allelefinder.py 9.16 KB
Newer Older
jhoogenboom's avatar
jhoogenboom committed
1
2
#!/usr/bin/env python
"""
3
4
Find true alleles in reference samples and detect possible
contaminations.
jhoogenboom's avatar
jhoogenboom committed
5
6
7

In each sample, the sequences with the highest read counts of each
marker are called alleles, with a user-defined maximum number of alleles
jhoogenboom's avatar
jhoogenboom committed
8
per marker.  The allele balance is kept within given bounds.  If the
jhoogenboom's avatar
jhoogenboom committed
9
10
11
highest non-allelic sequence exceeds a given limit, no alleles are
called for this marker.  If this happens for multiple markers in one
sample, no alleles are called for this sample at all.
jhoogenboom's avatar
jhoogenboom committed
12

jhoogenboom's avatar
jhoogenboom committed
13
14
15
16
17
18
19
The allele list obtained from allelefinder should always be checked
carefully before using it as the input of various other tools operating
on reference samples.  These tools rely heavily on the correctness of
this file to do their job.  One may use the allelefinder report
(-R/--report output argument) and the blame tool to get a quick overview
of what might be wrong.
"""
20
from ..lib import pos_int_arg, add_input_output_args, get_input_output_files, \
21
                  ensure_sequence_format, get_sample_data, \
22
                  add_sequence_format_args
jhoogenboom's avatar
jhoogenboom committed
23
24
25
26

__version__ = "0.1dev"


27
28
29
30
31
32
33
34
35
# Default values for parameters are specified below.

# Default minimum number of reads required for the highest allele.
# This value can be overridden by the -n command line option.
_DEF_MIN_READS = 50

# Default minimum number of reads required for an allele to be called,
# as a percentage of the number of reads of the highest allele.
# This value can be overridden by the -m command line option.
jhoogenboom's avatar
jhoogenboom committed
36
_DEF_MIN_ALLELE_PCT = 30.0
37
38
39
40
41
42

# Default maximum amount of noise to allow, as a percentage of the
# number of reads of the highest allele of each marker.  If any noise
# (i.e., non-allelic sequences) above this threshold are detected, the
# sample is considered 'noisy' for this marker.
# This value can be overridden by the -M command line option.
jhoogenboom's avatar
jhoogenboom committed
43
_DEF_MAX_NOISE_PCT = 10.0
44
45
46
47
48
49

# Default maximum number of noisy markers allowed per sample.
# This value can be overridden by the -x command line option.
_DEF_MAX_NOISY = 2


50
51
52
def find_alleles(samples_in, outfile, reportfile, min_reads, min_allele_pct,
                 max_noise_pct, max_alleles, max_noisy, stuttermark_column,
                 seqformat, library):
53
    library = library if library is not None else {}
54

jhoogenboom's avatar
jhoogenboom committed
55
    outfile.write("\t".join(["sample", "marker", "total", "allele"]) + "\n")
56
57
    allelelist = {}
    get_sample_data(
58
        samples_in,
59
60
        lambda tag, data: find_alleles_sample(
            data if stuttermark_column is None
jhoogenboom's avatar
jhoogenboom committed
61
62
                 else {key: data[key] for key in data if key[0] in
                       allelelist[tag]},
jhoogenboom's avatar
jhoogenboom committed
63
            outfile, reportfile, tag, min_reads, min_allele_pct, max_noise_pct,
64
65
66
            max_alleles, max_noisy, seqformat, library),
        allelelist,
        stuttermark_column)
67
68
69
#find_alleles


jhoogenboom's avatar
jhoogenboom committed
70
71
72
def find_alleles_sample(data, outfile, reportfile, tag, min_reads,
                        min_allele_pct, max_noise_pct, max_alleles, max_noisy,
                        seqformat, library):
jhoogenboom's avatar
jhoogenboom committed
73
74
75
76
    top_noise = {}
    top_allele = {}
    alleles = {}
    for marker, allele in data:
77
        reads = sum(data[marker, allele])
jhoogenboom's avatar
jhoogenboom committed
78

79
80
81
82
83
84
        if marker not in alleles:
            alleles[marker] = {allele: reads}
            top_allele[marker] = reads
            top_noise[marker] = ["-", 0]
        else:
            if reads > top_allele[marker]:
jhoogenboom's avatar
jhoogenboom committed
85
                # New highest allele!
86
87
88
89
90
91
92
93
94
95
                top_allele[marker] = reads
                for allelex in alleles[marker].keys():
                    if (alleles[marker][allelex] <
                            top_allele[marker] * (min_allele_pct/100.)):
                        if alleles[marker][allelex] > top_noise[marker][1]:
                            top_noise[marker] = [
                                allelex, alleles[marker][allelex]]
                        del alleles[marker][allelex]
                alleles[marker][allele] = reads
            elif reads >= top_allele[marker]*(min_allele_pct/100.):
jhoogenboom's avatar
jhoogenboom committed
96
97
                # New secundary allele!
                alleles[marker][allele] = reads
98
            elif reads >= top_noise[marker][1]:
jhoogenboom's avatar
jhoogenboom committed
99
                # New highest noise!
100
                top_noise[marker] = [allele, reads]
jhoogenboom's avatar
jhoogenboom committed
101

102
103
    # Find and eliminate noisy markers in this sample first.
    noisy_markers = 0
jhoogenboom's avatar
jhoogenboom committed
104
    for marker in alleles:
105
        if top_allele[marker] < min_reads:
jhoogenboom's avatar
jhoogenboom committed
106
107
108
109
            reportfile.write(
                "Sample %s is not suitable for marker %s:\n"
                "highest allele has only %i reads\n\n" %
                    (tag, marker, top_allele[marker]))
110
111
            alleles[marker] = {}
            continue
112
113
        expect = get_max_expected_alleles(max_alleles, marker, library)
        if len(alleles[marker]) > expect:
jhoogenboom's avatar
jhoogenboom committed
114
115
            allele_order = sorted(alleles[marker],
                                  key=lambda x: -alleles[marker][x])
116
117
            top_noise[marker] = [allele_order[expect],
                alleles[marker][allele_order[expect]]]
jhoogenboom's avatar
jhoogenboom committed
118
            alleles[marker] = {x: alleles[marker][x]
119
                               for x in allele_order[:expect]}
120
        if top_noise[marker][1] > top_allele[marker]*(max_noise_pct/100.):
jhoogenboom's avatar
jhoogenboom committed
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
            reportfile.write(
                "Sample %s is not suitable for marker %s:\n"
                "highest non-allele is %.1f%% of the highest allele\n" %
                (tag, marker, 100.*top_noise[marker][1]/top_allele[marker]))
            for allele in sorted(alleles[marker],
                                 key=lambda x: -alleles[marker][x]):
                seq = allele if seqformat is None \
                    else ensure_sequence_format(allele, seqformat,
                        library=library, marker=marker)
                reportfile.write("%i\tALLELE\t%s\n" %
                    (alleles[marker][allele], seq))
            seq = top_noise[marker][0] if seqformat is None \
                else ensure_sequence_format(top_noise[marker][0],
                    seqformat, library=library, marker=marker)
            reportfile.write("%i\tNOISE\t%s\n\n" % (top_noise[marker][1], seq))
136
137
138
139
140
            noisy_markers += 1
            alleles[marker] = {}

    # Drop this sample completely if it has too many noisy markers.
    if noisy_markers > max_noisy:
jhoogenboom's avatar
jhoogenboom committed
141
        reportfile.write("Sample %s appears to be contaminated!\n\n" % tag)
142
143
144
145
146
147
        return

    # The sample is OK, write out its alleles.
    for marker in alleles:
        for allele in sorted(alleles[marker],
                             key=lambda x: -alleles[marker][x]):
jhoogenboom's avatar
jhoogenboom committed
148
149
            seq = allele if seqformat is None else ensure_sequence_format(
                allele, seqformat, library=library, marker=marker)
jhoogenboom's avatar
jhoogenboom committed
150
151
            outfile.write("\t".join(
                [tag, marker, str(alleles[marker][allele]), seq]) + "\n")
152
#find_alleles_sample
jhoogenboom's avatar
jhoogenboom committed
153
154


155
156
157
158
159
160
161
162
163
def get_max_expected_alleles(max_alleles, marker, library):
    if max_alleles is not None:
        return max_alleles
    if "max_expected_copies" in library:
        return library["max_expected_copies"].get(marker, 2)
    return 2
#get_max_expected_alleles


jhoogenboom's avatar
jhoogenboom committed
164
def add_arguments(parser):
165
    add_input_output_args(parser, False, False, True)
jhoogenboom's avatar
jhoogenboom committed
166
167
168
    filtergroup = parser.add_argument_group("filtering options")
    filtergroup.add_argument('-m', '--min-allele-pct', metavar="PCT",
        type=float, default=_DEF_MIN_ALLELE_PCT,
jhoogenboom's avatar
jhoogenboom committed
169
        help="call heterozygous if the second allele is at least this "
jhoogenboom's avatar
jhoogenboom committed
170
171
             "percentage of the highest allele of a marker "
             "(default: %(default)s)")
jhoogenboom's avatar
jhoogenboom committed
172
173
    filtergroup.add_argument('-M', '--max-noise-pct', metavar="PCT",
        type=float, default=_DEF_MAX_NOISE_PCT,
174
175
        help="a sample is considered contaminated/unsuitable for a marker if "
             "the highest non-allelic sequence is at least this percentage of "
jhoogenboom's avatar
jhoogenboom committed
176
             "the highest allele of that marker (default: %(default)s)")
jhoogenboom's avatar
jhoogenboom committed
177
178
179
    filtergroup.add_argument('-n', '--min-reads', metavar="N",
        type=pos_int_arg, default=_DEF_MIN_READS,
        help="require at least this number of reads for the highest allele "
jhoogenboom's avatar
jhoogenboom committed
180
             "of each marker (default: %(default)s)")
jhoogenboom's avatar
jhoogenboom committed
181
    filtergroup.add_argument('-a', '--max-alleles', metavar="N",
182
183
184
185
        type=pos_int_arg,
        help="allow no more than this number of alleles per marker; if "
             "unspecified, the amounts given in the library file are used, "
             "which have a default value of 2")
jhoogenboom's avatar
jhoogenboom committed
186
187
    filtergroup.add_argument('-x', '--max-noisy', metavar="N",
        type=pos_int_arg, default=_DEF_MAX_NOISY,
188
189
        help="entirely reject a sample if more than this number of markers "
             "have a high non-allelic sequence (default: %(default)s)")
jhoogenboom's avatar
jhoogenboom committed
190
    filtergroup.add_argument('-c', '--stuttermark-column', metavar="COLNAME",
jhoogenboom's avatar
jhoogenboom committed
191
192
193
        help="name of column with Stuttermark output; if specified, sequences "
             "for which the value in this column does not start with ALLELE "
             "are ignored")
jhoogenboom's avatar
jhoogenboom committed
194
    add_sequence_format_args(parser)
jhoogenboom's avatar
jhoogenboom committed
195
196
197
198
#add_arguments


def run(args):
199
200
    files = get_input_output_files(args)
    if not files:
jhoogenboom's avatar
jhoogenboom committed
201
202
        raise ValueError("please specify an input file, or pipe in the output "
                         "of another program")
203

204
205
206
207
    find_alleles(files[0], files[1], args.report, args.min_reads,
                 args.min_allele_pct, args.max_noise_pct, args.max_alleles,
                 args.max_noisy, args.stuttermark_column, args.sequence_format,
                 args.library)
jhoogenboom's avatar
jhoogenboom committed
208
#run