This file is indexed.

/usr/bin/last-dotplot is in last-align 393-1.

This file is owned by root:root, with mode 0o755.

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
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
#! /usr/bin/env python

# Read pair-wise alignments in MAF or LAST tabular format: write an
# "Oxford grid", a.k.a. dotplot.

# TODO: Currently, pixels with zero aligned nt-pairs are white, and
# pixels with one or more aligned nt-pairs are black.  This can look
# too crowded for large genome alignments.  I tried shading each pixel
# according to the number of aligned nt-pairs within it, but the
# result is too faint.  How can this be done better?

import sys, os, re, itertools, optparse
import Image, ImageDraw, ImageFont, ImageColor

my_name = os.path.basename(sys.argv[0])
usage = """
  %prog --help
  %prog [options] last-tabular-output dotplot.png
  %prog [options] last-tabular-output dotplot.gif
  etc."""
parser = optparse.OptionParser(usage=usage)
# Replace "width" & "height" with a single "length" option?
parser.add_option("-x", "--width", type="int", dest="width", default=1000,
                  help="maximum width in pixels (default: %default)")
parser.add_option("-y", "--height", type="int", dest="height", default=1000,
                  help="maximum height in pixels (default: %default)")
parser.add_option("-f", "--fontfile", dest="fontfile",
                  help="TrueType or OpenType font file")
parser.add_option("-s", "--fontsize", type="int", dest="fontsize", default=11,
                  help="TrueType or OpenType font size (default: %default)")
parser.add_option("-c", "--forwardcolor", dest="forwardcolor", default="red",
                  help="Color for forward alignments (default: %default)")
parser.add_option("-r", "--reversecolor", dest="reversecolor", default="blue",
                  help="Color for reverse alignments (default: %default)")
(opts, args) = parser.parse_args()
if len(args) != 2: parser.error("2 arguments needed")

if opts.fontfile:  font = ImageFont.truetype(opts.fontfile, opts.fontsize)
else:              font = ImageFont.load_default()

# Make these options too?
text_color = "black"
background_color = "white"
pix_tween_seqs = 2  # number of border pixels between sequences
border_shade = 239, 239, 239  # the shade of grey to use for border pixels
label_space = 5     # minimum number of pixels between axis labels

image_mode = 'RGB'
forward_color = ImageColor.getcolor(opts.forwardcolor, image_mode)
reverse_color = ImageColor.getcolor(opts.reversecolor, image_mode)
overlap_color = tuple([(i+j)//2 for i, j in zip(forward_color, reverse_color)])

def isGapless(alignmentColumn):
    return "-" not in alignmentColumn

def matchAndInsertLengths(alignmentColumns):
    for k, v in itertools.groupby(alignmentColumns, isGapless):
        if k:
            matchLength = sum(1 for i in v)
            yield str(matchLength)
        else:
            blockRows = itertools.izip(*v)
            insertLengths = (len(i) - i.count("-") for i in blockRows)
            yield ":".join(map(str, insertLengths))

def alignmentInput(lines):  # read alignments in either tabular or MAF format
    for line in lines:
        w = line.split()
        if line[0].isdigit():  # tabular format
            yield w
        elif line[0] == "a":  # MAF format
            sLines = []
        elif line[0] == "s":  # MAF format
            sLines.append(w)
            if len(sLines) == 2:
                alignmentRows = (i[6] for i in sLines)
                alignmentColumns = itertools.izip(*alignmentRows)
                blocks = ",".join(matchAndInsertLengths(alignmentColumns))
                yield sLines[0][0:6] + sLines[1][1:6] + [blocks]

seq_size_dic1 = {}  # sizes of the first set of sequences
seq_size_dic2 = {}  # sizes of the second set of sequences
alignments = []

f = open(args[0])
sys.stderr.write(my_name + ": reading alignments...\n")
for w in alignmentInput(f):
    seq1, pos1, strand1, size1 = w[1], int(w[2]), w[4], int(w[5])
    seq2, pos2, strand2, size2 = w[6], int(w[7]), w[9], int(w[10])
    blocks = w[11]
    seq_size_dic1[seq1] = size1
    seq_size_dic2[seq2] = size2
    aln = seq1, seq2, pos1, pos2, strand1, strand2, blocks
    alignments.append(aln)
sys.stderr.write(my_name + ": done\n")
f.close()

if not alignments:
    sys.exit(my_name + ": there are no alignments")

def natural_sort_key(my_string):
    '''Return a sort key for "natural" ordering, e.g. chr9 < chr10.'''
    parts = re.split(r'(\d+)', my_string)
    parts[1::2] = map(int, parts[1::2])
    return parts

def get_text_sizes(my_strings):
    '''Get widths & heights, in pixels, of some strings.'''
    if opts.fontsize == 0: return [(0, 0) for i in my_strings]
    image_size = 1, 1
    im = Image.new(image_mode, image_size)
    draw = ImageDraw.Draw(im)
    return [draw.textsize(i, font=font) for i in my_strings]

def get_seq_info(seq_size_dic):
    '''Return miscellaneous information about the sequences.'''
    seq_names = seq_size_dic.keys()
    seq_names.sort(key=natural_sort_key)
    seq_sizes = [seq_size_dic[i] for i in seq_names]
    name_sizes = get_text_sizes(seq_names)
    margin = max(zip(*name_sizes)[1])  # maximum text height
    return seq_names, seq_sizes, name_sizes, margin

seq_names1, seq_sizes1, name_sizes1, margin1 = get_seq_info(seq_size_dic1)
seq_names2, seq_sizes2, name_sizes2, margin2 = get_seq_info(seq_size_dic2)

def div_ceil(x, y):
    '''Return x / y rounded up.'''
    q, r = divmod(x, y)
    return q + (r != 0)

def tot_seq_pix(seq_sizes, bp_per_pix):
    '''Return the total pixels needed for sequences of the given sizes.'''
    return sum([div_ceil(i, bp_per_pix) for i in seq_sizes])

def get_bp_per_pix(seq_sizes, pix_limit):
    '''Get the minimum bp-per-pixel that fits in the size limit.'''
    seq_num = len(seq_sizes)
    seq_pix_limit = pix_limit - pix_tween_seqs * (seq_num - 1)
    if seq_pix_limit < seq_num:
        sys.exit(my_name + ": can't fit the image: too many sequences?")
    lower_bound = div_ceil(sum(seq_sizes), seq_pix_limit)
    for bp_per_pix in itertools.count(lower_bound):  # slow linear search
        if tot_seq_pix(seq_sizes, bp_per_pix) <= seq_pix_limit: break
    return bp_per_pix

sys.stderr.write(my_name + ": choosing bp per pixel...\n")
bp_per_pix1 = get_bp_per_pix(seq_sizes1, opts.width  - margin1)
bp_per_pix2 = get_bp_per_pix(seq_sizes2, opts.height - margin2)
bp_per_pix = max(bp_per_pix1, bp_per_pix2)
sys.stderr.write(my_name + ": bp per pixel = " + str(bp_per_pix) + "\n")

def get_seq_starts(seq_pix, pix_tween_seqs, margin):
    '''Get the start pixel for each sequence.'''
    seq_starts = []
    pix_tot = margin - pix_tween_seqs
    for i in seq_pix:
        pix_tot += pix_tween_seqs
        seq_starts.append(pix_tot)
        pix_tot += i
    return seq_starts

def get_pix_info(seq_sizes, margin):
    '''Return pixel information about the sequences.'''
    seq_pix = [div_ceil(i, bp_per_pix) for i in seq_sizes]
    seq_starts = get_seq_starts(seq_pix, pix_tween_seqs, margin)
    tot_pix = seq_starts[-1] + seq_pix[-1]
    return seq_pix, seq_starts, tot_pix

seq_pix1, seq_starts1, width  = get_pix_info(seq_sizes1, margin1)
seq_pix2, seq_starts2, height = get_pix_info(seq_sizes2, margin2)
seq_start_dic1 = dict(zip(seq_names1, seq_starts1))
seq_start_dic2 = dict(zip(seq_names2, seq_starts2))
hits = [0] * (width * height)  # the image data

sys.stderr.write(my_name + ": processing alignments...\n")
for aln in alignments:
    seq1, seq2, pos1, pos2, strand1, strand2, blocks = aln
    last1 = seq_size_dic1[seq1] - 1
    last2 = seq_size_dic2[seq2] - 1
    seq_start1 = seq_start_dic1[seq1]
    seq_start2 = seq_start_dic2[seq2]
    my_start = seq_start2 * width + seq_start1
    if strand1 == strand2: store_value = 1
    else:                  store_value = 2
    for i in blocks.split(","):
        if ":" in i:  # it's a gap region: skip over it
            insertLength1, insertLength2 = i.split(":")
            pos1 += int(insertLength1)
            pos2 += int(insertLength2)
        else:  # it's a match region: draw pixels for it
            matchLength = int(i)
            end1 = pos1 + matchLength
            end2 = pos2 + matchLength
            if strand1 == '+': j = xrange(pos1, end1)
            else:              j = xrange(last1 - pos1, last1 - end1, -1)
            if strand2 == '+': k = xrange(pos2, end2)
            else:              k = xrange(last2 - pos2, last2 - end2, -1)
            for real_pos1, real_pos2 in itertools.izip(j, k):
                pix1 = real_pos1 // bp_per_pix
                pix2 = real_pos2 // bp_per_pix
                hits[my_start + pix2 * width + pix1] |= store_value
            pos1 = end1
            pos2 = end2
sys.stderr.write(my_name + ": done\n")

def make_label(text, text_size, range_start, range_size):
    '''Return an axis label with endpoint & sort-order information.'''
    text_width  = text_size[0]
    label_start = range_start + (range_size - text_width) // 2
    label_end   = label_start + text_width
    sort_key    = text_width - range_size
    return sort_key, label_start, label_end, text

def get_nonoverlapping_labels(labels):
    '''Get a subset of non-overlapping axis labels, greedily.'''
    nonoverlapping_labels = []
    for i in labels:
        if True not in [i[1] < j[2] + label_space and j[1] < i[2] + label_space
                        for j in nonoverlapping_labels]:
            nonoverlapping_labels.append(i)
    return nonoverlapping_labels

def get_axis_image(seq_names, name_sizes, seq_starts, seq_pix):
    '''Make an image of axis labels.'''
    min_pos = seq_starts[0]
    max_pos = seq_starts[-1] + seq_pix[-1]
    height = max(zip(*name_sizes)[1])
    labels = [make_label(i, j, k, l) for i, j, k, l in
              zip(seq_names, name_sizes, seq_starts, seq_pix)]
    labels = [i for i in labels if i[1] >= min_pos and i[2] <= max_pos]
    labels.sort()
    labels = get_nonoverlapping_labels(labels)
    image_size = max_pos, height
    im = Image.new(image_mode, image_size, border_shade)
    draw = ImageDraw.Draw(im)
    for i in labels:
        position = i[1], 0
        draw.text(position, i[3], font=font, fill=text_color)
    return im

image_size = width, height
im = Image.new(image_mode, image_size, background_color)

for i in range(height):
    for j in range(width):
        store_value = hits[i * width + j]
        xy = j, i
        if   store_value == 1: im.putpixel(xy, forward_color)
        elif store_value == 2: im.putpixel(xy, reverse_color)
        elif store_value == 3: im.putpixel(xy, overlap_color)

if opts.fontsize != 0:
    axis1 = get_axis_image(seq_names1, name_sizes1, seq_starts1, seq_pix1)
    axis2 = get_axis_image(seq_names2, name_sizes2, seq_starts2, seq_pix2)
    axis2 = axis2.rotate(270)
    im.paste(axis1, (0, 0))
    im.paste(axis2, (0, 0))

for i in seq_starts1[1:]:
    box = i - pix_tween_seqs, margin2, i, height
    im.paste(border_shade, box)

for i in seq_starts2[1:]:
    box = margin1, i - pix_tween_seqs, width, i
    im.paste(border_shade, box)

im.save(args[1])