File: createAugustusJoblist.pl

package info (click to toggle)
augustus 3.2.3%2Bdfsg-1
  • links: PTS, VCS
  • area: main
  • in suites: stretch
  • size: 289,676 kB
  • sloc: cpp: 48,711; perl: 13,339; ansic: 1,251; makefile: 859; sh: 58
file content (136 lines) | stat: -rwxr-xr-x 4,879 bytes parent folder | download
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
#!/usr/bin/perl
#
# createAugustusJoblist.pl
# Create a joblist with overlapping sequence chunks from multiple fasta files.
# Mario Stanke, 22.01.2007
#
#

use strict;
use Getopt::Long;

my $usage = "$0 -- Create a joblist with sequence chunks from multiple fasta files.\n\n";
$usage .= "Usage: $0\n\n";
$usage .= "parameters:\n";
$usage .= "--sequences seqs.lst input sequences, format: each line contains one sequence including the full path and its size, e.g.\n";
$usage .= "                     /cluster/data/panTro2/1/chr1.fa\t1\t229974691\n";
$usage .= "                     /cluster/data/panTro2/1/chr1_random\t1\t9420409\n";
$usage .= "                     /cluster/data/panTro2/2/chr2a\t1\t114460064\n";
$usage .= "                     or\n";
$usage .= "                     /cluster/data/panTro2/1/chr1_random\t/hints/chr1_random\t1\t9420409\n";
$usage .= "                     /cluster/data/panTro2/2/chr2a\t/hints/chr2a\t1\t114460064\n";
$usage .= "--outputdir s        directory, in which later the AUGUSTUS output will be written.\n";
$usage .= "--command s          AUGUSTUS command, e.g. \"augustus --species=human --maxDNAPieceSize=600000\".\n";
$usage .= "--joblist job.lst    filename with list of jobs as given to parasol.\n";
$usage .= "--chunksize n        chunk size. Each sequence is (imaginarily) cut into chunks of this size\n\n";
$usage .= "options:\n";
$usage .= "--overlap n          overlap. Neighboring chunks overlap by this number of bases.\n";
$usage .= "--padding n          padding on both sides (default 0).\n";
$usage .= "--errordir errdir    directory, in which later the AUGUSTUS error messages will be written.\n";
$usage .= "--check              insert parasol input/output checks.\n";
$usage .= "--wrap=s             have each job in a separate file, preceded by command s.\n";
$usage .= "--jobprefix=s        prefix of job name (default: \"job.\")\n";

if (@ARGV < 2) {
    print $usage;
    exit;
}

my ($seqfilename, $chunksize, $overlap, $outputdir, $errordir, $command, $joblist, $seqnr, $name, $predStart, $predEnd, $pE, $chunkid, $padding, $check, $wrap, $wholecommand);
my $jobnr = 0;
$padding = 0;
$check = 0;
$wrap = "";
my $jobprefix = "job.";

GetOptions('sequences=s'=>\$seqfilename,
	   'chunksize=i'=>\$chunksize,
	   'overlap:i'=>\$overlap,
	   'outputdir=s'=>\$outputdir,
	   'errordir=s'=>\$errordir,
	   'command=s'=>\$command,
	   'padding:i'=>\$padding,
	   'joblist=s'=>\$joblist,
	   'check!'=>\$check,
	   'wrap=s'=>\$wrap,
	   'jobprefix:s'=>\$jobprefix);

die ("Need to specify chunksize.\n") unless (defined $chunksize);
die ("Need to specify command.\n") unless (defined $command);
die ("Need to specify joblist.\n") unless (defined $joblist);
die ("Need to specify chunksize.\n") unless (defined $chunksize);
die ("Need to specify outputdir.\n") unless (defined $outputdir);

open(SEQ, "<$seqfilename") or die ("Could not open $seqfilename");
open (BATCH, ">$joblist") or die "Couldn't open $joblist\n";

$outputdir =~ s/\/$//;
$seqnr = 1;
while (<SEQ>) {
    my ($path, $start, $end, $hints);
    if (/(.+)\s+(.+)\s+(.+)\s+(.+)/) {
	$path = $1;
	$hints = $2;
	$start = $3;
	$end = $4;
    } elsif (/(.+)\s+(.+)\s+(.+)/){
	$path = $1;
	$start = $2;
	$end = $3;
	undef $hints;
    }
    $start -= $padding;
    $end += $padding;

    my $ovlp;
    if (!$overlap) {
	$ovlp = (($chunksize > 3000000)? 500000 : $$chunksize/6);
    } else {
	$ovlp = $overlap;
    }
    my $chunknr = 0;
    $name = $path;
    $name =~ s/.*\///;
    $predEnd = -1;
    for ($predStart = $start; $predStart <= $end && $predEnd < $end; $predStart = $predEnd+1-$ovlp) {
	$chunknr++;
	$predEnd = $predStart + $chunksize-1;
	$pE = $predEnd;
	if ($pE > $end){
	    $pE = $end;
	}
	$chunkid = sprintf ("%03d", $chunknr);
	my $gfffilename = "$outputdir/$seqnr.$chunkid.${name}.$predStart..$pE.gff";
	my $errorfilename;
	if (defined $errordir) {
	    $errorfilename = "$errordir/$seqnr.$chunkid.${name}.$predStart..$predEnd.err";
	} else {
	    $errorfilename = "$outputdir/$seqnr.$chunkid.${name}.$predStart..$pE.err";
	}
	$wholecommand = $command;
	if (defined $hints) {
	    if ($check){
		$wholecommand .= " --hintsfile={check in exists $hints}";
	    } else {
		$wholecommand .= " --hintsfile=$hints";
	    }
	}
	if ($check) {
	    $wholecommand .=  " --predictionStart=$predStart --predictionEnd=$pE {check in line+ $path} --outfile={check out line+ $gfffilename} --errfile=$errorfilename\n";
	} else {
	    $wholecommand .=  " --predictionStart=$predStart --predictionEnd=$pE $path --outfile=$gfffilename --errfile=$errorfilename\n";
	}
	if ($wrap eq "") {
	    print BATCH $wholecommand;
	} else {
	    $jobnr++;
	    my $jobname = "$jobprefix$jobnr";
	    open (JOB, ">$jobname");
	    print JOB "$wrap\n" . $wholecommand;
	    print BATCH "$jobname\n";
	    close JOB;
	    system("chmod +x $jobname");
	}
    }
    $seqnr++;
}