df['depth'].mean()
113 zero_df = coverage_df[coverage_df['depth'] == 0]
114 total_zero_coverage = len(zero_df)
115 total_coverage = total_length - total_zero_coverage
116 genome_coverage = "{:.2%}".format(total_coverage / total_length)
117 # Output a zero-coverage vcf fil and the metrics file.
118 output_files(vcf_file, total_zero_coverage, zero_df, output_vcf, average_coverage, genome_coverage, output_metrics)
140 119
141 120
142 if __name__ == '__main__': 121 if __name__ == '__main__':
143 parser = argparse.ArgumentParser() 122 parser = argparse.ArgumentParser()
144 123
124 parser.add_argument('--bam_input', action='store', dest='bam_input', help='bam input file')
145 parser.add_argument('--output_metrics', action='store', dest='output_metrics', required=False, default=None, help='Output metrics text file') 125 parser.add_argument('--output_metrics', action='store', dest='output_metrics', required=False, default=None, help='Output metrics text file')
146 parser.add_argument('--output_vcf', action='store', dest='output_vcf', required=False, default=None, help='Output VCF file') 126 parser.add_argument('--output_vcf', action='store', dest='output_vcf', required=False, default=None, help='Output VCF file')
147 parser.add_argument('--reference', action='store', dest='reference', help='Reference dataset') 127 parser.add_argument('--reference', action='store', dest='reference', help='Reference dataset')
148 parser.add_argument('--processes', action='store', dest='processes', type=int, help='User-selected number of processes to use for job splitting') 128 parser.add_argument('--vcf_input', action='store', dest='vcf_input', help='vcf input file')
149 129
150 args = parser.parse_args() 130 args = parser.parse_args()
151 131
152 # The assumption here is that the list of files 132 get_coverage_and_snp_count(args.bam_input, args.vcf_input, args.reference, args.output_metrics, args.output_vcf)
153 # in both INPUT_BAM_DIR and INPUT_VCF_DIR are
154 # equal in number and named such that they are
155 # properly matched if the directories contain
156 # more than 1 file (i.e., hopefully the bam file
157 # names and vcf file names will be something like
158 # Mbovis-01D6_* so they can be # sorted and properly
159 # associated with each other).
160 bam_files = []
161 for file_name in sorted(os.listdir(INPUT_BAM_DIR)):
162 file_path = os.path.abspath(os.path.join(INPUT_BAM_DIR, file_name))
163 bam_files.append(file_path)
164 vcf_files = []
165 for file_name in sorted(os.listdir(INPUT_VCF_DIR)):
166 file_path = os.path.abspath(os.path.join(INPUT_VCF_DIR, file_name))
167 vcf_files.append(file_path)
168
169 multiprocessing.set_start_method('spawn')
170 queue1 = multiprocessing.JoinableQueue()
171 num_files = len(bam_files)
172 cpus = set_num_cpus(num_files, args.processes)
173 # Set a timeout for get()s in the queue.
174 timeout = 0.05
175
176 # Add each associated bam and vcf file pair to the queue.
177 for i, bam_file in enumerate(bam_files):
178 vcf_file = vcf_files[i]
179 queue1.put((bam_file, vcf_file))
180
181 # Complete the get_coverage_and_snp_count task.
182 processes = [multiprocessing.Process(target=get_coverage_and_snp_count, args=(queue1, args.reference, args.output_metrics, args.output_vcf, timeout, )) for _ in range(cpus)]
183 for p in processes:
184 p.start()
185 for p in processes:
186 p.join()
187 queue1.join()
188
189 if queue1.empty():
190 queue1.close()
191 queue1.join_thread()