ProtoGene.pl 63.1 KB
Newer Older
Sebastien Moretti's avatar
Sebastien Moretti committed
1
2
3
4
5
6
7
8
9
10
11
12
#!/usr/bin/env perl

########################################################
#                                                      #
# ProtoGene : Bona-Fide Back Translation of Protein    #
#             Multiple Sequence Alignments             #
#                                                      #
########################################################

use strict;
use warnings;
use diagnostics;
Sebastien Moretti's avatar
Sebastien Moretti committed
13
14
use Carp;

Sebastien Moretti's avatar
Sebastien Moretti committed
15
use File::Which qw(which);                 # Locate external executable programs in the PATH
16
17
18
use Time::localtime;                       # Use localtime+PID for a pseudo-uniq temp file name
use Getopt::Long;                          # Options specifications

Sebastien Moretti's avatar
Sebastien Moretti committed
19

Sebastien Moretti's avatar
Sebastien Moretti committed
20
21
use Env qw(HOME);                          # Use only environmental (shell) HOME variable
use File::Copy qw(move);                   # Avoid external 'mv' command usage
22
use LWP::Simple;                           # To test gigablaster availability
Sebastien Moretti's avatar
Sebastien Moretti committed
23
24
25

use Mail::Send;                            # Send warnings and errors files by e-mail ==> only if the $userEMail variable is defined

26
27
use lib '/mnt/common/share/ProtoGene/';    # Local path for ProtoGene's own perl modules
###use lib '/mnt/local/lib/tcoffee_perl/';
Sebastien Moretti's avatar
Sebastien Moretti committed
28
use Exonerate;                             # Exonerate runner, parser, ...
29
use Views;                                 # Non-text outputs, e.g. HTML/CSS
30
#use CheckOutput;                         # Check output for cds consistancy with query
Sebastien Moretti's avatar
Sebastien Moretti committed
31
32
33



34
################## CONFIGURATION ##################
Sebastien Moretti's avatar
Sebastien Moretti committed
35
$ENV{'PATH'}        .= ':/mnt/local/bin/:./';                               # Additional path for executables
36
37
38
39
40
41

my $cachePath        = '/scratch/cluster/monthly/t_coffee/ProtoGene_Cache'; # Cache directory
my $cacheStorageTime = 15;                                                  # Do not update sequences younger than X days

my $userEMail        = 'moretti.sebastien@gmail.com';                       # To receive e-mails with encountered problems; leave blank to inactive
###################################################
Sebastien Moretti's avatar
Sebastien Moretti committed
42
43


44
my $VERSION  = '4.0.0';
Sebastien Moretti's avatar
Sebastien Moretti committed
45

Sebastien Moretti's avatar
Sebastien Moretti committed
46
47
48
49
my $webblast_exe  = 'webblast.pl';
my $blast_prog    = 'blastall';            # Or wu-blastall for Wu-BLAST; for local blast usage
my $blast_exe     = $blast_prog;
my $exonerate_exe = 'exonerate-1.0';       # Exonerate 1.0 because current parser only works with this version
50
51
52
53
54
55


################## Option management
my ($msa, $revtrans, $pep, $hideBOJ, $run_name, $template, $lim, $cache) = ('', 0, 0, 0, '', '', 0, 'update');
my ($debug, $tmp)                                                        = (0, 0);
my ($db, $species, $local, $giga)                                        = ('refseq_protein', 'All_organisms', 0, 0);
Sebastien Moretti's avatar
Sebastien Moretti committed
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
my %opts = ('msa|in=s'        => \$msa,        # Input sequences
            'revtrans'        => \$revtrans,   # Use to reverse-translate sequences with no match
            'pep'             => \$pep,        # Add the original peptide query beneath the related CDS seq
            'hideBOJ'         => \$hideBOJ,    # Hide BOJ output
            'run_name=s'      => \$run_name,   # Use another name, instead of input seq name, for result files
            'template=s'      => \$template,   # Use a template file
            'lim=i'           => \$lim,        # Limit number of input query sequences
            'cache=s'         => \$cache,      # Cache behavior

            'orgm|species=s'  => \$species,    # Organism(s) to blast against
            'db|database=s'   => \$db,         # Database to blast against
            'local'           => \$local,      # Use to specify a local db query, definied in $local_db
            'giga'            => \$giga,       # Use GigaBlaster server

            'version|V'       => sub { print "\n\tPROTOGENE : Bona-Fide Back Translation of Protein Multiple Sequence Alignments\n\tversion   : $VERSION\n\n";
                                       exit 0;
                                 },            # Print version information
            'help|H|?'        => sub { system('perldoc', "$0")==0 || print {*STDERR} "\n\tCannot open the documentation with 'perldoc'\n\n";
                                       exit 0;
                                     },        # Print full help message
            'debug'           => \$debug,      # Verbose output
            'tmp'             => \$tmp,        # To keep traces of fake intermediate files like fake xml from NCBI, fake aln, ...
Sebastien Moretti's avatar
Sebastien Moretti committed
78
           );
Sebastien Moretti's avatar
Sebastien Moretti committed
79
80
my $test_option_values = Getopt::Long::GetOptions(%opts);

81
82

################## Short help message
Sebastien Moretti's avatar
Sebastien Moretti committed
83
if ( !$test_option_values || $msa eq '' || $species eq '' || $db eq '' || $cache eq '' ){
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
    print {*STDERR} "\n\tCannot open the MSA file in FASTA format
\tTry:  $0 --msa=path_of_the_fasta_msa_file [Options]

\tOptions: --orgm=All_organisms, Bacteria, Viruses, Vertebrata,
\t                Eukaryota, Mammalia, Primates, Homo_sapiens,
\t                Gallus_gallus, Bos_taurus, Escherichia_coli,
\t                Arabidopsis_thaliana, Mus_musculus,
\t                Drosophila_Melanogaster, ...
\t                default is 'All_organisms'
\t         --db=nr, pdb, swissprot, refseq_protein
\t                default is 'refseq_protein'
\t         --local to execute a local BLAST query with
\t                --db=path_for_a_local_db_blast_formated\n
\t         --template to provide your own nucleotidic sequences
\t                following the cds file format
\t         --revtrans to reverse-translate sequences with no
\t                blast hit, in IUB (IUPAC) depiction code
\t                They are removed from the alignement by default
\t         --pep  to add the original peptide query beneath the
\t                back-translated sequence
\t         --cache=none, update, use, 'own_PATH_directory', old, empty
\t                to select the cache mode
\t                default is 'update'\n
\t         --version to print version information
\t         --help    to print a full help message\n\n";
109
    exit(1);
Sebastien Moretti's avatar
Sebastien Moretti committed
110
}
111
112


Sebastien Moretti's avatar
Sebastien Moretti committed
113
114
115
################## Check external software
checkExternalPrograms($webblast_exe, $exonerate_exe, $blast_exe); # Check external programs presence in the PATH
#FIXME ???
116
117
118
119
120
121
122
123
$blast_exe =~ s/${blast_prog}$//;

checkCacheAccessibility($cachePath); # Check cache directory accessibility



$local = 2 if ( $giga==1 );
if ( $blast_exe eq '' ){
124
    print {*STDERR} "\n\t$blast_prog program is not reachable\n\tIt could not be in your PATH or not installed\n\tIt is required only for local blast searches\n\n";
Sebastien Moretti's avatar
Sebastien Moretti committed
125
126
127
}

#Cache management
128
$cache = cacheManagement($cache, $cachePath) || '.';
Sebastien Moretti's avatar
Sebastien Moretti committed
129
130
131


#Open and check template file
132
my ($templG, $templP, $templSeq);
Sebastien Moretti's avatar
Sebastien Moretti committed
133
($templG, $templP, $templSeq) = checkTemplate($template) if ($template);
Sebastien Moretti's avatar
Sebastien Moretti committed
134
135
136


##Open and Check the fasta file
137
138
my $originalMSA   = $msa;
my $change        = 0;
Sebastien Moretti's avatar
Sebastien Moretti committed
139
($msa, $change)   = checkFastaFile($msa, $change);
140
my $fasta_checker = -1;
141
my (@input_order, @original_names, @original_seq); #Use 3 lists in parallel : original_names, seq, order(1->n)
142
open( my $MSA, '<', "$msa" ) if ( -e "$msa" );
143
144
while(<$MSA>){
    if ( $_ =~ /^>/ ){
145
146
147
148
149
150
151
        $fasta_checker++;
        my $name = $_;
        $name    =~ s/\r\n//g; #Remove return lines from windows OS '^M'
        $name    =~ s/^>[ \t]+/>/;
        chomp($name);
        @original_names = (@original_names, $name);
        @input_order    = (@input_order, $fasta_checker);
Sebastien Moretti's avatar
Sebastien Moretti committed
152
    }
153
    elsif ( $_ !~ /^>/ ){
154
155
156
        my $seqq = $_;
        $seqq    =~ s/\r\n//g;
        $seqq    =~ s/\./-/g; #for msa with '.' as gap
157
        $seqq    =~ s{[BJOUZ]}{X}ig; #U here for selenocystein
158
159
        $seqq    =~ s/[^A-Za-z\-\*\n\r]//g; #Remove all the non-gap or non-alphabetic characters from the seq
        chomp($seqq);
Sebastien Moretti's avatar
Sebastien Moretti committed
160
        #fasta sequence on 1 line
161
        if ( !exists($original_seq[$fasta_checker]) ){
162
163
164
165
166
            @original_seq = (@original_seq, $seqq);
        }
        else {
            $original_seq[$fasta_checker] = $original_seq[$fasta_checker].$seqq;
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
167
168
    }
}
169
close $MSA;
170
if ( $fasta_checker == -1 ){
Sebastien Moretti's avatar
Sebastien Moretti committed
171
    failure();
Sebastien Moretti's avatar
Sebastien Moretti committed
172
    print {*STDERR} "\tThe MSA file does NOT seem to be a protein FASTA format\n\tSee <a href='http://en.wikipedia.org/wiki/FASTA_format' target='_blank'>FASTA format</a>\n\n";
173
    exit(1);
Sebastien Moretti's avatar
Sebastien Moretti committed
174
}
175
elsif ( $lim >0 && $fasta_checker > $lim ){
Sebastien Moretti's avatar
Sebastien Moretti committed
176
    failure();
177
    print {*STDERR} "\tThe FASTA file is too large, try with less than $lim sequences\n\tor split your file\n\n";
178
179
    exit(1);
}
180
elsif ( exists( $original_seq[0] ) && $original_seq[0] =~ /[acgtu]/i ){
181
    my $first_seq = $original_seq[0];
182
    #Check if sequences are amino acids and not nucleotides
183
184
185
    my $a   = ($first_seq =~ s/[aA]//g)   || 0;
    my $c   = ($first_seq =~ s/[cC]//g)   || 0;
    my $g   = ($first_seq =~ s/[gG]//g)   || 0;
186
187
    my $t   = ($first_seq =~ s/[tTuU]//g) || 0;
    my $non = ($first_seq =~ s/[^aAcCgGtTuUXxNn-]//g) || 0;
188
    if ( ($a+$c+$g+$t) >= (($a+$c+$g+$t+$non)*80/100) ){
Sebastien Moretti's avatar
Sebastien Moretti committed
189
        failure();
190
        print {*STDERR} "\tYour sequences seem already to be nucleotides\n\tthis program purpose is to turn AMINO ACID alignments into CDS nucleotide alignments\n\n";
191
192
        exit(1);
    }
Sebastien Moretti's avatar
Sebastien Moretti committed
193
194
195
196
}
undef $fasta_checker;


197
198
199
200
201
#Check GigaBlaster status if used
if ( $giga==1 ){
    if ( head('http://www.igs.cnrs-mrs.fr/Giga2/~database/remoteblast.cgi') ){
    }
    else{
202
        $local = 0;
203
204
205
    }
}

Sebastien Moretti's avatar
Sebastien Moretti committed
206

207
208
my $date = sprintf("%d-%02d-%02d_%02d%02d", localtime->year() + 1900, localtime->mon(), localtime->mday(), localtime->hour, localtime->min)."_$$";

Sebastien Moretti's avatar
Sebastien Moretti committed
209
#Start main program with version # of programs and list original queries
210
print "\n\t   Protogene\t$VERSION\t$date\n\n";
211
open(my $EXONERATEISHERE, "$exonerate_exe --version |");
212
my $IsExonerateHere = <$EXONERATEISHERE>;
213
close $EXONERATEISHERE;
Sebastien Moretti's avatar
Sebastien Moretti committed
214
215
print "\t   $IsExonerateHere\n";
undef $IsExonerateHere;
216
undef $VERSION;
Sebastien Moretti's avatar
Sebastien Moretti committed
217
218
219
for(my $m=0; $m<=$#original_names; $m++){
    print "\t$original_names[$m]\n";
}
220
$date .= "_$$"; #Add PID
Sebastien Moretti's avatar
Sebastien Moretti committed
221
222
223
224



# run_name implementation for web server
225
$originalMSA = $run_name if ( $run_name ne '' );
Sebastien Moretti's avatar
Sebastien Moretti committed
226
227
228

unlink("${originalMSA}.cds");
unlink("${originalMSA}.cdsP");
Sebastien Moretti's avatar
Sebastien Moretti committed
229
unlink("${originalMSA}.cdsP.html");
Sebastien Moretti's avatar
Sebastien Moretti committed
230
231
232
233
234
235
unlink("${originalMSA}.out");
unlink("${originalMSA}.boj");



# Remove old files from the cache directory
236
cacheManagement('old', $cachePath, '1'); #everytime ProtoGene is running
Sebastien Moretti's avatar
Sebastien Moretti committed
237
238
239
240
241




##Build the sequences, from the alignment, to perform webblast
242
243
my $quiet = '';
EACH_SEQ:
Sebastien Moretti's avatar
Sebastien Moretti committed
244
for(my $r=0; $r<=$#input_order; $r++){
245
246
247
248
    my $right_name = $original_names[$r];
    my $templ_name = $original_names[$r];
    $right_name    =~ s/^>//;
    $templ_name    =~ s/^([^\s\t]+).*$/$1/;
Sebastien Moretti's avatar
Sebastien Moretti committed
249

250
    open(my $READY2BLAST, '>', "$cache/${date}_seq2blast.fas${r}");
251
252
    my $noGap = $original_seq[$r];
    $noGap    =~ s/[\.-]//g; #remove gaps
253
254
    print {$READY2BLAST} ">$input_order[$r]\n$noGap\n";
    close $READY2BLAST;
Sebastien Moretti's avatar
Sebastien Moretti committed
255
256
257
258
    undef $noGap;


    my @multipleequalblasthits;
259
260
261
    if ( exists($templG->{$templ_name}) && $templG->{$templ_name} ne '' ){
        @multipleequalblasthits = (@multipleequalblasthits, 'My_own_seq'); #To change for ?
        print "\n\n$r done\n\nNucleotidic template for $r:\t$templG->{$templ_name}\n\n**********************************************************************\n\n";
Sebastien Moretti's avatar
Sebastien Moretti committed
262
    }
263
264
265
    elsif ( exists($templP->{$templ_name}) && $templP->{$templ_name} ne '' ){
        @multipleequalblasthits = (@multipleequalblasthits, $templP->{$templ_name});
        print "\n\n$r done\n\nBlastP template for $r:\t$templP->{$templ_name}\n\n**********************************************************************\n\n";
Sebastien Moretti's avatar
Sebastien Moretti committed
266
267
    }
    else{
268
    ##   --> send them to webblast
269
        $quiet = '-quiet=on ' if ( $r>0 ); #Show webblast parameters only for the first webblast run
270
        launchWebblast("$cache/${date}_seq2blast.fas${r}", $quiet, $local, $db, $species, $blast_exe, "$cache/${date}.blastp${r}");
Sebastien Moretti's avatar
Sebastien Moretti committed
271

272
    ##Get the BLASTp hit(s)  acc number
273
274
275
276
277
278
279
280
        open(my $BLASTP, '<', "${cache}/${date}.blastp${r}");
        while(<$BLASTP>){
            @multipleequalblasthits = (@multipleequalblasthits, $1) if ( $_ =~ /^>\s*\d+@?\w?\w?\w?__([^\s\|\.]+).*$/ ); #Warning: double '_'
        }
        close $BLASTP;
        unlink("$cache/${date}.blastp${r}", "${cache}/${date}_seq2blast.fas${r}") if ( $tmp == 0 || exists($multipleequalblasthits[0]) );
        if ( !exists($multipleequalblasthits[0]) ){
            print "\n$input_order[$r] ...\tNo blast result found above filter thresholds for $right_name\n\n";
Sebastien Moretti's avatar
Sebastien Moretti committed
281
            buildFailureOutputFiles($r, 'No_BLASTp_Result', 'Unavailable', '');
282
283
284
285
286
            undef $original_seq[$r];
            undef $original_names[$r];
            undef $input_order[$r];
            next EACH_SEQ;
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
287
288
289
290
291
    }



    #When multiple equal blast hits  ==> use, and add, every hits
292
293
294
    my ($resultPOS, $resultBOJ) = ('', '');
    my @failureStatus = ('');
    my @intronStatus  = ('');
Sebastien Moretti's avatar
Sebastien Moretti committed
295
    for(my $qq=0; $qq<=$#multipleequalblasthits; $qq++){
296
297
298
299
300
301
        print "\n${right_name}  -->  [${multipleequalblasthits[$qq]}]\n";


        my @nt_GIs;
        my $intronStep = 0;
        if ( exists($templG->{$templ_name}) && $templG->{$templ_name} ne '' && $templG->{$templ_name} !~ /^My_Seq$/i ){
Sebastien Moretti's avatar
Sebastien Moretti committed
302
            downloadSeq($cache, $date, '', '', $templG->{$templ_name});
303
304
305
306
307
308
309
310
311
312
            @nt_GIs = $templG->{$templ_name};
        }
        elsif ( exists($templG->{$templ_name}) && $templG->{$templ_name} ne '' && $templG->{$templ_name} =~ /^My_Seq$/i ){
            open(my $MYSEQ, '>', "$cache/My_Seq-$$.fas");
            print {$MYSEQ} ">My_Seq-$$ for $templ_name\n".$templSeq->{$templ_name}."\n";
            close $MYSEQ;
            @nt_GIs = "My_Seq-$$";
        }
        else{
        #Prot ACC -> PUID
Sebastien Moretti's avatar
Sebastien Moretti committed
313
            my $protGI = blastPAcc2PGI($multipleequalblasthits[$qq]);
314
315
316
317
318
            if ( $protGI eq '' ){
                print "\tNo protein (prot GI) link found for ${multipleequalblasthits[$qq]} in $right_name\n\n";
                @failureStatus = ('PUI_unavailable', ${multipleequalblasthits[$qq]});
                next;
            }
Sebastien Moretti's avatar
Sebastien Moretti committed
319
320
#           print "\t$protGI\n";

321
        #PUID -> nt GIs & GeneID
Sebastien Moretti's avatar
Sebastien Moretti committed
322
            my ($ntGIs, $geneID) = protGI2NTGIs($protGI, $multipleequalblasthits[$qq]);
323
324
325
326
327
328
329
330
331
332
            if ( $ntGIs eq '' && $geneID eq '' ){
                print "\tNo nucleotide (nt GIs) link found for ${multipleequalblasthits[$qq]} in $right_name\n\n";
                @failureStatus = ('No_nt_link', ${multipleequalblasthits[$qq]});
                next;
            }
            print " linked with [$ntGIs $geneID]\n";


        #Get transcript and contig seq
            @nt_GIs = split(/,/, $ntGIs);
Sebastien Moretti's avatar
Sebastien Moretti committed
333
            downloadSeqFromGIs($cache, $date, '', '', @nt_GIs);
334
335
336
337
        #GeneID -> Chr
            my @geneIDs = split(/,/, $geneID);
            while(<@geneIDs>){
                my $geneID = $_;
Sebastien Moretti's avatar
Sebastien Moretti committed
338
                my ($chr, $amont, $aval) = geneID2Chr($geneID, $multipleequalblasthits[$qq]);
339
340
341
342
                print "\n\tNo gene locus found for ${multipleequalblasthits[$qq]} with $geneID in $right_name\n\n" if ($chr eq '' and $geneID ne '');

                if ( $chr ne '' ){
            #Get Chr seq
Sebastien Moretti's avatar
Sebastien Moretti committed
343
                    downloadSeq($cache, $date, $amont, $aval, $chr);
344
345
346
347
348
349
350
351
352
                    $intronStep = 1;

                    @nt_GIs = ("$chr:$amont-$aval", @nt_GIs);
                }
            }
        }


    #Align Nt seq with our protein query seq
353
        my ($POS, $BOJ) = runExonerate($exonerate_exe, $cache, $date, ${input_order[$r]}, ${original_seq[$r]}, $right_name, $multipleequalblasthits[$qq], $r, @nt_GIs);
354

Sebastien Moretti's avatar
Sebastien Moretti committed
355
        my ($POSresult, $bestNucleotide) = prepareResults4CDS($POS, $multipleequalblasthits[$qq], ${original_seq[$r]}, $right_name);
356
357
358
359
360
361
362
363
        $resultPOS .= $POSresult if ( $bestNucleotide ne '' && $POSresult ne '' && $resultPOS !~ /_G_$bestNucleotide/ );

        @intronStatus = ($POS->{0}, ${multipleequalblasthits[$qq]}, '')              if ( $intronStep==1 && exists($POS->{0}) && $POS->{0} ne '' && $POS->{0} =~ /\:/ );
        @intronStatus = ($POS->{0}, ${multipleequalblasthits[$qq]}, $bestNucleotide) if ( $intronStep==0 && ($bestNucleotide =~ /^N[CTWZG]_/ || $bestNucleotide =~ /^AC_/) );

        my ($intronLess, $nameLess) = ('', '');
        $intronLess = $POS->{0}       if ( %$BOJ eq 0 && exists($POS->{0}) && $POS->{0} ne '' );
        $nameLess   = $bestNucleotide if ( %$BOJ eq 0 && $intronStep==0 && ($bestNucleotide =~ /^N[CTWZG]_/ || $bestNucleotide =~ /^AC_/) );
Sebastien Moretti's avatar
Sebastien Moretti committed
364
        my ($BOJresult, $bestGenomic) = prepareResults4BOJ($BOJ, $multipleequalblasthits[$qq], ${original_seq[$r]}, $right_name, $intronLess, $nameLess);
365
        $resultBOJ .= $BOJresult      if ( $bestGenomic ne '' && $BOJresult ne '' && $resultBOJ !~ /_G_$bestGenomic[ \-]/ );
366
#    print " $intronStep $POS->{0} [@intronStatus] {$BOJresult} $bestNucleotide - $bestGenomic \t $BOJ->{0}\n";
Sebastien Moretti's avatar
Sebastien Moretti committed
367
368
369
    }


370
    if ( $resultPOS eq '' ){
Sebastien Moretti's avatar
Sebastien Moretti committed
371
372
        buildFailureOutputFiles($r, $failureStatus[1], $failureStatus[0], '')     if ( $failureStatus[0] ne '' );
        buildFailureOutputFiles($r, $multipleequalblasthits[0], 'Failed_Aln', '') if ( $failureStatus[0] eq '' );
373
        next;
Sebastien Moretti's avatar
Sebastien Moretti committed
374
    }
375
376
    elsif ( $resultBOJ eq '' ){
        if ( $failureStatus[0] eq 'PUI_unavailable' || $failureStatus[0] eq 'No_nt_link' ){
Sebastien Moretti's avatar
Sebastien Moretti committed
377
            buildFailureBOJOutputFile($failureStatus[1], $failureStatus[0], $right_name, ${original_seq[$r]});
378
379
        }
        elsif ( $failureStatus[0] eq '' && $intronStatus[0] ne '' ){
Sebastien Moretti's avatar
Sebastien Moretti committed
380
            buildIntronlessBOJOutputFile($intronStatus[1], $intronStatus[0], $intronStatus[2], $right_name, ${original_seq[$r]});
381
382
        }
        else{
Sebastien Moretti's avatar
Sebastien Moretti committed
383
            buildFailureBOJOutputFile($multipleequalblasthits[0], 'Unavailable', $right_name, ${original_seq[$r]});
384
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
385
        createCDSOutputFile($resultPOS, ${original_seq[$r]}, $right_name);
Sebastien Moretti's avatar
Sebastien Moretti committed
386
387
    }
    else {
Sebastien Moretti's avatar
Sebastien Moretti committed
388
389
        createCDSOutputFile($resultPOS, ${original_seq[$r]}, $right_name);
        createBOJOutputFile($resultBOJ, ${original_seq[$r]});
Sebastien Moretti's avatar
Sebastien Moretti committed
390
391
392
393
394
395
396
397
398
    }


    undef $original_seq[$r];
    undef $original_names[$r];
    undef $input_order[$r];
}


Sebastien Moretti's avatar
Sebastien Moretti committed
399
400
checkAndCleanStderrFiles("$cache/${date}_ExonerateError") if ( -e "$cache/${date}_ExonerateError" );
checkOutputFiles();
401
402
unlink("$msa") if ( $change==1 );
my @tmpFiles = glob($cache."/${date}*");
Sebastien Moretti's avatar
Sebastien Moretti committed
403
404
if ( exists($tmpFiles[0]) ){
    mkdir("$cache/${date}_TMP");
405
    move("$_", "$cache/${date}_TMP") while(<@tmpFiles>);
Sebastien Moretti's avatar
Sebastien Moretti committed
406
407
408
409
410
411
412
}


#Ouput Checker


#STDOUT Log output for our server
413
if ( -s "${originalMSA}.cds" || -s "${originalMSA}.out"){
Sebastien Moretti's avatar
Sebastien Moretti committed
414
415
    print "\n\nTERMINATION STATUS: SUCCESS\n";
    print "OUTPUT RESULTS\n";
Sebastien Moretti's avatar
Sebastien Moretti committed
416
417
418
419
420
    print "    #### File Type=        MSA Format= fasta_CDS Name= ${originalMSA}.cds\n" 
        if ( -s "${originalMSA}.cds");
    print "    #### File Type=        MSA Format= fasta_CDS+Query Name= ${originalMSA}.cdsP\n" 
        if ( -s "${originalMSA}.cdsP");
    print "    #### File Type=        MSA Format= colored_fasta_CDS Name= ${originalMSA}.cdsP.html\n" 
Sebastien Moretti's avatar
Sebastien Moretti committed
421
        if ( Views::Html("${originalMSA}.cdsP") && -s "${originalMSA}.cdsP.html");
Sebastien Moretti's avatar
Sebastien Moretti committed
422
423
424
425
    print "    #### File Type=        MSA Format= Rejected_seq Name= ${originalMSA}.out\n" 
        if ( -s "${originalMSA}.out");
    print "    #### File Type=        MSA Format= fasta_Exon-Boundaries Name= ${originalMSA}.boj\n" 
        if ( -s "${originalMSA}.boj" && -s "${originalMSA}.cds" && $hideBOJ==0);
Sebastien Moretti's avatar
Sebastien Moretti committed
426
427
}
else{
Sebastien Moretti's avatar
Sebastien Moretti committed
428
    failure();
429
    exit(1);
Sebastien Moretti's avatar
Sebastien Moretti committed
430
431
}

432
exit(0);
Sebastien Moretti's avatar
Sebastien Moretti committed
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514


=head1 NAME

ProtoGene.pl - Converts a peptidic alignment to a nucleotidic alignment through database search

=head1 SYNOPSIS

B<ProtoGene.pl> --msa=path_of_the_fasta_msa_file [options]

with these options available:

=over 4

=item I<--orgm>=targeted species in the database search

=item All_organisms [default], Bacteria, Viruses, Vertebrata, Eukaryota, Mammalia, Primates, Homo_sapiens, Gallus_gallus, Bos_taurus, Escherichia_coli, Arabidopsis_thaliana, Mus_musculus, Drosophila_Melanogaster, ...

=item I<--db>=targeted database in the database search

=item nr, pdb, swissprot, refseq_protein [default]

=item I<--local> to execute a local BLAST query with  --db=path_for_a_local_db_blast_formated

=item I<--template> to provide your own nucleotidic sequences following the cds file format

=item I<--revtrans> to reverse-translate sequences with no blast hit, in IUB (IUPAC) depiction code

=item They are removed from the alignement by default

=item I<--pep> to add the original peptide query beneath the back-translated sequence

=item I<--cache>=none, update, use, own_path_directory, old, empty

=item none: no cache usage, none temporary files are stored

=item update: use cache but update old files [default]

=item use: force use cache, whatever the age of files

=item own_path_directory: use my own directory, and its files

=item old: remove the old files in the cache directory

=item empty: empty the whole cache directory $HOME/.ProtoGene/

=item I<--version> to print version information

=item I<--help> to print this help message

=back

=head1 DESCRIPTION

PROTOGENE : Bona-Fide Back Translation of Protein Multiple Sequence Alignments
It converts a peptidic alignment to a nucleotidic alignment through database search.
Blast queries allow to get nucleotidic sequences from where peptidic sequences come from.
Exonerate aligns nucleotidic and peptidic sequences together.
PROTOGENE re-builds the original alignment with nucleotidic information it has gotten.

=head1 REQUIREMENT

=over 4

=item B<Perl 5.6 or better> is required !

=item Standard Perl modules B<lib>, B<strict>,  B<warnings>,  B<diagnostics>, B<Env> are required

=item and some other current ones : B<Getopt::Long>,  B<File::Which>,  B<File::Copy>, B<Mail::Send>

=item -

=item I<exonerate> from http://www.ebi.ac.uk/~guy/exonerate/

=item I<blast> from http://www.ncbi.nlm.nih.gov/BLAST/download.shtml or http://blast.wustl.edu/

=back

=head1 VERSION

=over 8

515
=item version 4.0.0
Sebastien Moretti's avatar
Sebastien Moretti committed
516

517
=item on Apr 02nd, 2011
Sebastien Moretti's avatar
Sebastien Moretti committed
518
519
520
521
522
523
524
525
526

=back

=head1 AUTHORS

=over 8

=item Sebastien MORETTI

Sebastien Moretti's avatar
Sebastien Moretti committed
527
=item moretti.sebastien [AT] gmail.com
Sebastien Moretti's avatar
Sebastien Moretti committed
528

529
=item Vital-IT computing center
Sebastien Moretti's avatar
Sebastien Moretti committed
530

531
=item Swiss Institute of Bioinformatics
Sebastien Moretti's avatar
Sebastien Moretti committed
532

533
=item Lausanne, Switzerland
Sebastien Moretti's avatar
Sebastien Moretti committed
534

535
=item http://www.vital-it.ch/
Sebastien Moretti's avatar
Sebastien Moretti committed
536
537
538
539
540
541
542
543
544

=back

=cut


######################  Management  ######################

#Failure declaration
Sebastien Moretti's avatar
Sebastien Moretti committed
545
sub failure {
546
547
548
    print {*STDERR} "\n\nFATAL\n\nTERMINATION STATUS: FAILURE\n\n";
    return;
}
Sebastien Moretti's avatar
Sebastien Moretti committed
549

Sebastien Moretti's avatar
Sebastien Moretti committed
550
551
552
553
554
555
556
557
# Check external programs presence in the PATH
sub checkExternalPrograms{
    for my $exe ( @_ ){
        if ( ! which($exe) ){
            failure();
            print {*STDERR} "\t'$exe' program is not reachable\n\tIt could not be in your PATH or not installed\n\n";
            exit(1);
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
558
    }
Sebastien Moretti's avatar
Sebastien Moretti committed
559
    return;
Sebastien Moretti's avatar
Sebastien Moretti committed
560
561
562
563
}

#Check cache directory accessibility: Test and change rights, if needs, of the $HOME/.ProtoGene/ cache directory
sub checkCacheAccessibility{
564
    my ($cacheDir) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
565

566
    if ( -d "$cacheDir/" && -w "$cacheDir/" && -x "$cacheDir/" && -r "$cacheDir/" ){
Sebastien Moretti's avatar
Sebastien Moretti committed
567
    }
568
569
570
    elsif ( -d "$cacheDir/" ){
        if ( -o "$cacheDir/" ){
            chmod 0754, "$cacheDir/";
571
        }
572
        else{
Sebastien Moretti's avatar
Sebastien Moretti committed
573
            print {*STDERR} "\n\t** Warning ** :\n\tPermissions do NOT seem to be valid for the current cache directory\n\n";
574
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
575
576
    }
    else{
577
578
579
        mkdir "$cacheDir/";
        if ( -o "$cacheDir/" ){
            chmod 0754, "$cacheDir/"; #0754 est en octal, donc 754 ou '0754' ne sont pas bons !
580
581
        }
        else{
Sebastien Moretti's avatar
Sebastien Moretti committed
582
            print {*STDERR} "\n\t** Warning ** :\n\tPermissions do NOT seem to be valid for the current cache directory\n\n";
583
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
584
    }
585
    return;
Sebastien Moretti's avatar
Sebastien Moretti committed
586
587
588
589
}

#Cache Management
sub cacheManagement{
590
    my ($cache, $cacheDir, $inScript) = @_;
591
592
    $inScript = 0 if ( !defined $inScript );

593
594
    if ( $cache eq 'empty' ){
        my @temp_files = glob($cacheDir.'/*.fas');
595
596
597
        if ( exists ($temp_files[0]) ){
            unlink $_ while(<@temp_files>);
        }
598
        @temp_files = glob($cacheDir.'/*_ExonerateError');
599
600
601
        if ( exists ($temp_files[0]) ){
            unlink $_ while(<@temp_files>);
        }
602
603
        unlink("$cacheDir/webblast.log", "$cacheDir/web_tempo.result", "$cacheDir/debug.tempo");
        print "\n\tcache directory [$cacheDir] is empty\n\n";
604
        exit 0;
Sebastien Moretti's avatar
Sebastien Moretti committed
605
    }
606
607
608
    elsif ( $cache eq 'old' ){
        my @temp_files = glob($cacheDir.'/*.fas');
        my @tempFiles  = grep { -M $_ > $cacheStorageTime } @temp_files; #Sort list only with $cacheStorageTime days old files
609
610
611
        if ( exists ($tempFiles[0]) ){
            unlink $_ while(<@tempFiles>);
        }
612
        print "\n\tcache directory [$cacheDir] is empty of its old files\n\n" if ($inScript==0);
613
        exit 0 if ( $inScript==0 );
Sebastien Moretti's avatar
Sebastien Moretti committed
614
    }
615
616
    elsif ( $cache eq 'update' || $cache eq 'use' ){
        $cache = $cacheDir; #Limit cache files to less than $cacheStorageTime days old ones for 'update'
Sebastien Moretti's avatar
Sebastien Moretti committed
617
    }
618
    elsif ( $cache eq 'none' ){
Sebastien Moretti's avatar
Sebastien Moretti committed
619
    }
620
621
622
623
    elsif ( -d $cache ){
        die "\n\tProtoGene cannot access to your directory\n\n"  if ( ! -x $cache );
        die "\n\tProtoGene cannot write into your directory\n\n" if ( ! -w $cache );
        die "\n\tProtoGene cannot read into your directory\n\n"  if ( ! -r $cache );
Sebastien Moretti's avatar
Sebastien Moretti committed
624
625
    }
    else {
626
        die "\n\tWrong cache argument, or your cache folder doesn't seem to exist\n\n";
Sebastien Moretti's avatar
Sebastien Moretti committed
627
628
629
630
631
632
633
634
635
636
637
638
    }

    return($cache);
}

##########################################################


######################    Fasta    ######################

#Check input aln
sub checkFastaFile{
639
    my ($msa, $change) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
640

641
    if ( -e "$msa" ){
Sebastien Moretti's avatar
Sebastien Moretti committed
642
    #Convert no fasta format to fasta with seq_reformat if available
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
        my $seqreformat = which('t_coffeeX') || ''; #desactive tant que t_coffee seq_reformat pose prob
        if ( $seqreformat ne '' ){
            my @formatType = `$seqreformat -other_pg seq_reformat -in $msa -print_format`;
            my $format     = '';
            while(<@formatType>){
                if ( $_ =~ /FORMAT:/ ){
                    $format = $_;
                    chomp($format);
                    last;
                }
            }
            if ( $format ne '' && $format !~ /FORMAT:fasta_/ ){
                system("$seqreformat -other_pg seq_reformat -in $msa -output fasta_aln -out=${msa}.007") if ($format =~ /_aln *$/);
                system("$seqreformat -other_pg seq_reformat -in $msa -output fasta_seq -out=${msa}.007") if ($format =~ /_seq *$/);
                $msa    .= '.007';
                $change  = 1;
            }
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
661
662
    }
    else {
Sebastien Moretti's avatar
Sebastien Moretti committed
663
        failure();
664
665
        print {*STDERR} "\tThe file \'$msa\' doesn't not seem to be reachable\n\n";
        exit(1);
Sebastien Moretti's avatar
Sebastien Moretti committed
666
667
668
669
670
671
672
    }

    return($msa,$change);
}

#Prepare Fasta Header of the query name to incorporate our annotations
sub fastaHeaders4ProtoGene{
673
    my ($QueryName) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
674
675

    $QueryName =~ s/  +/ /g;
676
677
678
679
    my ($cqacc, $cqdesc) = ($QueryName, $QueryName);
    $cqacc  =~ s/^ *([^ ]*) .*$/$1/;
    $cqacc  =~ s/[\s\t]*$//g;
    $cqacc  =  $cqacc.'_G_@@';
Sebastien Moretti's avatar
Sebastien Moretti committed
680
681
682
683
    $cqdesc =~ s/^ *[^ ]* *(.*)$/$1/;
    $cqdesc =~ s/^[\| \.]*//;
    $cqdesc =~ s/[\|\. ]*$//;

684
    $QueryName = $cqacc.$cqdesc;
Sebastien Moretti's avatar
Sebastien Moretti committed
685
686
687
688
689
    return($QueryName);
}

#Prepare Fasta Header to put it on the fasta header of the result outputs
sub getFastaHeaderAnnot{
690
    my ($bestTarget) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
691
692

    #Get annotation from nucleotide file header
693
694
695
    open(my $BEST, '<', "$cache/${bestTarget}.fas");
    my $annot=<$BEST> || '';
    close $BEST;
Sebastien Moretti's avatar
Sebastien Moretti committed
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
    chomp($annot);
    $annot =~ s/^ *[^ ]* //;
    $annot =~ s/[\. ]*$//;
    $annot =~ s/^ *//;
    $annot =~ s/  +/ /g;

    return($annot);
}

#########################################################


###################### Translation ######################

#Direct Reverse translation
sub reverse_trad{
712
    my ($aa) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
713
714
715
716
717
718
719

    my %reverse_code=('A' => 'GCN', 'F' => 'TTy', 'K' => 'AAr', 'P' => 'CCN', 'T' => 'ACN',
                      'C' => 'TGy', 'G' => 'GGN', 'L' => 'yTN', 'Q' => 'CAr', 'V' => 'GTN',
                      'D' => 'GAy', 'H' => 'CAy', 'M' => 'ATG', 'R' => 'mGN', 'W' => 'TGG',
                      'E' => 'GAr', 'I' => 'ATh', 'N' => 'AAy', 'S' => 'wsN', 'Y' => 'TAy',
                      'X' => 'NNN', '-' => '---', '*' => 'Trr'); # '*' is for stop codon

720
721
722
723
724
725
    my $triplet = 'nnn';
    while( my ($x,$y) =each(%reverse_code) ){
        if ( uc($aa) eq $x ){
            $triplet = $y;
            last;
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
726
727
728
729
730
731
732
733
734
735
    }

    return($triplet);
}

#########################################################


####################### webblast #######################
sub launchWebblast{
736
737
    my ($infile, $quiet, $whatBlast, $db, $species, $blast_exe, $outfile) = @_;
    my $blast_at = '';
Sebastien Moretti's avatar
Sebastien Moretti committed
738

739
740
741
    $blast_at    = "-database=$db"                         if ( $whatBlast==0 );
    $blast_at    = "-database=$db -blast_dir=${blast_exe}" if ( $whatBlast==1 );
    $blast_at    = "-database=$db -gigablast=yes"          if ( $whatBlast==2 );
Sebastien Moretti's avatar
Sebastien Moretti committed
742
743
744
745
746
747
748
749
750
751
752
753
754

#    print "             Evalue treshold : 0.05
#             Matrix : BLOSUM62
#             Filter : F
#             **********************
#             For RefSeq Protein db:
#             Blast_identity_threshold : 100
#             Cover threshold : 100
#             **********************
#             For NR Protein db:
#             Blast_identity_threshold : 95
#             Cover threshold : 95";

755
    system("$webblast_exe -program=blastp ${blast_at} -infile=$infile -matrix=BLOSUM62 -evalue=0.05 -method=geneid -filter=Off -organism=$species -identity=100 -cover=100 -hits=10 $quiet -outfile=$outfile");
Sebastien Moretti's avatar
Sebastien Moretti committed
756
757


758
759
760
761
762
763
764
    if ( -z "$outfile" ){ #Another blastP query if no hit with refseq and db was not nr !
        $quiet = '-quiet=on';
        print {*STDERR} "run BLAST.. against NR because there was no acceptable hit against $db\n"
            if ( $whatBlast != 1 && $db ne 'nr' );
        print {*STDERR} "run BLAST.. again, with decreased thresholds, because there was no acceptable hit against $db\n"
            if ( $whatBlast == 1 || $db eq 'nr' );
        $blast_at =~ s/-database=[^\s]*/-database=nr/ if ( $whatBlast != 1 );
765
        system("$webblast_exe -program=blastp ${blast_at} -infile=$infile -matrix=BLOSUM62 -evalue=0.05 -method=geneid -filter=Off -organism=$species -identity=95 -cover=95 -hits=10 $quiet -outfile=$outfile");
Sebastien Moretti's avatar
Sebastien Moretti committed
766
767
    }

768
769
770
771
772
773
774
    move('webblast.log',     "$cache"); #Move temporary webblast.pl files
    move('web_tempo.result', "$cache");
    move('debug.tempo',      "$cache");

    unlink("$cache/webblast.log")     if ( -z "$cache/webblast.log" );
    unlink("$cache/web_tempo.result") if ( -z "$cache/web_tempo.result" );
    unlink("$cache/debug.tempo")      if ( -z "$cache/debug.tempo" );
Sebastien Moretti's avatar
Sebastien Moretti committed
775

776
    return;
Sebastien Moretti's avatar
Sebastien Moretti committed
777
778
779
780
781
782
783
784
785
786
}


########################################################


##################### NCBI requests #####################

#Prot ACC -> PUID
sub blastPAcc2PGI{
787
    my ($blastHit) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
788

789
790
791
    my $protGI = '';
    my $count  = 0;
    GET_PUI:
Sebastien Moretti's avatar
Sebastien Moretti committed
792
    for(my $rep=0;$rep <= 4; $rep++){
793
        $count++;
794
795
#        system("wget -q -O $cache/${date}_${blastHit}protGi.tmp 'http://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?db=protein&term=${blastHit}[pacc]&retmode=xml&tool=ProtoGene&email=smoretti\@unil.ch'");
        system("wget -q -O $cache/${date}_${blastHit}protGi.tmp 'http://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?db=protein&term=${blastHit}&retmode=xml&tool=ProtoGene&email=smoretti\@unil.ch'");
796
797
798
        open(my $GIP, '<', "$cache/${date}_${blastHit}protGi.tmp");
        PROT_GI:
        while(<$GIP>){
799
            if ( $_ =~ /^.*<Id>(\d+)<\/Id>.*$/ ){
800
801
802
803
804
805
806
                $protGI = $1;
                last PROT_GI; #OK if only one PUID
            }
        }
        close $GIP;
        $rep = $rep-15 if ( -z "$cache/${date}_${blastHit}protGi.tmp" && $count==1 );
        last GET_PUI if ( $protGI =~ /^\d+$/ );
Sebastien Moretti's avatar
Sebastien Moretti committed
807
    }
808
    unlink("$cache/${date}_${blastHit}protGi.tmp") if ( $tmp == 0 || $protGI ne '' );
Sebastien Moretti's avatar
Sebastien Moretti committed
809
810
811
812
813
814

    return($protGI);
}

#PUID -> nt GIs
sub protGI2NTGIs{
815
    my ($protGI, $blastHit) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
816

817
818
819
820
    my $ntGIs  = '';
    my $geneID = '';
    my $count  = 0;
    GET_NTUI:
Sebastien Moretti's avatar
Sebastien Moretti committed
821
    for(my $rep=0;$rep <= 4; $rep++){
822
        $count++;
Sebastien Moretti's avatar
Sebastien Moretti committed
823
        #nt GIs
824
        system("wget -q -O $cache/${date}_${blastHit}nucleo.tmp 'http://eutils.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=protein&db=nuccore,nucleotide,gene&id=$protGI&retmode=xml&tool=ProtoGene&email=smoretti\@unil.ch'");
825
826
827
828
829
830
831
832
833
834
835
        open(my $GIN, '<', "$cache/${date}_${blastHit}nucleo.tmp");
        my $flag = 0;
        NT_GI:
        while(<$GIN>){
            if ( $_   =~ /<LinkName>protein_nuc[a-z]+<\/LinkName>/ ){ #for nuccleotide and nuccore
                $flag = 1;
            }
            elsif ( $_ =~ /\<LinkName\>protein_gene\<\/LinkName\>/ ){
                $flag  = 2;
            }
            elsif ( $flag==1 && $_ =~ /^.*\<Id\>(\d+)\<\/Id\>.*$/ ){
836
837
838
                my $match = $1;
                $ntGIs  .= ",$match" if ( $ntGIs  ne '' && $ntGIs  !~ /,$match,/ && $ntGIs  !~ /^$match,/ );
                $ntGIs   = $match    if ( $ntGIs  eq '' );
839
840
            }
            elsif ( $flag==2 && $_ =~ /^.*\<Id\>(\d+)\<\/Id\>.*$/ ){
841
842
843
                my $match = $1;
                $geneID .= ",$match" if ( $geneID ne '' && $geneID !~ /,$match,/ && $geneID !~ /^$match,/ );
                $geneID  = $match    if ( $geneID eq '' );
844
845
846
847
848
            }
        }
        close $GIN;
        $rep = $rep-15 if ( -z "$cache/${date}_${blastHit}nucleo.tmp" && $count==1 );
        last GET_NTUI if ( $ntGIs ne '' || $geneID ne '' );
Sebastien Moretti's avatar
Sebastien Moretti committed
849
    }
850
    unlink("$cache/${date}_${blastHit}nucleo.tmp") if ( $tmp==0 || ($ntGIs ne '' || $geneID ne '') );
Sebastien Moretti's avatar
Sebastien Moretti committed
851

852
    return($ntGIs, $geneID);
Sebastien Moretti's avatar
Sebastien Moretti committed
853
854
855
}

sub geneID2Chr{
856
    my ($geneID, $blastHit) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
857

858
859
860
861
    my $chr   = '';
    my ($amont, $aval) = ('', '');
    my $count = 0;
    GET_CHR:
Sebastien Moretti's avatar
Sebastien Moretti committed
862
    for(my $rep=0;$rep <= 8; $rep++){
863
        $count++;
Sebastien Moretti's avatar
Sebastien Moretti committed
864
        #Gene Acc
865
866
#        system("wget -q -O $cache/${date}_${blastHit}gene.tmp 'http://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=gene&id=$geneID&retmode=xml&tool=ProtoGene&email=smoretti\@unil.ch'");
        system("wget -q -O $cache/${date}_${blastHit}gene.tmp 'http://eutils.ncbi.nlm.nih.gov/entrez/eutils/esummary.fcgi?db=gene&id=$geneID&retmode=xml&tool=ProtoGene&email=smoretti\@unil.ch'");
867
868
869
870
871
872
873
        open(my $GACC, '<', "$cache/${date}_${blastHit}gene.tmp");
        my $flag = 0;
        GENE_CHR:
        while(<$GACC>){
            if ( $_  =~ /\<ERROR\>Empty id list \- nothing todo\<\/ERROR\>/ && $flag==0 ){
                $rep = $rep-15 if ( $count==1 );
            }
874
875
876
#            if ( $_   =~ /\<Entrezgene_locus\>/ && $flag==0 ){
            if ( $_   =~ /<Item Name="GenomicInfoType" Type="Structure">/ && $flag==0 ){
               $flag = 1;
877
            }
878
879
880
881
882
883
884
#            elsif ( $_ =~ /\<Gene-commentary_type value=\"genomic\"\>/ && $flag==1 ){
#                $flag  = 2;
#            }
#            elsif ( $_ =~ /\<Gene-commentary_type value=/ && $flag==2 ){
#                last GENE_CHR;
#            }
#            elsif ( $_ =~ /\<Gene-commentary_accession\>([\w\_\-\.]+)\<\/Gene-commentary_accession\>/ && $flag==2 ){
Sebastien Moretti's avatar
Sebastien Moretti committed
885
            elsif ( $_ =~ /<Item Name="ChrAccVer" Type="String">([^<]+?)\.?\d*<\/Item>/ && $flag==1 ){
886
887
888
                $chr   = $1;
                $flag  = 3;
            }
889
890
#            elsif ( $_ =~ /\<Seq-interval_from\>(\d+)\<\/Seq-interval_from\>/ && $flag==3 ){
            elsif ( $_ =~ /<Item Name="ChrStart" Type="Integer">(\d+)<\/Item>/ && $flag==3 ){
891
892
893
                $amont = $1;
                $flag  = 4;
            }
894
895
#            elsif ( $_ =~ /\<Seq-interval_to\>(\d+)\<\/Seq-interval_to\>/ && $flag==4 ){
            elsif ( $_ =~ /<Item Name="ChrStop" Type="Integer">(\d+)<\/Item>/ && $flag==4 ){
896
897
898
899
900
901
902
                $aval  = $1;
                last GENE_CHR;
            }
        }
        close $GACC;
        $rep = $rep-15 if ( -z "$cache/${date}_${blastHit}gene.tmp" && $count==1 );
        last GET_CHR   if ( $chr ne '' && $amont =~ /^\d+$/ && $aval =~ /^\d+$/ );
Sebastien Moretti's avatar
Sebastien Moretti committed
903
    }
904
905
906
907
    unlink("$cache/${date}_${blastHit}gene.tmp") if ( $tmp == 0 || $chr ne '' );
    if ( $amont eq '' || $aval eq '' ){
        $amont = '';
        $aval = '';
Sebastien Moretti's avatar
Sebastien Moretti committed
908
    }
909
910
911
912
913
914
915
916
917
    elsif ( $amont ne '' && $aval ne '' ){
        if ( $amont > $aval ){
            $amont = $amont+5000;
            $aval  = $aval-5000;
        }
        else{
            $amont = $amont-5000;
            $aval  = $aval+5000;
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
918
    }
919
920
    $amont = 1 if ( $amont ne '' && $amont <= 0 );
    $aval  = 1 if ( $aval ne ''  && $aval <= 0 );
Sebastien Moretti's avatar
Sebastien Moretti committed
921

922
    return($chr, $amont, $aval);
Sebastien Moretti's avatar
Sebastien Moretti committed
923
924
925
}

sub downloadSeqFromGIs{
926
    my ($cache, $date, $amont, $aval, @acc) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
927

928
    GET_SEQ:
Sebastien Moretti's avatar
Sebastien Moretti committed
929
    for(my $a=0; $a<=$#acc; $a++){
930
931
932
933
        my $whatNumber = 0;
        if ( $amont =~ /^\d+$/ && $aval =~ /^\d+$/ ){
            GET_CHR_SEQ:
            for(my $rep=0;$rep <= 4; $rep++){
934
                system("wget -q -O $cache/${acc[$a]}-${amont}.fas 'http://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=nucleotide&id=${acc[$a]}&rettype=fasta&retmode=text&from=$amont&to=$aval&tool=ProtoGene&email=smoretti\@unil.ch'") if ( !-e "$cache/${acc[$a]}-${amont}.fas" || -z "$cache/${acc[$a]}-${amont}.fas" || $cache eq 'none' || ($cache eq 'update' && -M "$cache/${acc[$a]}-${amont}.fas" > $cacheStorageTime ) );
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
                open(my $CIBLE, '<', "$cache/${acc[$a]}-${amont}.fas");
                my $counter = 0;
                my $lines   = 0;
                CHECK_CHR_SEQ:
                while(<$CIBLE>){
                    $lines++              if ( $_ !~ /^>/ );
                    $counter = $counter+2 if ( $counter==1 && $_ !~ /^\w/ && $lines==1 && $_ !~ /^>/ );
                    $counter++            if ( $_ =~ /^>/ );
                    $counter = $counter+2 if ( $_ !~ /^>/ && ($_ =~ /Error:/ || $_ =~ /[<>]/) );
#                   last CHECK_CHR_SEQ    if ( $_ !~ /^>/ && $counter==1 && $_ =~ /^\w/ );
                }
                close $CIBLE;
                $whatNumber++;
                last GET_CHR_SEQ if ( $whatNumber==20 );
                if ( $counter != 1 ){
                    $rep = $rep-1;
                    unlink("$cache/${acc[$a]}-${amont}.fas");
                }
            }
        }
        else{
            GET_OTHER_SEQ:
            for(my $rep=0;$rep <= 4; $rep++){
958
                system("wget -q -O $cache/${acc[$a]}.fas 'http://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=nucleotide&id=${acc[$a]}&rettype=fasta&retmode=text&tool=ProtoGene&email=smoretti\@unil.ch'") if ( !-e "$cache/${acc[$a]}.fas" || -z "$cache/${acc[$a]}.fas" || $cache eq 'none' || ($cache eq 'update' && -M "$cache/${acc[$a]}.fas" > $cacheStorageTime ) );
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
                open(my $CIBLE, '<', "$cache/${acc[$a]}.fas");
                my $counter = 0;
                my $lines   = 0;
                CHECK_OTHER_SEQ:
                while(<$CIBLE>){
                    $lines++              if ( $_ !~ /^>/ );
                    $counter = $counter+2 if ( $counter==1 && $_ !~ /^\w/ && $lines==1 && $_ !~ /^>/ );
                    $counter++            if ( $_ =~ /^>/ );
                    $counter = $counter+2 if ( $_ !~ /^>/ && ($_ =~ /Error:/ || $_ =~ /[<>]/) );
                }
                close $CIBLE;
                $whatNumber++;
                last GET_OTHER_SEQ if ( $whatNumber==20 );
                if ( $counter != 1 ){
                    $rep=$rep-1;
                    unlink("$cache/${acc[$a]}.fas");
                }
            }
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
978
    }
979
980

    return;
Sebastien Moretti's avatar
Sebastien Moretti committed
981
982
983
}

sub downloadSeq{
984
    my ($cache, $date, $amont, $aval, @acc) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
985

986
987
988
989
    my $cp   = 0;
    my $from = $amont;
    my $to   = $aval;
    DOWNL_SEQ:
Sebastien Moretti's avatar
Sebastien Moretti committed
990
    for(my $a=0; $a<=$#acc; $a++){
991
992
993
        my $pacc2puid = $acc[$a];
        if ( $pacc2puid !~ /^[NAX][CGTWZM]_/ ){
        #pacc = primary acc NOT prot acc !   #265666 -> S55551
994
            system("wget -q -O $cache/${date}_${acc[$a]}.gui 'http://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?&db=nucleotide&term=${pacc2puid}[pacc]&tool=ProtoGene&email=smoretti\@unil.ch'");
995
996
997
998
999
1000
            open(my $GUI, '<', "$cache/${date}_${acc[$a]}.gui");
            DOWNL:
            while(<$GUI>){
                if ( $_ =~ /\<Id\>(\d+)\<\/Id\>/ ){
                    $pacc2puid = $1;
                    last DOWNL;