ProtoGene.pl 57.4 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;

15
16
17
18
19
use File::Which qw(which);                 # Locate external executable programs in the PATH
use Time::localtime;                       # Use localtime+PID for a pseudo-uniq temp file name
use Getopt::Long;                          # Options specifications
use File::Copy qw(move);                   # Avoid external 'mv' command usage
use LWP::Simple;                           # To test gigablaster availability
Sebastien Moretti's avatar
Sebastien Moretti committed
20

21
use Mail::Send;                            # Send warnings and errors files by e-mail ==> only if the $userEMail variable is defined
Sebastien Moretti's avatar
Sebastien Moretti committed
22

23
24
25
26
use lib '/mnt/common/share/ProtoGene/';    # Local path for ProtoGene's own perl modules
use Exonerate;                             # Exonerate runner, parser, ...
use Views;                                 # Non-text outputs, e.g. HTML/CSS
#use CheckOutput;                         # Check output for cds consistancy with query
Sebastien Moretti's avatar
Sebastien Moretti committed
27
28
29



30
################## CONFIGURATION ##################
31
#$ENV{'PATH'}        .= ':/mnt/local/bin/:./';                               # Additional path for executables
32

33
34
35
36
my $cachePath        = '/scratch/fhgfs/tcoffee/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
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53

### BLAST parameters ###
my $blast_param = { 'evalue'    => 0.05,
                    'matrix'    => 'BLOSUM62',
                    'filter'    => 'Off',
                    'species'   => 'All_organisms',

                    'db1'       => 'refseq_protein', # NCBI RefSeq protein
                    'identity1' => 100,
                    'cover1'    => 100,
                    'db2'       => 'nr', # NCBI nr protein
                    'identity2' =>  95,
                    'cover2'    =>  95,

                    'nbesthits' => 10,
                    'core'      => 1,
                  };
54
###################################################
Sebastien Moretti's avatar
Sebastien Moretti committed
55
56


57
my $VERSION  = '4.2.0';
Sebastien Moretti's avatar
Sebastien Moretti committed
58

59
my $webblast_exe  = '/mnt/common/share/ProtoGene/webblast.pl';
60
61
my $blast_exe     = 'blastall';        # Or wu-blastall for Wu-BLAST; for local blast usage
my $exonerate_exe = 'exonerate';       # Exonerate 1.0 because current parser only works with this version
62
63
64


################## Option management
65
66
67
my ($msa, $revtrans, $pep, $hideBOJ, $run_name, $template, $lim, $cache) = ('', 0, 0, 0, '', '', 0, 'update');
my ($debug, $tmp)                                                        = (0, 0);
my ($db, $species, $local, $giga)                                        = ($blast_param->{'db1'}, $blast_param->{'species'}, 0, 0);
Sebastien Moretti's avatar
Sebastien Moretti committed
68
my %opts = ('msa|in=s'        => \$msa,        # Input sequences
69
            'revtrans:s'      => \$revtrans,   # Use to reverse-translate sequences with no match
Sebastien Moretti's avatar
Sebastien Moretti committed
70
71
72
73
74
            '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
75
            'cache=s'         => \$cache,      # Cache behavior
Sebastien Moretti's avatar
Sebastien Moretti committed
76
77
78

            'orgm|species=s'  => \$species,    # Organism(s) to blast against
            'db|database=s'   => \$db,         # Database to blast against
79
            'local'           => \$local,      # Use to specify a local db query, defined in $local_db
Sebastien Moretti's avatar
Sebastien Moretti committed
80
81
82
83
84
85
86
87
88
89
            '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
90
           );
Sebastien Moretti's avatar
Sebastien Moretti committed
91
92
my $test_option_values = Getopt::Long::GetOptions(%opts);

93
94

################## Short help message
95
if ( !$test_option_values || ($msa eq '' && $cache ne 'empty' && $cache ne 'old') ){
96
97
98
    print {*STDERR} "\n\tCannot open the MSA file in FASTA format
\tTry:  $0 --msa=path_of_the_fasta_msa_file [Options]

99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
\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 '$blast_param->{'species'}'
\t         --db=nr, pdb, swissprot, refseq_protein
\t                default is '$blast_param->{'db1'}'
\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 reverse-translates sequences with no
\t                blast hit, in IUB (IUPAC) depiction code
\t                They are removed from the alignement by default
\t         --pep  adds 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         --debug   prints extra information when running
\t         --version prints version information
\t         --help    prints a full help message\n\n";
122
    exit(1);
Sebastien Moretti's avatar
Sebastien Moretti committed
123
}
124
125


126
################## Check cache directory accessibility
Sebastien Moretti's avatar
Sebastien Moretti committed
127
128
129
if ( -d $cache ){
    $cachePath = $cache;
}
130
checkCacheAccessibility($cachePath);
131

132
133
# Cache management
$cache = cacheManagement($cache, $cachePath, 0) || '.';
134

135
136
# Remove old files from the cache directory everytime ProtoGene is running
cacheManagement('old', $cachePath, 1);
137
138


139
140

################## Check external software in the PATH
141
checkExternalSoftware($webblast_exe, $exonerate_exe);
142
143
if ( $local ){
    checkExternalSoftware($blast_exe);
Sebastien Moretti's avatar
Sebastien Moretti committed
144
145
146
147
}



148
149
150
################## Check input files
# Open and check template file
my ($template_NT, $template_AA, $template_SEQ);
151
#TODO: template is fasta only now
152
153
($template_NT, $template_AA, $template_SEQ) = checkTemplate($template) if ( $template );

Sebastien Moretti's avatar
Sebastien Moretti committed
154

Sebastien Moretti's avatar
Sebastien Moretti committed
155
156
157
158
159
# Open and Check the input fasta file
my $originalMSA    = $msa;
my $converted      = 0; # Accept only fasta format
#TODO: Do a better file converter and identifier is msa is not fasta
#($msa, $converted) = isFastaFile($msa, $converted);
Sebastien Moretti's avatar
Sebastien Moretti committed
160

161
my $fasta_checker = -1;
162
my (@original_names, @original_seq); #Use 2 lists in parallel : original_names, seq
Sebastien Moretti's avatar
Sebastien Moretti committed
163
164
open( my $MSA, '<', "$msa" ) or die "\n\tCannot open your MSA file\n\n";
READ_MSA:
165
while(<$MSA>){
Sebastien Moretti's avatar
Sebastien Moretti committed
166
167
    next READ_MSA if ( $_ =~ /^#/ );

168
    if ( $_ =~ /^>/ ){
169
170
        $fasta_checker++;
        my $name = $_;
Sebastien Moretti's avatar
Sebastien Moretti committed
171
172
173
        $name    =~ s{\r\n}{}g; #Remove return lines from windows OS '^M'
        $name    =~ s{\r}{\n}g;
        $name    =~ s{^>[ \t]+}{>};
174
        chomp($name);
Sebastien Moretti's avatar
Sebastien Moretti committed
175
176

        push @original_names, $name;
Sebastien Moretti's avatar
Sebastien Moretti committed
177
    }
Sebastien Moretti's avatar
Sebastien Moretti committed
178
179
180
181
182
183
184
185
186
187
    elsif ( $_ =~ /[A-Z\-\.]/i ){
        my $sequence = $_;
        $sequence    =~ s{\r\n}{}g;
        $sequence    =~ s{\r}{\n}g;
        $sequence    =~ s{\.}{-}g; #for msa with '.' as gap
        #TODO because exonerate cannot work with selenocystein
        $sequence    =~ s{[BJOUZ]}{X}ig; #U here for selenocystein
        $sequence    =~ s{[^A-Za-z\-\*\n]}{}g; #Remove all the non-gap or non-alphabetic characters from the seq
        chomp($sequence);

Sebastien Moretti's avatar
Sebastien Moretti committed
188
        #fasta sequence on 1 line
189
        if ( !exists($original_seq[$fasta_checker]) ){
Sebastien Moretti's avatar
Sebastien Moretti committed
190
            push @original_seq, $sequence;
191
192
        }
        else {
Sebastien Moretti's avatar
Sebastien Moretti committed
193
            $original_seq[$fasta_checker] .= $sequence;
194
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
195
196
    }
}
197
close $MSA;
Sebastien Moretti's avatar
Sebastien Moretti committed
198
199

#Are there FASTA sequences ?
200
if ( $fasta_checker == -1 ){
Sebastien Moretti's avatar
Sebastien Moretti committed
201
    failure();
Sebastien Moretti's avatar
Sebastien Moretti committed
202
203
    print {*STDERR} "\tThe MSA file does NOT seem to be a protein FASTA format
\tSee an example of FASTA format <a href='http://en.wikipedia.org/wiki/FASTA_format' target='_blank'>here</a>\n\n";
204
    exit(1);
Sebastien Moretti's avatar
Sebastien Moretti committed
205
}
206
elsif ( $lim >0 && $fasta_checker > $lim ){
Sebastien Moretti's avatar
Sebastien Moretti committed
207
    failure();
208
    print {*STDERR} "\tThe FASTA file is too large, try with less than $lim sequences\n\tor split your file\n\n";
209
210
    exit(1);
}
211
elsif ( exists( $original_seq[0] ) && $original_seq[0] =~ /[acgtu]/i ){
212
    my $first_seq = $original_seq[0];
Sebastien Moretti's avatar
Sebastien Moretti committed
213
214
215
216
217
218
    #Check if sequences are amino acids and not nucleotides ONLY on the 1st sequence
    my $a         = ($first_seq =~ s{[aA]}{}g)               || 0;
    my $c         = ($first_seq =~ s{[cC]}{}g)               || 0;
    my $g         = ($first_seq =~ s{[gG]}{}g)               || 0;
    my $t         = ($first_seq =~ s{[tTuU]}{}g)             || 0;
    my $non       = ($first_seq =~ s{[^aAcCgGtTuUXxNn-]}{}g) || 0;
219
    if ( ($a+$c+$g+$t) >= (($a+$c+$g+$t+$non)*80/100) ){
Sebastien Moretti's avatar
Sebastien Moretti committed
220
        failure();
Sebastien Moretti's avatar
Sebastien Moretti committed
221
222
        print {*STDERR} "\tYour sequences look already to be nucleotides
\tThe goal of this program is to turn AMINO ACID alignments into CDS nucleotide alignments\n\n";
223
224
        exit(1);
    }
Sebastien Moretti's avatar
Sebastien Moretti committed
225
226
227
228
}
undef $fasta_checker;


Sebastien Moretti's avatar
Sebastien Moretti committed
229
# Check GigaBlaster status if used
230
if ( $giga==1 ){
231
    $local = 0;
Sebastien Moretti's avatar
Sebastien Moretti committed
232
    $local = 2  if ( head('http://www.igs.cnrs-mrs.fr/public_g/remoteblast.cgi') );
233
234
}

Sebastien Moretti's avatar
Sebastien Moretti committed
235

Sebastien Moretti's avatar
Sebastien Moretti committed
236
# Temporary file extension
237
my $date = sprintf("%d-%02d-%02d_%02dh%02d", localtime->year() + 1900, localtime->mon() + 1, localtime->mday(), localtime->hour, localtime->min);
Sebastien Moretti's avatar
Sebastien Moretti committed
238
239


240

Sebastien Moretti's avatar
Sebastien Moretti committed
241
#Start main program with version # of programs and list original queries
242
print "\n\t   Protogene\t$VERSION\t$date\n\n";
243

244
open(my $EXONERATEISHERE, "$exonerate_exe --version |");
245
my $IsExonerateHere = <$EXONERATEISHERE>;
246
close $EXONERATEISHERE;
Sebastien Moretti's avatar
Sebastien Moretti committed
247
248
print "\t   $IsExonerateHere\n";
undef $IsExonerateHere;
249
undef $VERSION;
250

Sebastien Moretti's avatar
Sebastien Moretti committed
251
for(my $m=0; $m<=$#original_names; $m++){
252
    printf("\t>%-3s  is  %s\n", $m, $original_names[$m]);
Sebastien Moretti's avatar
Sebastien Moretti committed
253
}
Sebastien Moretti's avatar
Sebastien Moretti committed
254
$date .= "_$$"; #Add PID to be more uniq
Sebastien Moretti's avatar
Sebastien Moretti committed
255
256
257
258



# run_name implementation for web server
259
$originalMSA = $run_name if ( $run_name ne '' );
Sebastien Moretti's avatar
Sebastien Moretti committed
260

261
262
unlink("$originalMSA.cds", "$originalMSA.cdsP", "$originalMSA.cdsP.html",
       "$originalMSA.out", "$originalMSA.boj",  "$originalMSA.log");
Sebastien Moretti's avatar
Sebastien Moretti committed
263
264


265
# Build the sequences, from the alignment, to perform blast search
266
EACH_SEQ:
267
268
269
270
271
272
for(my $r=0; $r<=$#original_names; $r++){
    my $fasta_header = $original_names[$r];
    my $short_name   = $original_names[$r];
    $fasta_header    =~ s{^>}{};               # Full FASTA header
    $short_name      =~ s{^([^\s\t]+).*$}{$1}; # Short name without description

Sebastien Moretti's avatar
Sebastien Moretti committed
273

Sebastien Moretti's avatar
Sebastien Moretti committed
274
    my @equivalent_blast_hits; #Multiple BLAST hits with the same highest e-value/score
275
276

    if ( exists($template_NT->{$short_name}) && $template_NT->{$short_name} ne '' ){
Sebastien Moretti's avatar
Sebastien Moretti committed
277
278
279
        # First try nucleotide template
        push @equivalent_blast_hits, 'My_own_seq';
        print "\n\n$r done\n\nNucleotide template for $r:\t$template_NT->{$short_name}\n\n*******************************************************************\n\n";
Sebastien Moretti's avatar
Sebastien Moretti committed
280
    }
281
    elsif ( exists($template_AA->{$short_name}) && $template_AA->{$short_name} ne '' ){
Sebastien Moretti's avatar
Sebastien Moretti committed
282
283
284
        # Second try amino acid/BLASTp template
        push @equivalent_blast_hits, $template_AA->{$short_name};
        print "\n\n$r done\n\nBlastP template for $r:\t$template_AA->{$short_name}\n\n*******************************************************************\n\n";
Sebastien Moretti's avatar
Sebastien Moretti committed
285
286
    }
    else{
287
    # else send them to webblast
Sebastien Moretti's avatar
Sebastien Moretti committed
288
289
290
291
292
293
294
        # Prepare sequence $r for BLAST
        open(my $READY2BLAST, '>', "$cache/$date.seq2blast.fas.$r");
        my $noGap = $original_seq[$r];
        $noGap    =~ s{\-}{}g; #remove gaps
        print {$READY2BLAST} ">$r\n$noGap\n";
        close $READY2BLAST;

295
296
297
298
299
        if ($r==0){
            # Show BLAST parameters only for the first BLAST run
            display_blast_param();
        }
        run_BLAST("$cache/$date.seq2blast.fas.$r", $local, $blast_param->{'db1'}, $blast_exe, "$cache/$date.blastp.$r");
Sebastien Moretti's avatar
Sebastien Moretti committed
300

301
302
303

        # Get the BLASTp hit(s)  acc number
        open(my $BLASTP, '<', "$cache/$date.blastp.$r");
304
        while(<$BLASTP>){
305
            push @equivalent_blast_hits, $1  if ( $_ =~ /^>\s*\d+@?\w?\w?\w?__([^\s\|\.]+).*$/ ); #Warning: double '_'
306
307
        }
        close $BLASTP;
308
        unlink("$cache/$date.blastp.$r", "$cache/$date.seq2blast.fas.$r") if ( $tmp == 0 || exists($equivalent_blast_hits[0]) );
Sebastien Moretti's avatar
Sebastien Moretti committed
309
        if ( !exists($equivalent_blast_hits[0]) ){
310
            print "\n$r ...\tNo blast result found above thresholds for $fasta_header\n\n";
311
            buildFailureOutputFiles($r, 'No_BLASTp_Result', 'Not_searched');
312
313
314
315
            undef $original_seq[$r];
            undef $original_names[$r];
            next EACH_SEQ;
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
316
317
318
319
    }



320
    #When multiple equivalent blast hits  ==> use, and add, every hits
321
    my ($resultPOS, $resultBOJ) = ('', '');
322
    my @failureStatus = (''); # To always have $failureStatus[0] defined
323
    my @intronStatus  = ('');
324
    HIT_LINK:
Sebastien Moretti's avatar
Sebastien Moretti committed
325
326
    for(my $qq=0; $qq<=$#equivalent_blast_hits; $qq++){
        print "\n$fasta_header  -->  [$equivalent_blast_hits[$qq]]\n";
327
328
329
330


        my @nt_GIs;
        my $intronStep = 0;
331
332
333
        # Get Nt sequence from template
        if ( exists($template_NT->{$short_name}) && $template_NT->{$short_name} ne '' ){
            if ( $template_NT->{$short_name} !~ /^My_Seq$/i ){
334
                download_seq($cache, '', '', $template_NT->{$short_name});
335
336
337
338
339
340
341
342
                @nt_GIs = $template_NT->{$short_name};
            }
            else {
                open(my $MYSEQ, '>', "$cache/My_Seq-$$.fas");
                print {$MYSEQ} ">My_Seq-$$ for $short_name\n".$template_SEQ->{$short_name}."\n";
                close $MYSEQ;
                @nt_GIs = "My_Seq-$$";
            }
343
        }
344
        # Get Nt sequence through prot acc -> prot gi -> nt linked gi -> nt acc
345
        else{
Sebastien Moretti's avatar
Sebastien Moretti committed
346
            # Prot ACC -> PUID
Sebastien Moretti's avatar
Sebastien Moretti committed
347
            my $protGI = blastPAcc2PGI($equivalent_blast_hits[$qq]);
348
            if ( $protGI eq '' ){
Sebastien Moretti's avatar
Sebastien Moretti committed
349
350
                print "\tNo protein (prot GI) link found for $equivalent_blast_hits[$qq] in $fasta_header\n\n";
                @failureStatus = ('PUI_unavailable', $equivalent_blast_hits[$qq]);
351
                next HIT_LINK;
352
            }
Sebastien Moretti's avatar
Sebastien Moretti committed
353

Sebastien Moretti's avatar
Sebastien Moretti committed
354
            # PUID -> nt GIs & GeneID
Sebastien Moretti's avatar
Sebastien Moretti committed
355
            my ($ntGIs, $geneID) = protGI2NTGIs($protGI, $equivalent_blast_hits[$qq]);
356
            if ( $ntGIs eq '' && $geneID eq '' ){
Sebastien Moretti's avatar
Sebastien Moretti committed
357
358
                print "\tNo nucleotide (nt GIs) link found for $equivalent_blast_hits[$qq] in $fasta_header\n\n";
                @failureStatus = ('No_nt_link', $equivalent_blast_hits[$qq]);
359
                next HIT_LINK;
360
361
362
363
            }
            print " linked with [$ntGIs $geneID]\n";


Sebastien Moretti's avatar
Sebastien Moretti committed
364
            # Get transcript and contig seq
Sebastien Moretti's avatar
Sebastien Moretti committed
365
            if ( $ntGIs ne '' ){
Sebastien Moretti's avatar
Sebastien Moretti committed
366
                @nt_GIs = split(/,/, $ntGIs);
Sebastien Moretti's avatar
Sebastien Moretti committed
367
368
                downloadSeqFromGIs($cache, '', '', @nt_GIs);
            }
369

Sebastien Moretti's avatar
Sebastien Moretti committed
370
            # GeneID -> Chr
371
372
373
374
375
376
377
            if ( $geneID ne '' ){
                my @geneIDs = split(/,/, $geneID);
                while(<@geneIDs>){
                    my $geneID = $_;
                    my ($chr, $amont, $aval) = geneID2Chr($geneID, $equivalent_blast_hits[$qq]);
                    print "\n\tNo gene locus found for $equivalent_blast_hits[$qq] with $geneID in $fasta_header\n\n" if ($chr eq '' and $geneID ne '');

Sebastien Moretti's avatar
Sebastien Moretti committed
378
                    # Get Chr seq
379
380
                    if ( $chr ne '' ){
                        download_seq($cache, $amont, $aval, $chr);
Sebastien Moretti's avatar
Sebastien Moretti committed
381
                        $intronStep = 1;
382

383
                        unshift(@nt_GIs, "$chr--$amont-$aval");
384
                    }
385
386
387
388
389
                }
            }
        }


Sebastien Moretti's avatar
Sebastien Moretti committed
390
391
        # Align Nt seq with our protein query seq
        my ($POS, $BOJ) = runExonerate($cache, $date, $r, $original_seq[$r], $fasta_header, @nt_GIs);
392
        next HIT_LINK if ( !defined $POS );
393

Sebastien Moretti's avatar
Sebastien Moretti committed
394
        my ($POSresult, $bestNucleotide) = prepareResults4CDS($POS, $equivalent_blast_hits[$qq], $original_seq[$r], $fasta_header);
395
396
        $resultPOS .= $POSresult if ( $bestNucleotide ne '' && $POSresult ne '' && $resultPOS !~ /_G_$bestNucleotide/ );

397
398
        @intronStatus = ($POS->{0}, $equivalent_blast_hits[$qq], '')              if ( $intronStep==1 && exists($POS->{0}) && $POS->{0} =~ /\:/ );
        @intronStatus = ($POS->{0}, $equivalent_blast_hits[$qq], $bestNucleotide) if ( $intronStep==0 && $bestNucleotide =~ /^NA[CTWZGS]_/ );
399
400
401

        my ($intronLess, $nameLess) = ('', '');
        $intronLess = $POS->{0}       if ( %$BOJ eq 0 && exists($POS->{0}) && $POS->{0} ne '' );
402
        $nameLess   = $bestNucleotide if ( %$BOJ eq 0 && $intronStep==0 && $bestNucleotide =~ /^NA[CTWZGS]_/ );
Sebastien Moretti's avatar
Sebastien Moretti committed
403
        my ($BOJresult, $bestGenomic) = prepareResults4BOJ($BOJ, $equivalent_blast_hits[$qq], $original_seq[$r], $fasta_header, $intronLess, $nameLess);
404
        $resultBOJ .= $BOJresult      if ( $bestGenomic ne '' && $BOJresult ne '' && $resultBOJ !~ /_G_$bestGenomic[ \-]/ );
Sebastien Moretti's avatar
Sebastien Moretti committed
405
406
407
    }


408
    if ( $resultPOS eq '' ){
409
        if ( $failureStatus[0] ne '' ){
410
            buildFailureOutputFiles($r, $failureStatus[1],         $failureStatus[0]);
411
412
        }
        else {
413
            buildFailureOutputFiles($r, $equivalent_blast_hits[0], 'Alignment_failure');
414
        }
415
        next EACH_SEQ;
Sebastien Moretti's avatar
Sebastien Moretti committed
416
    }
417
418
    elsif ( $resultBOJ eq '' ){
        if ( $failureStatus[0] eq 'PUI_unavailable' || $failureStatus[0] eq 'No_nt_link' ){
419
            buildFailureBOJOutputFile($failureStatus[1], $failureStatus[0], $fasta_header, $original_seq[$r]);
420
421
        }
        elsif ( $failureStatus[0] eq '' && $intronStatus[0] ne '' ){
422
            buildIntronlessBOJOutputFile($intronStatus[1], $intronStatus[0], $intronStatus[2], $fasta_header, $original_seq[$r]);
423
424
        }
        else{
Sebastien Moretti's avatar
Sebastien Moretti committed
425
            buildFailureBOJOutputFile($equivalent_blast_hits[0], 'Unavailable', $fasta_header, $original_seq[$r]);
426
        }
427
        createCDSOutputFile($resultPOS, $original_seq[$r], $fasta_header);
Sebastien Moretti's avatar
Sebastien Moretti committed
428
429
    }
    else {
430
        createCDSOutputFile($resultPOS, $original_seq[$r], $fasta_header);
431
        createBOJOutputFile($resultBOJ);
Sebastien Moretti's avatar
Sebastien Moretti committed
432
433
434
435
436
437
438
439
    }


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


Sebastien Moretti's avatar
Sebastien Moretti committed
440
checkAndCleanStderrFiles("$cache/$date.ExonerateError") if ( -e "$cache/$date.ExonerateError" );
Sebastien Moretti's avatar
Sebastien Moretti committed
441
checkOutputFiles();
Sebastien Moretti's avatar
Sebastien Moretti committed
442
unlink("$msa") if ( $converted==1 );
443
unlink 'blast_result.txt';
Sebastien Moretti's avatar
Sebastien Moretti committed
444
my @tmpFiles = glob($cache."/$date*");
Sebastien Moretti's avatar
Sebastien Moretti committed
445
if ( exists($tmpFiles[0]) ){
Sebastien Moretti's avatar
Sebastien Moretti committed
446
447
    mkdir("$cache/$date.TMP");
    move("$_", "$cache/$date.TMP") while(<@tmpFiles>);
Sebastien Moretti's avatar
Sebastien Moretti committed
448
449
450
451
452
453
454
}


#Ouput Checker


#STDOUT Log output for our server
Sebastien Moretti's avatar
Sebastien Moretti committed
455
if ( -s "$originalMSA.cds" || -s "$originalMSA.out" ){
Sebastien Moretti's avatar
Sebastien Moretti committed
456
    print "\n\nTERMINATION STATUS: SUCCESS\n";
457
    print 'OUTPUT RESULTS', "\n";
458
    print "    #### File Type=        MSA Format= fasta_CDS Name= $originalMSA.cds\n"
459
        if ( -s "$originalMSA.cds" );
460
    print "    #### File Type=        MSA Format= fasta_CDS+Query Name= $originalMSA.cdsP\n"
461
        if ( -s "$originalMSA.cdsP" );
462
    print "    #### File Type=        MSA Format= colored_fasta_CDS Name= $originalMSA.cdsP.html\n"
463
        if ( Views::Html("$originalMSA.cdsP") && -s "$originalMSA.cdsP.html" );
464
    print "    #### File Type=        MSA Format= Rejected_seq Name= $originalMSA.out\n"
465
        if ( -s "$originalMSA.out" );
466
    print "    #### File Type=        MSA Format= fasta_Exon-Boundaries Name= $originalMSA.boj\n"
467
        if ( -s "$originalMSA.boj" && -s "$originalMSA.cds" && $hideBOJ==0 );
Sebastien Moretti's avatar
Sebastien Moretti committed
468
469

    unlink "$cache/web_tempo.result";
Sebastien Moretti's avatar
Sebastien Moretti committed
470
471
}
else{
Sebastien Moretti's avatar
Sebastien Moretti committed
472
    failure();
473
    exit(1);
Sebastien Moretti's avatar
Sebastien Moretti committed
474
475
}

476
exit(0);
Sebastien Moretti's avatar
Sebastien Moretti committed
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


=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

509
=item I<--cache>=none, update, use, own_path_directory, old, empty
Sebastien Moretti's avatar
Sebastien Moretti committed
510
511
512
513
514
515
516

=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

517
518
=item own_path_directory: use my own directory, and its files

Sebastien Moretti's avatar
Sebastien Moretti committed
519
520
=item old: remove the old files in the cache directory

521
=item empty: empty completely the cache directory
Sebastien Moretti's avatar
Sebastien Moretti committed
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542

=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 !

543
=item Standard Perl modules B<lib>, B<strict>,  B<warnings>,  B<diagnostics>, B<Carp> are required
Sebastien Moretti's avatar
Sebastien Moretti committed
544

545
=item as well as some other current ones : B<Getopt::Long>,  B<File::Which>,  B<File::Copy>, B<Mail::Send>, B<Time::localtime>, B<LWP::Simple>
Sebastien Moretti's avatar
Sebastien Moretti committed
546
547
548

=item -

549
=item I<exonerate> from http://www.ebi.ac.uk/~guy/exonerate/
Sebastien Moretti's avatar
Sebastien Moretti committed
550

551
=item I<blast> from http://www.ncbi.nlm.nih.gov/BLAST/download.shtml or http://blast.wustl.edu/
Sebastien Moretti's avatar
Sebastien Moretti committed
552
553
554
555
556
557
558

=back

=head1 VERSION

=over 8

559
=item version 4.2.0
Sebastien Moretti's avatar
Sebastien Moretti committed
560

561
=item on Aug 08th, 2013
Sebastien Moretti's avatar
Sebastien Moretti committed
562
563
564
565
566
567
568
569
570

=back

=head1 AUTHORS

=over 8

=item Sebastien MORETTI

Sebastien Moretti's avatar
Sebastien Moretti committed
571
=item moretti.sebastien [AT] gmail.com
Sebastien Moretti's avatar
Sebastien Moretti committed
572

573
=item Vital-IT computing center
Sebastien Moretti's avatar
Sebastien Moretti committed
574

575
=item Swiss Institute of Bioinformatics
Sebastien Moretti's avatar
Sebastien Moretti committed
576

577
=item Lausanne, Switzerland
Sebastien Moretti's avatar
Sebastien Moretti committed
578

579
=item http://www.vital-it.ch/
Sebastien Moretti's avatar
Sebastien Moretti committed
580
581
582
583
584
585
586
587

=back

=cut


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

588
# Failure declaration
Sebastien Moretti's avatar
Sebastien Moretti committed
589
sub failure {
590
591
592
    print {*STDERR} "\n\nFATAL\n\nTERMINATION STATUS: FAILURE\n\n";
    return;
}
Sebastien Moretti's avatar
Sebastien Moretti committed
593

Sebastien Moretti's avatar
Sebastien Moretti committed
594
# Check external programs presence in the PATH
595
sub checkExternalSoftware {
Sebastien Moretti's avatar
Sebastien Moretti committed
596
    for my $exe ( @_ ){
597
        if ( ! which($exe) && !-x $exe ){
Sebastien Moretti's avatar
Sebastien Moretti committed
598
599
600
601
            failure();
            print {*STDERR} "\t'$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
602
    }
Sebastien Moretti's avatar
Sebastien Moretti committed
603
    return;
Sebastien Moretti's avatar
Sebastien Moretti committed
604
605
}

606
607
# Check cache directory accessibility: Test and change rights, if required, of the cache directory
sub checkCacheAccessibility {
608
    my ($cacheDir) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
609

610
    if ( -d "$cacheDir/" && -w "$cacheDir/" && -x "$cacheDir/" && -r "$cacheDir/" ){
611
        # Good
Sebastien Moretti's avatar
Sebastien Moretti committed
612
    }
613
614
615
    elsif ( -d "$cacheDir/" ){
        if ( -o "$cacheDir/" ){
            chmod 0754, "$cacheDir/";
616
        }
617
        else{
618
619
620
            failure();
            print {*STDERR} "\tPermissions do NOT seem to be valid for the '$cacheDir' cache directory\n\n";
            exit(2);
621
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
622
623
    }
    else{
624
        mkdir "$cacheDir/"; # FIXME: mkdir is not recursive !
625
        if ( -o "$cacheDir/" ){
626
            chmod 0754, "$cacheDir/"; #0754 is in octal, so 754 or '0754' are not good !
627
628
        }
        else{
629
630
631
            failure();
            print {*STDERR} "\tPermissions do NOT seem to be valid for the '$cacheDir' cache directory\n\n";
            exit(2);
632
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
633
    }
634

635
    return;
Sebastien Moretti's avatar
Sebastien Moretti committed
636
637
}

638
# Cache Behavior Management
Sebastien Moretti's avatar
Sebastien Moretti committed
639
sub cacheManagement{
640
    my ($cache, $cacheDir, $inScript) = @_;
641

642
    if ( $cache eq 'empty' ){
643
        unlink <$cacheDir/*.fas>;
Sebastien Moretti's avatar
Sebastien Moretti committed
644
        unlink <$cacheDir/*.ExonerateError>;
645
646
        unlink("$cacheDir/webblast.log", "$cacheDir/web_tempo.result", "$cacheDir/debug.tempo");
        print "\n\tcache directory [$cacheDir] is empty\n\n";
647
        exit 0;
Sebastien Moretti's avatar
Sebastien Moretti committed
648
    }
649
    elsif ( $cache eq 'old' ){
650
651
652
653
654
        unlink    grep { -M $_ > $cacheStorageTime }    glob($cacheDir.'/*.fas'); # Grep list only with $cacheStorageTime days old files

        if ( $inScript==0 ){
            print "\n\tcache directory [$cacheDir] is empty of its old files\n\n";
            exit 0;
655
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
656
    }
657
    elsif ( $cache eq 'update' || $cache eq 'use' ){
658
659
        # DEFAULT: Limit cached files to less than $cacheStorageTime days old ones for 'update'
        $cache = $cacheDir;
Sebastien Moretti's avatar
Sebastien Moretti committed
660
    }
661
    elsif ( $cache eq 'none' ){
662
        # Do not use cache
Sebastien Moretti's avatar
Sebastien Moretti committed
663
    }
664
665
666
667
    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 $cache );
Sebastien Moretti's avatar
Sebastien Moretti committed
668
669
    }
    else {
670
671
672
        failure();
        print {*STDERR} "\n\tWrong cache argument, or your cache folder '$cacheDir' doesn't seem to exist\n\n";
        exit(3);
Sebastien Moretti's avatar
Sebastien Moretti committed
673
674
675
676
677
678
679
680
681
682
    }

    return($cache);
}

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


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

Sebastien Moretti's avatar
Sebastien Moretti committed
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
# Check input aln format
sub isFastaFile{
    my ($msa, $converted) = @_;

    if ( -s "$msa" ){ # Check is the file exists and is not empty
        # Is FASTA format ?
        my $firstLine = `head $msa`;
        return($msa, 0) if ( $firstLine =~ /^>/ );

        # Convert no fasta format to fasta with readseq
        my $reformat = which('readseq') || '';
        if ( $reformat ne '' ){
            system("$reformat -a -f8 $msa > $msa.007")==0 or do {failure(); print {*STDERR} "\n\tEnable to convert your file to fasta format\n\n";};
            $msa      .= '.007';
            $converted = 1;
698
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
699
700
    }
    else {
Sebastien Moretti's avatar
Sebastien Moretti committed
701
        failure();
702
703
        print {*STDERR} "\tThe file \'$msa\' doesn't not seem to be reachable\n\n";
        exit(1);
Sebastien Moretti's avatar
Sebastien Moretti committed
704
705
    }

Sebastien Moretti's avatar
Sebastien Moretti committed
706
    return($msa, $converted);
Sebastien Moretti's avatar
Sebastien Moretti committed
707
708
709
710
}

#Prepare Fasta Header of the query name to incorporate our annotations
sub fastaHeaders4ProtoGene{
711
    my ($QueryName) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
712

713
    $QueryName =~ s{  +}{ }g;
714
    my ($acc, $desc) = ($QueryName, $QueryName);
715
716
    $acc  =~ s{^ *([^ ]*) .*$}{$1};
    $acc  =~ s{\s+$}{}g;
717
718
    $acc  =  $acc.'_G_@@';

719
720
721
    $desc =~ s{^ *[^ ]* *(.*)$}{$1};
    $desc =~ s{^[\| \.]+}{};
    $desc =~ s{[\|\. ]+$}{};
722
723

    return($acc.$desc);
Sebastien Moretti's avatar
Sebastien Moretti committed
724
725
726
727
}

#Prepare Fasta Header to put it on the fasta header of the result outputs
sub getFastaHeaderAnnot{
728
    my ($bestTarget) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
729
730

    #Get annotation from nucleotide file header
Sebastien Moretti's avatar
Sebastien Moretti committed
731
732
    open(my $BEST, '<', "$cache/$bestTarget.fas");
    my $annot = <$BEST> || '';
733
    close $BEST;
Sebastien Moretti's avatar
Sebastien Moretti committed
734

Sebastien Moretti's avatar
Sebastien Moretti committed
735
    chomp($annot);
736
737
738
739
    $annot =~ s{^ *[^ ]* }{};
    $annot =~ s{[\. ]*$}{};
    $annot =~ s{^ *}{};
    $annot =~ s{  +}{ }g;
Sebastien Moretti's avatar
Sebastien Moretti committed
740
741
742
743
744
745
746
747
748
749
750

    return($annot);
}

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


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

#Direct Reverse translation
sub reverse_trad{
751
    my ($aa) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
752

753
754
755
756
757
758
    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
Sebastien Moretti's avatar
Sebastien Moretti committed
759

760
    my $triplet = 'nnn';
761
    while( my ($x, $y) = each(%reverse_code) ){
762
763
764
765
        if ( uc($aa) eq $x ){
            $triplet = $y;
            last;
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
766
767
768
769
770
771
772
773
774
    }

    return($triplet);
}

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


####################### webblast #######################
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
sub display_blast_param {
    print "
             Program :                   BLASTp
             Evalue threshold :          $blast_param->{'evalue'}
             Matrix :                    $blast_param->{'matrix'}
             Filter :                    $blast_param->{'filter'}
             Equivalent best hits used : $blast_param->{'nbesthits'}

             Database:                   NCBI $blast_param->{'db1'}
             Blast_identity_threshold :  $blast_param->{'identity1'}
             Cover threshold :           $blast_param->{'cover1'}
          if no match
             Database:                   NCBI $blast_param->{'db2'}
             Blast_identity_threshold :   $blast_param->{'identity2'}
             Cover threshold :            $blast_param->{'cover2'}
\n";
    return;
}

sub run_BLAST{
    my ($infile, $whatBlast, $db, $blast_exe, $outfile) = @_;

    my $blast_at = $whatBlast==2  ? '-gigablast=yes'
                 : $whatBlast==1  ? "-blast_dir=$blast_exe"
                 :                  '';

    # BLAST run 1
    system("$webblast_exe -program=blastp -database=$db $blast_at -infile=$infile -matrix=$blast_param->{'matrix'} -evalue=$blast_param->{'evalue'} -method=geneid -filter=$blast_param->{'filter'} -organism=$species -identity=$blast_param->{'identity1'} -cover=$blast_param->{'cover1'} -hits=$blast_param->{'nbesthits'} -quiet=on -outfile=$outfile");


    if ( -z "$outfile" ){
    # BLAST run 2 if no hit with 100% identity and coverage: refseq100 -> nr95  OR  nr100 -> nr95
Sebastien Moretti's avatar
Sebastien Moretti committed
807
        print "run BLAST.. against $blast_param->{'db2'} because there was no acceptable hit against $blast_param->{'db1'}\n"
808
            if ( $whatBlast != 1 && $db ne 'nr' );
Sebastien Moretti's avatar
Sebastien Moretti committed
809
        print "run BLAST.. again, with decreased thresholds, because there was no acceptable hit against $db\n"
810
            if ( $whatBlast == 1 || $db eq 'nr' );
811
812

        system("$webblast_exe -program=blastp -database=$blast_param->{'db2'} $blast_at -infile=$infile -matrix=$blast_param->{'matrix'} -evalue=$blast_param->{'evalue'} -method=geneid -filter=$blast_param->{'filter'} -organism=$species -identity=$blast_param->{'identity2'} -cover=$blast_param->{'cover2'} -hits=$blast_param->{'nbesthits'} -quiet=on -outfile=$outfile");
Sebastien Moretti's avatar
Sebastien Moretti committed
813
814
    }

815
816
817
818
819
820
821
    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
822

823
    return;
Sebastien Moretti's avatar
Sebastien Moretti committed
824
825
826
827
828
829
830
}


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


##################### NCBI requests #####################
831
832
833
sub fetch {
    my ($url) = @_;

834
    XML:
835
836
    for (my $tries=0; $tries <20; $tries++ ){
        my $content = get($url);
837
        print {*STDERR} "[$content]\n\n" if ($debug);
838
        if ( defined $content ){
839
            next XML  if ( $content =~ /<ERROR>/i && $content !~ /<ERROR>Can not find description/i );
840
841
            return $content;
        }
842
    }
Sebastien Moretti's avatar
Sebastien Moretti committed
843
    print {*STDERR} "\n\tERROR: Problem with NCBI eutils, please try again later\n";
844
    print {*STDOUT} "\n\tERROR: Problem with NCBI eutils, please try again later\n";
Sebastien Moretti's avatar
Sebastien Moretti committed
845
    exit(4);
846
}
847

848
849
850
sub fetch_fasta {
    my ($url, $outfile) = @_;

851
    return 1 if ( -s "$outfile" && $cache eq 'update' && -M "$outfile" <= $cacheStorageTime );
852

853
854
855
856
857
858
859
860
861
862
863
    GET_FASTA:
    for (my $tries=0; $tries <20; $tries++ ){
        my $HTTP_status = getstore($url, $outfile);
        if ( is_error($HTTP_status) ){
            next GET_FASTA;
        }
        else {
            open(my $FASTA, '<', "$outfile");
            my $counter = 0;
            my $lines   = 0;
            while(<$FASTA>){
864
#                $lines++              if ( $_ !~ /^>/ ); #Allow download of entries with only fasta header and no sequence associated
865
866
867
868
869
                $counter = $counter+2 if ( $counter==1 && $_ !~ /^\w/ && $lines==1 && $_ !~ /^>/ );
                $counter++            if ( $_ =~ /^>/ );
                $counter = $counter+2 if ( $_ !~ /^>/ && ($_ =~ /Error:/ || $_ =~ /[<>]/) );
            }
            close $FASTA;
870

871
872
873
            return 1 if ( $counter == 1 );
        }
    }
874
    print {*STDERR} "[[$url]]\n" if ( $debug );
875
    unlink("$outfile");
Sebastien Moretti's avatar
Sebastien Moretti committed
876
    print {*STDERR} "\n\tERROR: Problem with NCBI eutils, please try again later.\n";
877
    print {*STDOUT} "\n\tERROR: Problem with NCBI eutils, please try again later.\n";
Sebastien Moretti's avatar
Sebastien Moretti committed
878
    exit(5);
879
880
}

Sebastien Moretti's avatar
Sebastien Moretti committed
881
882
883

#Prot ACC -> PUID
sub blastPAcc2PGI{
884
    my ($blastHit) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
885

886
    my $protGI = '';
887
    #FIXME: should be ${blastHit}[pacc] but something is broken at NCBI
888
    my $content = fetch("http://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?db=protein&term=$blastHit&retmode=xml&tool=ProtoGene&email=smoretti\@unil.ch");
889
890
    if ( $content =~ /<Id>(\d+)<\/Id>/ ){
        $protGI = $1;
Sebastien Moretti's avatar
Sebastien Moretti committed
891
892
893
894
895
896
897
    }

    return($protGI);
}

#PUID -> nt GIs
sub protGI2NTGIs{
898
    my ($protGI, $blastHit) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
899

900
901
    my $ntGIs  = '';
    my $geneID = '';
902
    my $content = fetch("http://eutils.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=protein&db=nuccore,gene&id=$protGI&retmode=xml&tool=ProtoGene&email=smoretti\@unil.ch");
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919

    my @xml  = split("\n", $content);
    my $flag = 0;
    for my $line (@xml){
        if (    $line =~ /<LinkName>protein_nuc[a-z]+<\/LinkName>/ ){ #for nuccleotide and nuccore
            $flag = 1;
        }
        elsif ( $line =~ /<LinkName>protein_gene<\/LinkName>/ ){
            $flag  = 2;
        }
        elsif ( $flag==1 && $line =~ /^.*<Id>(\d+)<\/Id>.*$/ ){
            my $match = $1;
            $ntGIs  .= "$match,$match,";
        }
        elsif ( $flag==2 && $line =~ /^.*<Id>(\d+)<\/Id>.*$/ ){
            my $match = $1;
            $geneID .= "$match,$match,";
920
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
921
922
    }

923
    #Remove redundancy if any
924
    chop $ntGIs; #Remove last ','
925
926
927
928
929
930
    my %hash_NT = split(',', $ntGIs);
    $ntGIs      = join(',',  keys(%hash_NT) );
    chop $geneID;
    my %hash_GE = split(',', $geneID);
    $geneID     = join(',',  keys(%hash_GE) );

931
    return($ntGIs, $geneID);
Sebastien Moretti's avatar
Sebastien Moretti committed
932
933
934
}

sub geneID2Chr{
935
    my ($geneID, $blastHit) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
936

937
938
    my $chr   = '';
    my ($amont, $aval) = ('', '');
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
#   my $content = fetch("http://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=gene&id=$geneID&retmode=xml&tool=ProtoGene&email=smoretti\@unil.ch");
    my $content = fetch("http://eutils.ncbi.nlm.nih.gov/entrez/eutils/esummary.fcgi?db=gene&id=$geneID&retmode=xml&tool=ProtoGene&email=smoretti\@unil.ch");

    my @xml  = split("\n", $content);
    my $flag = 0;
    GENE_INFO:
    for my $line (@xml){
        if ( $flag==0    && $line =~ /<Item Name="GenomicInfoType" Type="Structure">/ ){
            $flag = 1;
        }
        elsif ( $flag==1 && $line =~ /<Item Name="ChrAccVer" Type="String">([^<]+?)\.?\d*<\/Item>/ ){
            $chr   = $1;
            $flag  = 3;
        }
        elsif ( $flag==3 && $line =~ /<Item Name="ChrStart" Type="Integer">(\d+)<\/Item>/ ){
            $amont = $1;
            $flag  = 4;
        }
        elsif ( $flag==4 && $line =~ /<Item Name="ChrStop" Type="Integer">(\d+)<\/Item>/ ){
            $aval  = $1;
            last GENE_INFO;
960
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
961
    }
962

963
964
    if ( $amont eq '' || $aval eq '' ){
        $amont = '';
965
        $aval  = '';
Sebastien Moretti's avatar
Sebastien Moretti committed
966
    }
967
968
    elsif ( $amont ne '' && $aval ne '' ){
        if ( $amont > $aval ){
969
970
            $amont = $amont + 5000;
            $aval  = $aval  - 5000;
971
972
        }
        else{
973
974
            $amont = $amont - 5000;
            $aval  = $aval  + 5000;
975
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
976
    }
977
    $amont = 1 if ( $amont ne '' && $amont <= 0 );
978
    $aval  = 1 if ( $aval  ne '' && $aval  <= 0 );
Sebastien Moretti's avatar
Sebastien Moretti committed
979

980
    return($chr, $amont, $aval);
Sebastien Moretti's avatar
Sebastien Moretti committed
981
982
983
}

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

986
    GET_SEQ:
Sebastien Moretti's avatar
Sebastien Moretti committed
987
    for(my $a=0; $a<=$#acc; $a++){
988
        if ( $amont =~ /^\d+$/ && $aval =~ /^\d+$/ ){
989
            fetch_fasta("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", "$cache/$acc[$a]-$amont.fas") ;
990
991
        }
        else{
992
            fetch_fasta("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'", "$cache/$acc[$a].fas");
993
        }