Mercurial > repos > mvdbeek > mismatch_frequencies

comparison mismatch_frequencies.py @ 0:77de5fc623f9 draft

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
planemo upload for repository https://bitbucket.org/drosofff/gedtools/
author mvdbeek
date Wed, 27 May 2015 13:40:23 -0400
parents
children 2974c382105c
comparison
equal deleted inserted replaced
-1:000000000000 0:77de5fc623f9
1 import pysam, re, string
2 import matplotlib.pyplot as plt
3 import pandas as pd
4 import json
5 from collections import defaultdict
6 from collections import OrderedDict
7 import argparse
8 import itertools
9
10 class MismatchFrequencies:
11 '''Iterate over a SAM/BAM alignment file, collecting reads with mismatches. One
12 class instance per alignment file. The result_dict attribute will contain a
13 nested dictionary with name, readlength and mismatch count.'''
14 def __init__(self, result_dict={}, alignment_file=None, name="name", minimal_readlength=21,
15 maximal_readlength=21,
16 number_of_allowed_mismatches=1,
17 ignore_5p_nucleotides=0,
18 ignore_3p_nucleotides=0,
19 possible_mismatches = [
20 'AC', 'AG', 'AT',
21 'CA', 'CG', 'CT',
22 'GA', 'GC', 'GT',
23 'TA', 'TC', 'TG'
24 ]):
25
26 self.result_dict = result_dict
27 self.name = name
28 self.minimal_readlength = minimal_readlength
29 self.maximal_readlength = maximal_readlength
30 self.number_of_allowed_mismatches = number_of_allowed_mismatches
31 self.ignore_5p_nucleotides = ignore_5p_nucleotides
32 self.ignore_3p_nucleotides = ignore_3p_nucleotides
33 self.possible_mismatches = possible_mismatches
34
35 if alignment_file:
36 self.pysam_alignment = pysam.Samfile(alignment_file)
37 self.references = self.pysam_alignment.references #names of fasta reference sequences
38 result_dict[name]=self.get_mismatches(
39 self.pysam_alignment,
40 minimal_readlength,
41 maximal_readlength,
42 possible_mismatches
43 )
44
45 def get_mismatches(self, pysam_alignment, minimal_readlength,
46 maximal_readlength, possible_mismatches):
47 mismatch_dict = defaultdict(int)
48 rec_dd = lambda: defaultdict(rec_dd)
49 len_dict = rec_dd()
50 for alignedread in pysam_alignment:
51 if self.read_is_valid(alignedread, minimal_readlength, maximal_readlength):
52 chromosome = pysam_alignment.getrname(alignedread.rname)
53 try:
54 len_dict[int(alignedread.rlen)][chromosome]['total valid reads'] += 1
55 except TypeError:
56 len_dict[int(alignedread.rlen)][chromosome]['total valid reads'] = 1
57 MD = alignedread.opt('MD')
58 if self.read_has_mismatch(alignedread, self.number_of_allowed_mismatches):
59 (ref_base, mismatch_base)=self.read_to_reference_mismatch(MD, alignedread.seq, alignedread.is_reverse)
60 if ref_base == None:
61 continue
62 else:
63 for i, base in enumerate(ref_base):
64 if not ref_base[i]+mismatch_base[i] in possible_mismatches:
65 continue
66 try:
67 len_dict[int(alignedread.rlen)][chromosome][ref_base[i]+mismatch_base[i]] += 1
68 except TypeError:
69 len_dict[int(alignedread.rlen)][chromosome][ref_base[i]+mismatch_base[i]] = 1
70 return len_dict
71
72 def read_is_valid(self, read, min_readlength, max_readlength):
73 '''Filter out reads that are unmatched, too short or
74 too long or that contian insertions'''
75 if read.is_unmapped:
76 return False
77 if read.rlen < min_readlength:
78 return False
79 if read.rlen > max_readlength:
80 return False
81 else:
82 return True
83
84 def read_has_mismatch(self, read, number_of_allowed_mismatches=1):
85 '''keep only reads with one mismatch. Could be simplified'''
86 NM=read.opt('NM')
87 if NM <1: #filter out reads with no mismatch
88 return False
89 if NM >number_of_allowed_mismatches: #filter out reads with more than 1 mismtach
90 return False
91 else:
92 return True
93
94 def mismatch_in_allowed_region(self, readseq, mismatch_position):
95 '''
96 >>> M = MismatchFrequencies()
97 >>> readseq = 'AAAAAA'
98 >>> mismatch_position = 2
99 >>> M.mismatch_in_allowed_region(readseq, mismatch_position)
100 True
101 >>> M = MismatchFrequencies(ignore_3p_nucleotides=2, ignore_5p_nucleotides=2)
102 >>> readseq = 'AAAAAA'
103 >>> mismatch_position = 1
104 >>> M.mismatch_in_allowed_region(readseq, mismatch_position)
105 False
106 >>> readseq = 'AAAAAA'
107 >>> mismatch_position = 4
108 >>> M.mismatch_in_allowed_region(readseq, mismatch_position)
109 False
110 '''
111 mismatch_position+=1 # To compensate for starting the count at 0
112 five_p = self.ignore_5p_nucleotides
113 three_p = self.ignore_3p_nucleotides
114 if any([five_p > 0, three_p > 0]):
115 if any([mismatch_position <= five_p,
116 mismatch_position >= (len(readseq)+1-three_p)]): #Again compensate for starting the count at 0
117 return False
118 else:
119 return True
120 else:
121 return True
122
123 def read_to_reference_mismatch(self, MD, readseq, is_reverse):
124 '''
125 This is where the magic happens. The MD tag contains SNP and indel information,
126 without looking to the genome sequence. This is a typical MD tag: 3C0G2A6.
127 3 bases of the read align to the reference, followed by a mismatch, where the
128 reference base is C, followed by 10 bases aligned to the reference.
129 suppose a reference 'CTTCGATAATCCTT'
130 ||| || ||||||
131 and a read 'CTTATATTATCCTT'.
132 This situation is represented by the above MD tag.
133 Given MD tag and read sequence this function returns the reference base C, G and A,
134 and the mismatched base A, T, T.
135 >>> M = MismatchFrequencies()
136 >>> MD='3C0G2A7'
137 >>> seq='CTTATATTATCCTT'
138 >>> result=M.read_to_reference_mismatch(MD, seq, is_reverse=False)
139 >>> result[0]=="CGA"
140 True
141 >>> result[1]=="ATT"
142 True
143 >>>
144 '''
145 search=re.finditer('[ATGC]',MD)
146 if '^' in MD:
147 print 'WARNING insertion detected, mismatch calling skipped for this read!!!'
148 return (None, None)
149 start_index=0 # refers to the leading integer of the MD string before an edited base
150 current_position=0 # position of the mismatched nucleotide in the MD tag string
151 mismatch_position=0 # position of edited base in current read
152 reference_base=""
153 mismatched_base=""
154 for result in search:
155 current_position=result.start()
156 mismatch_position=mismatch_position+1+int(MD[start_index:current_position]) #converts the leading characters before an edited base into integers
157 start_index=result.end()
158 reference_base+=MD[result.end()-1]
159 mismatched_base+=readseq[mismatch_position-1]
160 if is_reverse:
161 reference_base=reverseComplement(reference_base)
162 mismatched_base=reverseComplement(mismatched_base)
163 mismatch_position=len(readseq)-mismatch_position-1
164 if mismatched_base=='N':
165 return (None, None)
166 if self.mismatch_in_allowed_region(readseq, mismatch_position):
167 return (reference_base, mismatched_base)
168 else:
169 return (None, None)
170
171 def reverseComplement(sequence):
172 '''do a reverse complement of DNA base.
173 >>> reverseComplement('ATGC')=='GCAT'
174 True
175 >>>
176 '''
177 sequence=sequence.upper()
178 complement = string.maketrans('ATCGN', 'TAGCN')
179 return sequence.upper().translate(complement)[::-1]
180
181 def barplot(df, library, axes):
182 df.plot(kind='bar', ax=axes, subplots=False,\
183 stacked=False, legend='test',\
184 title='Mismatch frequencies for {0}'.format(library))
185
186 def df_to_tab(df, output):
187 df.to_csv(output, sep='\t')
188
189 def reduce_result(df, possible_mismatches):
190 '''takes a pandas dataframe with full mismatch details and
191 summarises the results for plotting.'''
192 alignments = df['Alignment_file'].unique()
193 readlengths = df['Readlength'].unique()
194 combinations = itertools.product(*[alignments, readlengths]) #generate all possible combinations of readlength and alignment files
195 reduced_dict = {}
196 frames = []
197 last_column = 3+len(possible_mismatches)
198 for combination in combinations:
199 library_subset = df[df['Alignment_file'] == combination[0]]
200 library_readlength_subset = library_subset[library_subset['Readlength'] == combination[1]]
201 sum_of_library_and_readlength = library_readlength_subset.iloc[:,3:last_column+1].sum()
202 if not reduced_dict.has_key(combination[0]):
203 reduced_dict[combination[0]] = {}
204 reduced_dict[combination[0]][combination[1]] = sum_of_library_and_readlength.to_dict()
205 return reduced_dict
206
207 def plot_result(reduced_dict, args):
208 names=reduced_dict.keys()
209 nrows=len(names)/2+1
210 fig = plt.figure(figsize=(16,32))
211 for i,library in enumerate (names):
212 axes=fig.add_subplot(nrows,2,i+1)
213 library_dict=reduced_dict[library]
214 df=pd.DataFrame(library_dict)
215 df.drop(['total aligned reads'], inplace=True)
216 barplot(df, library, axes),
217 axes.set_ylabel('Mismatch count / all valid reads * readlength')
218 fig.savefig(args.output_pdf, format='pdf')
219
220 def format_result_dict(result_dict, chromosomes, possible_mismatches):
221 '''Turn nested dictionary into preformatted tab seperated lines'''
222 header = "Reference sequence\tAlignment_file\tReadlength\t" + "\t".join(
223 possible_mismatches) + "\ttotal aligned reads"
224 libraries = result_dict.keys()
225 readlengths = result_dict[libraries[0]].keys()
226 result = []
227 for chromosome in chromosomes:
228 for library in libraries:
229 for readlength in readlengths:
230 line = []
231 line.extend([chromosome, library, readlength])
232 try:
233 line.extend([result_dict[library][readlength][chromosome].get(mismatch, 0) for mismatch in possible_mismatches])
234 line.extend([result_dict[library][readlength][chromosome].get(u'total valid reads', 0)])
235 except KeyError:
236 line.extend([0 for mismatch in possible_mismatches])
237 line.extend([0])
238 result.append(line)
239 df = pd.DataFrame(result, columns=header.split('\t'))
240 last_column=3+len(possible_mismatches)
241 df['mismatches/per aligned nucleotides'] = df.iloc[:,3:last_column].sum(1)/(df.iloc[:,last_column]*df['Readlength'])
242 return df
243
244 def setup_MismatchFrequencies(args):
245 resultDict=OrderedDict()
246 kw_list=[{'result_dict' : resultDict,
247 'alignment_file' :alignment_file,
248 'name' : name,
249 'minimal_readlength' : args.min,
250 'maximal_readlength' : args.max,
251 'number_of_allowed_mismatches' : args.n_mm,
252 'ignore_5p_nucleotides' : args.five_p,
253 'ignore_3p_nucleotides' : args.three_p,
254 'possible_mismatches' : args.possible_mismatches }
255 for alignment_file, name in zip(args.input, args.name)]
256 return (kw_list, resultDict)
257
258 def nested_dict_to_df(dictionary):
259 dictionary = {(outerKey, innerKey): values for outerKey, innerDict in dictionary.iteritems() for innerKey, values in innerDict.iteritems()}
260 df=pd.DataFrame.from_dict(dictionary).transpose()
261 df.index.names = ['Library', 'Readlength']
262 return df
263
264 def run_MismatchFrequencies(args):
265 kw_list, resultDict=setup_MismatchFrequencies(args)
266 references = [MismatchFrequencies(**kw_dict).references for kw_dict in kw_list]
267 return (resultDict, references[0])
268
269 def main():
270 result_dict, references = run_MismatchFrequencies(args)
271 df = format_result_dict(result_dict, references, args.possible_mismatches)
272 reduced_dict = reduce_result(df, args.possible_mismatches)
273 plot_result(reduced_dict, args)
274 reduced_df = nested_dict_to_df(reduced_dict)
275 df_to_tab(reduced_df, args.output_tab)
276 if not args.expanded_output_tab == None:
277 df_to_tab(df, args.expanded_output_tab)
278 return reduced_dict
279
280 if __name__ == "__main__":
281
282 parser = argparse.ArgumentParser(description='Produce mismatch statistics for BAM/SAM alignment files.')
283 parser.add_argument('--input', nargs='*', help='Input files in SAM/BAM format')
284 parser.add_argument('--name', nargs='*', help='Name for input file to display in output file. Should have same length as the number of inputs')
285 parser.add_argument('--output_pdf', help='Output filename for graph')
286 parser.add_argument('--output_tab', help='Output filename for table')
287 parser.add_argument('--expanded_output_tab', default=None, help='Output filename for table')
288 parser.add_argument('--possible_mismatches', default=[
289 'AC', 'AG', 'AT','CA', 'CG', 'CT', 'GA', 'GC', 'GT', 'TA', 'TC', 'TG'
290 ], nargs='+', help='specify mismatches that should be counted for the mismatch frequency. The format is Reference base -> observed base, eg AG for A to G mismatches.')
291 parser.add_argument('--min', '--minimal_readlength', type=int, help='minimum readlength')
292 parser.add_argument('--max', '--maximal_readlength', type=int, help='maximum readlength')
293 parser.add_argument('--n_mm', '--number_allowed_mismatches', type=int, default=1, help='discard reads with more than n mismatches')
294 parser.add_argument('--five_p', '--ignore_5p_nucleotides', type=int, default=0, help='when calculating nucleotide mismatch frequencies ignore the first N nucleotides of the read')
295 parser.add_argument('--three_p', '--ignore_3p_nucleotides', type=int, default=1, help='when calculating nucleotide mismatch frequencies ignore the last N nucleotides of the read')
296 #args = parser.parse_args(['--input', '3mismatches_ago2ip_s2.bam', '3mismatches_ago2ip_ovary.bam','--possible_mismatches','AC','AG', 'CG', 'TG', 'CT','--name', 'Siomi1', 'Siomi2' , '--five_p', '3','--three_p','3','--output_pdf', 'out.pdf', '--output_tab', 'out.tab', '--expanded_output_tab', 'expanded.tab', '--min', '20', '--max', '22'])
297 args = parser.parse_args()
298 reduced_dict = main()
299
300