This file is indexed.

/usr/lib/python2.7/dist-packages/biotools/sequence.py is in python-biotools 1.2.12-2.

This file is owned by root:root, with mode 0o644.

The actual contents of the file can be viewed below.

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
from biotools.annotation import Annotation


def chop(seq, length=70):
    '''
    Yields a chunk of a sequence of no more than `length` characters,
    it is meant to be used to print fasta files.
    '''

    while seq:
        try:
            piece, seq = seq[:length], seq[length:]
        except IndexError:
            piece, seq = seq, ''
        yield piece
    raise StopIteration()


def isprot(seq, nucleotides='ATUCGNYRatucgnyr- '):
    '''
    Check whether the current sequence is a protein or nucleotide sequence.
    '''

    for c in seq:
        if c not in nucleotides:
            return True
    else:
        return False


class Sequence(object):
    '''
    A wrapper class for sequences.
    '''

    def __init__(self, name, seq, **kwargs):
        '''
        Instantiates a Sequence object with a given sequence. Some other
        useful parameters that the `Sequence` constructor can handle are:
            * `qual` => the quality scores (an array of integers) of the
                sequence,
            * `type` => the type of the sequence, either prot or nucl,
            * `start` => the starting position of the sequence within a
                supersequence,
            * `end` => the ending position of the sequnece within a
                supersequence,
            * `step` => the 'step' of the sequence, usually +1 for top-strand
                sequences, and -1 for bottom-strand sequences, but can handle
                other values as well,
            * `original` => the original `Sequence` object from which this one
                derives,
            * `defline` => the definition line for this sequnce from a fasta
                file.
        If one of these are not given, they will default to the most logical
        value that can be determined from the other values and sequence (e.g.,
        if `end < start`, then `step` is probably -1).
        '''

        self.name = name
        self.seq = seq
        self.qual = kwargs.get('qual', None)
        if 'type' in kwargs:
            self.type = kwargs['type']
        self.start = kwargs.get('start', 1)
        self.end = kwargs.get('end', self.start - 1 + len(seq))
        self.step = kwargs.get('step', -1 if self.start > self.end else 1)
        self.original = kwargs.get('original', self)
        self.defline = kwargs.get('defline', '')

    def __getattr__(self, attr):
        if attr == 'type':
            self.type = 'prot' if isprot(self.seq) else 'nucl'
            return self.type
        raise AttributeError('%r object has no attribute %r' % 
                             (self.__class__.__name__, attr))

    def __getitem__(self, key):
        '''
        sequence[i] or sequence[start:end] or sequence[start:end:step]
        constructs a new Sequence that is a subsequence as described by the
        way this function is called. This function will automatically fill in
        the name, sequence, start, stop, step, original, and type of the
        subsequence. It also tries to fill in the annotations, but annotations
        are handled pretty poorly right now, so it's probably best not to
        worry about those, but it will work if you really want to.
        '''

        try:
            start, stop, step = key.indices(len(self.seq))
        except AttributeError:
            start, stop, step = key, key + 1, 1

        order = abs(self.step) / self.step
        r = stop - (stop - start) % step - step
        seq = ''.join(self.seq[x] for x in xrange(start, stop, step))
        qual = self.qual and [self.qual[x] for x in xrange(start, stop, step)]
        info = (self.name, start, stop, step)
        self.type
        return Sequence("subsequence(%s, %d, %d, %d)" % info, seq,
                        qual=qual, original=self.original, type=self.type,
                        start=self.start + start * order,
                        end=self.start + r * order,
                        step=step * self.step)

    '''
    Some other things you can do with a Sequence object:
    * len(sequence) => gives the length of the sequence.
    * for character in sequence: => allows you to loop over each character in
        the sequence.
    * dictionary[sequence] => allows sequences to be used as keys for
        dictionaries and allows you to have sequences in sets. This relies on
        the test seqA == seqB, described next.
    * seqA == seqB => compare two sequences. The sequences are the same if
        they have the same sequence AND name. Therefore, two sequences with
        different names are treated as separate items in a set and separate
        keys in a dictionary. If you need to match only the sequence, use
        seqA.seq == seqB.seq.
    * print sequence => print a fasta / fastq (depending on whether there are
        any quality scores) representation of the sequence. Sequence objects
        in any other data structure (e.g., list, dictionary) are printed as
        (e.g., <Sequence 0x000000>). If you want to change that, you can do:
            def __repr__(self):
                return self.__str__()
    '''

    def upper(self):
        return Sequence(self.name, self.seq.upper(), type=self.type,
                        qual=self.qual, original=self.original,
                        defline=self.defline, start=self.start, step=self.step,
                        end=self.end)

    def __iter__(self):
        for c in self.seq:
            yield c
        raise StopIteration()

    def __len__(self):
        return len(self.seq)

    def __hash__(self):
        return hash(self.seq)

    def __eq__(self, other):
        try:
            return self.seq == other.seq and self.name == other.name
        except AttributeError:
            return (self.seq == other)

    def __str__(self):
        if self.qual:
            return '@%s\n%s\n+\n%s' % (self.name, self.seq,
                                       ''.join(chr(ord('A') - 1 + q)
                                               for q in self.qual))
        else:
            return '>%s %s\n%s' % (self.name, self.defline,
                                   '\n'.join(chop(self.seq, 70)))


def annotation(seq, source, type, **kwargs):
    '''
    Creates an `Annotation` object for the given sequence from a source
    (e.g., "phytozome7.0") of a particular type (e.g., "gene").
    '''
    try:
        sname = source.name
    except AttributeError:
        sname = source

    start, end = min(seq.start, seq.end), max(seq.start, seq.end)
    strand = '+' if seq.step == 1 else '-'
    attrs = ';'.join('%s=%s' % (key, str(kwargs[key])) for key in kwargs)

    return Annotation(seq.original.name, sname, type, start, end, '.',
                      strand, start % 3, attrs)