Skip to content

Latest commit

 

History

History
246 lines (188 loc) · 6.25 KB

README.rst

File metadata and controls

246 lines (188 loc) · 6.25 KB
Raw

pyWFA

A python wrapper for wavefront alignment using WFA2-lib

Installation

To download from pypi:

pip install pywfa

From conda:

conda install -c bioconda pywfa

Build from source:

git clone https://github.com/kcleal/pywfa
cd pywfa
pip install .

Overview

Alignment of pattern and text strings can be performed by accessing WFA2-lib functions directly:

from pywfa import WavefrontAligner

pattern = "TCTTTACTCGCGCGTTGGAGAAATACAATAGT"
text =    "TCTATACTGCGCGTTTGGAGAAATAAAATAGT"
a = WavefrontAligner(pattern)
score = a.wavefront_align(text)
assert a.status == 0  # alignment was successful
assert a.cigarstring == "3M1X4M1D7M1I9M1X6M"
assert a.score == -24
a.cigartuples
>>> [(0, 3), (8, 1), (0, 4), (2, 1), (0, 7), (1, 1), (0, 9), (8, 1), (0, 6)]
a.cigar_print_pretty()

>>> ALIGNMENT       3M1X4M1D7M1I9M1X6M
  ALIGNMENT.COMPACT 1X1D1I1X
  PATTERN    TCTTTACTCGCGCGTT-GGAGAAATACAATAGT
             ||| |||| ||||||| ||||||||| ||||||
  TEXT       TCTATACT-GCGCGTTTGGAGAAATAAAATAGT

The output of cigar_pretty_print can be directed to a file, rather than stdout using:

a.cigar_print_pretty("file.txt")

To obtain a python str of this print out, access the results object (see below).

Cigartuples follow the convention:

Operation Code
M 0
I 1
D 2
N 3
S 4
H 5
= 7
X 8
B 9

For convenience, a results object can be obtained by calling the WavefrontAligner with a pattern and text:

pattern = "TCTTTACTCGCGCGTTGGAGAAATACAATAGT"
text =    "TCTATACTGCGCGTTTGGAGAAATAAAATAGT"
a = WavefrontAligner(pattern)
result = a(text)  # alignment result
result.__dict__
>>> {'pattern_length': 32, 'text_length': 32, 'pattern_start': 0, 'pattern_end': 32, 'text_start': 0, 'text_end': 32, 'cigartuples': [(0, 3), (8, 1), (0, 4), (2, 1), (0, 7), (1, 1), (0, 9), (8, 1), (0, 6)], 'score': -24, 'pattern': 'TCTTTACTCGCGCGTTGGAGAAATACAATAGT', 'text': 'TCTATACTGCGCGTTTGGAGAAATAAAATAGT', 'status': 0}

# Alignment can also be called with a pattern like this:
a(text, pattern)

# obtain a string in the same format as cigar_print_pretty
a.pretty
>>> 3M1X4M1D7M1I9M1X6M      ALIGNMENT
    1X1D1I1X      ALIGNMENT.COMPACT
          PATTERN    TCTTTACTCGCGCGTT-GGAGAAATACAATAGT
                     |||*|||| ||||||| |||||||||*||||||
          TEXT       TCTATACT-GCGCGTTTGGAGAAATAAAATAGT

Configure

To configure the WaveFrontAligner, options can be provided during initialization:

from pywfa import WavefrontAligner

a = WavefrontAligner(scope="score",
                     distance="affine2p",
                     span="end-to-end",
                     heuristic="adaptive")

Supported distance metrics are "affine" (default) and "affine2p". Scope can be "full" (default) or "score". Span can be "ends-free" (default) or "end-to-end". Heuristic can be None (default), "adaptive" or "X-drop".

When using heuristic functions it is recommended to check the status attribute:

pattern = "AAAAACCTTTTTAAAAAA"
text = "GGCCAAAAACCAAAAAA"
a = WavefrontAligner(heuristic="adaptive")
a(pattern, text)
a.status
>>> 0   # successful alignment, -1 indicates the alignment was stopped due to the heuristic

Default options

The WavefrontAligner will be initialized with the following default options:

Parameter Default value
pattern None
distance "affine"
match 0
gap_opening 6
gep_extension 2
gap_opening2 24
gap_extension2 1
scope "full"
span "ends-free"
pattern_begin_free 0
pattern_end_free 0
text_begin_free 0
text_end_free 0
heuristic None
min_wavefront_length 10
max_distance_threshold 50
steps_between_cutoffs 1
xdrop 20

Modifying the cigar

If desired the cigar can be modified so the end operation is either a soft-clip or a match, this makes the alignment cigar resemble those produced by bwa, for example:

pattern = "AAAAACCTTTTTAAAAAA"
text = "GGCCAAAAACCAAAAAA"
a = WavefrontAligner(pattern)

res = a(text, clip_cigar=False)
print(cigartuples_to_str(res.cigartuples))
>>> 4I7M5D6M

res(text, clip_cigar=True)
print(cigartuples_to_str(res.cigartuples))
>>> 4S7M5D6M

An experimental feature is to trim short matches at the end of alignments. This results in alignments that approximate local alignments:

pattern = "AAAAAAAAAAAACCTTTTAAAAAAGAAAAAAA"
text = "ACCCCCCCCCCCAAAAACCAAAAAAAAAAAAA"
a = WavefrontAligner(pattern)

# The unmodified cigar may have short matches at the end:
res = a(text, clip_cigar=False)
res.cigartuples
>>> [(0, 1), (1, 5), (8, 6), (0, 7), (2, 5), (0, 5), (8, 1), (0, 7)]
res.aligned_text
>>> ACCCCCCCCCCCAAAAACCAAAAAAAAAAAAA
res.text_start, res.text_end
>>> 0, 32

# The minimum allowed block of matches can be set at e.g. 5 bp, which will trim off short matches
res = a(text, clip_cigar=True, min_aligned_bases_left=5, min_aligned_bases_right=5)
res.cigartuples
>>> [(4, 12), (0, 7), (2, 5), (0, 5), (8, 1), (0, 7)]
res.aligned_text
>>> AAAAACCAAAAAAAAAAAAA
res.text_start, res.text_end
>>> 12, 32

# Mismatch operations X can also be elided, note this occurs after the clip_cigar stage
res = a(text, clip_cigar=True, min_aligned_bases_left=5, min_aligned_bases_right=5, elide_mismatches=True)
res.cigartuples
>>> [(4, 12), (0, 7), (2, 5), (0, 13)]
res.aligned_text
>>> AAAAACCAAAAAAAAAAAAA

Notes: The alignment score is not modified currently by trimming the cigar, however the pattern_start, pattern_end, test_start and text_end are modified when the cigar is modified.