forked from thegenemyers/DALIGNER
-
Notifications
You must be signed in to change notification settings - Fork 0
/
README
331 lines (273 loc) · 19.6 KB
/
README
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
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
/************************************************************************************\
* *
* Copyright (c) 2014, Dr. Eugene W. Myers (EWM). All rights reserved. *
* *
* Redistribution and use in source and binary forms, with or without modification, *
* are permitted provided that the following conditions are met: *
* *
* · Redistributions of source code must retain the above copyright notice, this *
* list of conditions and the following disclaimer. *
* *
* · Redistributions in binary form must reproduce the above copyright notice, this *
* list of conditions and the following disclaimer in the documentation and/or *
* other materials provided with the distribution. *
* *
* · The name of EWM may not be used to endorse or promote products derived from *
* this software without specific prior written permission. *
* *
* THIS SOFTWARE IS PROVIDED BY EWM ”AS IS” AND ANY EXPRESS OR IMPLIED WARRANTIES, *
* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND *
* FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL EWM BE LIABLE *
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES *
* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS *
* OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY *
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING *
* NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN *
* IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. *
* *
* For any issues regarding this software and its use, contact EWM at: *
* *
* Eugene W. Myers Jr. *
* Bautzner Str. 122e *
* 01099 Dresden *
* GERMANY *
* Email: [email protected] *
* *
\************************************************************************************/
The Daligner Overlap Library
Author: Gene Myers
First: July 17, 2013
Current: August 29, 2014
The commands below permit one to find all significant local alignments between reads
encoded in Dazzler database. The assumption is that the reads are from a PACBIO RS II
long read sequencer. That is the reads are long and noisy, up to 15% on average.
Recall that a database has a current partition that divides it into blocks of a size
that can conveniently be handled by calling the "dalign" overlapper on all the pairs of
blocks producing a collection of .las local alignment files that can then be sorted and
merged into an ordered sequence of sorted files containing all alignments between reads
in the data set. The alignment records are parsimonious in that they do not record an
alignment but simply a set of trace points, typically every 100bp or so, that allow the
efficient reconstruction of alignments on demand.
1. daligner [-vbd] [-k<int(14)>] [-w<int(6)>] [-h<int(35)>] [-t<int>]
[-e<double(.70)] [-l<int(1000)] [-s<int(100)>]
[-H<int>] [-M<int>] <subject:db> <target:db> ...
Compare sequences in the trimmed <subject> block against those in the list of <target>
blocks searching for local alignments involving at least -l base pairs (default 1000)
or more, that have an average correlation rate of -e (default 70%). The local
alignments found will be output in a sparse encoding where a trace point on the
alignment is recorded every -s base pairs of the a-read (default 100bp). The subject
reads are compared in both orientations against each target read and local alignments
meeting the criteria are output to one of several created files described below. The
-v option turns on a verbose reporting mode that gives statistics on each major step of
the computation.
There cannot be more than 65,535 reads in a given db block, and each read must be less
than 65,535 characters long. DBsplit is guaranteed to produce blocks that meet this
condition.
The options -k, -h, and -w control the initial filtration search for possible matches
between reads. Specifically, our search code looks for a pair of diagonal bands of
width 2^w (default 2^6 = 64) that contain a collection of exact matching k-mers
(default 14) between the two reads, such that the total number of bases covered by the
k-mer hits is h (default 35). k cannot be larger than 16 in the current implementation.
If the -b option is set, then the daligner assumes the data has a strong compositional
bias (e.g. >65% AT rich), and at the cost of a bit more time, dynamically adjusts k-mer
sizes depending on compositional bias, so that the mers used have an effective
specificity of 4^k. If the -d option is set, then the daligner looks for a ".dust"
track produced by DBdust and if it finds it, then it ignores k-mers that contain any
bases from a low-complexity interval annotated by DBdust. Lastly, some k-mers are
significantly over-represented (e.g. homopolymer runs). These can be suppressed with
the -t option: any k-mer that occurs more than t times in either the subject or target
is not counted in the heuristic. If the -t option is absent then t is effectively
infinity.
bias (e.g. >65% AT rich), and at the cost of a bit more time, dynamically adjusts k-mer
Invariably, some k-mers are significantly over-represented (e.g. homopolymer runs).
These k-mers create an excessive number of matching k-mer pairs and left unaddressed
would cause daligner to overflow the available physical memory. One way to deal with
this is to explicitly set the -t parameter which suppresses the use of any k-mer that
occurs more than t times in either the subject or target block. However, a better way
to handle the situation is to let the program automatically select a value of t that
meets a given memory usage limit specified (in Gb) by the -M parameter. By default
daligner will use the amount of physical memory as the choice for -M. If you want to
use less, say only 8Gb on a 24Gb HPC cluster node because you want to run 3 daligner
jobs on the node, then specify -M8. Specifying -M0 basically indicates that you do not
want daligner to self adjust k-mer suppression to fit within a given amount of memory.
For each subject, target pair of blocks, say X and Y, the program produces files
containing alignments with the names X.Y.[C|N]#.las and if X !≠ Y then also files
Y.X.[C|N]#.las, where C indicates that the b-reads are complemented and N indicates
they are not (both comparisons are performed) and # is the thread that detected and
wrote out the collection of overlaps. For example, if the defined constant NTHREAD (in
file filter.h) is 4 (the default) when the program is compiled, then 8 or 16 files are
output for each subject,target pair depending on whether or not the subject and target
are the same block. Each found alignment is recorded as -- a[ab,ae] x b^o[bb,be] --
where a and b are the indices (in the trimmed DB) of the reads that overlap, o
indicates whether the b-read is from the same or opposite strand, and [ab,ae] and
[bb,be] are the intervals of a and b^o, respectively, that align. The file X.Y.O#.las
contains the alignments produced by thread # for which the a-read is from X and the
b-read is from Y and in orientation O.
While the default parameter settings are good for raw Pacbio data, daligner can be used
for efficiently finding alignments in corrected reads or other less noisy reads. For
example, after error correction, we run "daligner -k15 -s5 -h60 -e.95 -m.80 -s500" on
the corrected reads and at these settings it is very fast.
2. LAsort [-v] <align:las> ...
Sort each .las alignment file specified on the command line. For each file it reads in
all the overlaps in the file and sorts them in lexicographical order of (a,b,o,ab)
assuming each alignment is recorded as a[ab,ae] x b^o[bb,be]. It then writes them all
to a file named <align>.S.las (assuming that the input file was <align>.las). With the
-v option set then the program reports the number of records read and written.
3. LAmerge [-v] <merge:las> <parts:las> ...
Merge the .las files <parts> into a singled sorted file <merge>, where it is assumed
that the input <parts> files are sorted. Due to operating system limits, the number of
<parts> files must be <= 252. With the -v option set the program reports the # of
records read and written.
Used correctly, LAmerge and LAsort together allow one to perform an "external" sort
that produces a collection of sorted files containing in aggregate all the local
alignments found by the daligner, such that their concatenation is sorted in order of
(a,b,o,ab). In particular, this means that all the alignments for a given a-read will
be found consecutively in one of the files. So computations that need to look at all
the alignments for a given read can operate in simple sequential scans of these
sorted files.
4. LAshow [-coU] [-(a|r):<db>] [-i<int(4)>] [-w<int(100)>] [-b<int(10)>]
<align:las> [ <reads:range> ... ]
LAshow produces a printed listing of the local alignments contained in the specified
.las file. If a list of read ranges is given then only the overlaps for which the
a-read is in the set specified by the list are displayed. See DBshow for an explanation
of how the list of read ranges is interpreted.
If the -c option is given then a cartoon rendering is displayed, and if -a or -r option
is set then an alignment of the local alignment is displayed. If both are set then the
cartoon is first, and if neither are set then a simple one line summary of the local
alignment is displayed. To display alignments the name of the database containing the
reads must be given (and it must refer to the whole database, not a block thereof).
The -a option puts exactly -w columns per segment of the display, whereas the -r option
puts exactly -w a-read symbols in each segment of the display. The -r display mode is
useful when one wants to visually compare two alignments involving the same a-read.
The -i option sets the indent for the cartoon and/or alignment if they are requested.
The -w option sets the width of an alignment, the -b sets the number of symbols on
either side of the aligned segments to display, and -U specifies that uppercase should
be used for DNA sequence instead of the default lowercase. If the -o option is set then
only proper overlaps are shown.
5. LAcat <source:las> > <target>.las
Given argument <source>, find all files <source>.1.las, <source>.2.las, ...
<source>.n.<las> where <source>.i.las exists for every i in [1,n]. Then
concatenate these files in order into a single .las file and pipe the result
to the standard output.
6. LAsplit <target:las> (<parts:int> | <path:db>) < <source>.las
If the second argument is an integer n, then divide the alignment file <source>, piped
in through the standard input, as evenly as possible into n alignment files with the
name <target>.i.las for i in [1,n], subject to the restriction that all alignment
records for a given a-read are in the same file.
If the second argument refers to a database <path>.db that has been partitioned, then
divide the input alignment file into block .las files where all records whose a-read is
in <path>.i.db are in <align>.i.las.
7. LAcheck [-vS] [-a:<db>] <align:las> ...
LAcheck checks each .las file for structural integrity. That is, it makes sure each
file makes sense as a plausible .las file, e.g. values are not out of bound, the number
of records is correct, the number of trace points for a record is correct, and so on.
If the -S option is set then it further checks that the alignments are in sorted order,
and if the database is given with the -a option, then read ranges and lengths are
double checked. If the -v option is set then a line is output for each .las file
saying either the file is OK or reporting the first error. If the -v option is not set
then the program runs silently. The exit status is 0 if every file is deemed good, and
1 if at least one of the files looks corrupted.
8. HPCdaligner [-vbd] [-k<int(14)>] [-w<int(6)>] [-h<int(35)>] [-t<int>]
[-e<double(.70)] [-l<int(1000)] [-s<int(100)>]
[-H<int>] [-M<int>] [-dal<int(4)>] [-mrg<int(25)>]
<path:db> [<first:int>[-<last:int>]]
HPCdaligner writes a UNIX shell script to the standard output that consists of a
sequence of commands that effectively run daligner on all pairs of blocks of a split
database and then externally sorts and merges them using LAsort and LAmerge into a
collection of alignment files with names <path>.#.las where # ranges from 1 to the
number of blocks the data base is split into. These sorted files if concatenated by say
LAcat would contain all the alignments in sorted order (of a-read, then b-read, ...).
Moreover, all overlaps for a given a-read are guaranteed to not be split across files,
so one can run artifact analyzers or error correction on each sorted file in parallel.
The data base must have been previously split by DBsplit and all the parameters, except
-v, -dal, and -mrg, are passed through to the calls to daligner. The defaults for these
parameters are as for daligner. The -v flag, for verbose-mode, is also passed to all
calls to LAsort and LAmerge. -dal and -mrg options are described later. For a database
divided into N sub-blocks, the calls to daligner will produce in total 2TN^2 .las files
assuming daligner runs with T threads. These will then be sorted and merged into N^2
sorted .las files, one for each block pair. These are then merged in ceil(log_mrg N)
phases where the number of files decreases geometrically in -mrg until there is 1 file
per row of the N x N block matrix. So at the end one has N sorted .las files that when
concatenated would give a single large sorted overlap file.
The -dal option (default 4) gives the desired number of block comparisons per call to
daligner. Some must contain dal-1 comparisons, and the first dal-2 block comparisons
even less, but the HPCdaligner "planner" does the best it can to give an average load
of dal block comparisons per command. The -mrg option (default 25) gives the maximum
number of files that will be merged in a single LAmerge command. The planner makes the
most even k-ary tree of merges, where the number of levels is ceil(log_mrg N).
If the integers <first> and <last> are missing then the script produced is for every
block in the database. If <first> is present then HPCdaligner produces an incremental
script that compares blocks <first> through <last> (<last> = <first> if not present)
against each other and all previous blocks 1 through <first>-1, and then incrementally
updates the .las files for blocks 1 through <first>-1, and creates the .las files for
blocks <first> through <last>.
Each UNIX command line output by the HPCdaligner can be a batch job (we use the &&
operator to combine several commands into one line to make this so). Dependencies
between jobs can be maintained simply by first running all the daligner jobs, then all
the initial sort jobs, and then all the jobs in each phase of the external merge sort.
Each of these phases is separated by an informative comment line for your scripting
convenience.
9. LA2gfa [-a:<db>] <align:las>
Write the overlaps and containments found by daligner in Graphical Fragment Assembly
(gfa) format. This textual output is quite large due to the long CIGAR strings but
is suitable for small assembly projects and useful for prototyping algorithms.
Example:
// Recall G.db from the example in DAZZ_DB/README
> cat G.db
files = 1
1862 G Sim
blocks = 2
size = 11 cutoff = 0 all = 0
0 0
1024 1024
1862 1862
> HPCplanner -d -t5 G | csh -v // Run the HPCdazzler script
# Dazzler jobs (2)
dazzler -d -t5 G.1 G.1
dazzler -d -t5 G.2 G.1 G.2
# Initial sort jobs (4)
LAsort G.1.G.1.*.las && LAmerge G.L1.1.1 G.1.G.1.*.S.las && rm G.1.G.1.*.S.las
LAsort G.1.G.2.*.las && LAmerge G.L1.1.2 G.1.G.2.*.S.las && rm G.1.G.2.*.S.las
LAsort G.2.G.1.*.las && LAmerge G.L1.2.1 G.2.G.1.*.S.las && rm G.2.G.1.*.S.las
LAsort G.2.G.2.*.las && LAmerge G.L1.2.2 G.2.G.2.*.S.las && rm G.2.G.2.*.S.las
# Level 1 jobs (2)
LAmerge G.1 G.L1.1.1 G.L1.1.2 && rm G.L1.1.1.las G.L1.1.2.las
LAmerge G.2 G.L1.2.1 G.L1.2.2 && rm G.L1.2.1.las G.L1.2.2.las
> LAshow -c -a:G -w50 G.1 | more // Take a look at the result !
G.1: 34,510 records
1 9 c [ 0.. 1,876] x [ 9,017..10,825] ( 18 trace pts)
12645
A ---------+====> dif/(len1+len2) = 398/(1876+1808) = 21.61%
B <====+---------
9017
1 ..........gtg-cggt--caggggtgcctgc-t-t-atcgcaatgtta
|||*||||**||||||||*||||*|*|*||**|*|*||||
9008 gagaggccaagtggcggtggcaggggtg-ctgcgtcttatatccaggtta 27.5%
35 ta-ctgggtggttaaacttagccaggaaacctgttgaaataa-acggtgg
||*|||||||||||||*|**|*||*|*||||||*|**|||||*|*|||||
9057 tagctgggtggttaaa-tctg-ca-g-aacctg-t--aataacatggtgg 24.0%
83 -ctagtggcttgccgtttacccaacagaagcataatgaaa-tttgaaagt
*||||||||*||||||||*||**||||*|||**|||||||*||||*||||
9100 gctagtggc-tgccgttt-ccgcacag-agc--aatgaaaatttg-aagt 20.0%
131 ggtaggttcctgctgtct-acatacagaacgacggagcgaaaaggtaccg
||*|||||||||||||*|*||||*|*|*||||||||||*||||||||||*
9144 gg-aggttcctgctgt-tcacat-c-ggacgacggagc-aaaaggtacc- 16.0%
...
> LAcat G >G.las // Combine G.1.las & G.2.las into a single .las file
> LAshow G | more // Take another look, now at G.las
G: 62,654 records
1 9 c [ 0.. 1,876] x [ 9,017..10,825] : < 398 diffs ( 18 trace pts)
1 38 c [ 0.. 7,107] x [ 5,381..12,330] : < 1,614 diffs ( 71 trace pts)
1 49 n [ 5,493..14,521] x [ 0.. 9,065] : < 2,028 diffs ( 91 trace pts)
1 68 n [12,809..14,521] x [ 0.. 1,758] : < 373 diffs ( 17 trace pts)
1 147 c [ 0..13,352] x [ 854..14,069] : < 2,993 diffs (133 trace pts)
1 231 n [10,892..14,521] x [ 0.. 3,735] : < 816 diffs ( 37 trace pts)
1 292 c [ 3,835..14,521] x [ 0..10,702] : < 2,353 diffs (107 trace pts)
1 335 n [ 7,569..14,521] x [ 0.. 7,033] : < 1,544 diffs ( 70 trace pts)
1 377 c [ 9,602..14,521] x [ 0.. 5,009] : < 1,104 diffs ( 49 trace pts)
1 414 c [ 6,804..14,521] x [ 0.. 7,812] : < 1,745 diffs ( 77 trace pts)
1 415 c [ 0.. 3,613] x [ 7,685..11,224] : < 840 diffs ( 36 trace pts)
1 445 c [ 9,828..14,521] x [ 0.. 4,789] : < 1,036 diffs ( 47 trace pts)
1 464 n [ 0.. 1,942] x [12,416..14,281] : < 411 diffs ( 19 trace pts)
> LA2gfa -a:G G.las > G.gfa
...