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;

Sebastien Moretti's avatar
Sebastien Moretti committed
15
use File::Which qw(which);                 # Locate external executable programs in the PATH
16
17
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
18
use File::Copy qw(move);                   # Avoid external 'mv' command usage
19
use LWP::Simple;                           # To test gigablaster availability
Sebastien Moretti's avatar
Sebastien Moretti committed
20
21
22

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

23
use lib '/mnt/common/share/ProtoGene/';    # Local path for ProtoGene's own perl modules
24
#use lib '/mnt/local/lib/tcoffee_perl/';
Sebastien Moretti's avatar
Sebastien Moretti committed
25
use Exonerate;                             # Exonerate runner, parser, ...
26
use Views;                                 # Non-text outputs, e.g. HTML/CSS
27
#use CheckOutput;                         # Check output for cds consistancy with query
Sebastien Moretti's avatar
Sebastien Moretti committed
28
29
30



31
################## CONFIGURATION ##################
32
#$ENV{'PATH'}        .= ':/mnt/local/bin/:./';                               # Additional path for executables
33

34
my $cachePath        = '/scratch/fhgfs/tcoffee/ProtoGene_Cache';            # Cache directory
35
36
37
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
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54

### 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,
                  };
55
###################################################
Sebastien Moretti's avatar
Sebastien Moretti committed
56
57


Sebastien Moretti's avatar
Sebastien Moretti committed
58
my $VERSION  = '4.2.0';
Sebastien Moretti's avatar
Sebastien Moretti committed
59

60
my $webblast_exe  = '/mnt/common/share/ProtoGene/webblast.pl';
61
62
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
63
64
65
66
67


################## Option management
my ($msa, $revtrans, $pep, $hideBOJ, $run_name, $template, $lim, $cache) = ('', 0, 0, 0, '', '', 0, 'update');
my ($debug, $tmp)                                                        = (0, 0);
68
my ($db, $species, $local, $giga)                                        = ($blast_param->{'db1'}, $blast_param->{'species'}, 0, 0);
Sebastien Moretti's avatar
Sebastien Moretti committed
69
70
71
72
73
74
75
76
77
78
79
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
80
            'local'           => \$local,      # Use to specify a local db query, defined in $local_db
Sebastien Moretti's avatar
Sebastien Moretti committed
81
82
83
84
85
86
87
88
89
90
            '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
91
           );
Sebastien Moretti's avatar
Sebastien Moretti committed
92
93
my $test_option_values = Getopt::Long::GetOptions(%opts);

94
95

################## Short help message
96
if ( !$test_option_values || ($msa eq '' && $cache ne 'empty' && $cache ne 'old') ){
97
98
99
100
101
102
103
104
    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, ...
105
\t                default is '$blast_param->{'species'}'
106
\t         --db=nr, pdb, swissprot, refseq_protein
107
\t                default is '$blast_param->{'db1'}'
108
109
110
111
\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
112
\t         --revtrans reverse-translates sequences with no
113
114
\t                blast hit, in IUB (IUPAC) depiction code
\t                They are removed from the alignement by default
115
\t         --pep  adds the original peptide query beneath the
116
117
118
119
\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
120
121
122
\t         --debug   prints extra information when running
\t         --version prints version information
\t         --help    prints a full help message\n\n";
123
    exit(1);
Sebastien Moretti's avatar
Sebastien Moretti committed
124
}
125
126


127
################## Check cache directory accessibility
Sebastien Moretti's avatar
Sebastien Moretti committed
128
129
130
if ( -d $cache ){
    $cachePath = $cache;
}
131
checkCacheAccessibility($cachePath);
132

133
134
# Cache management
$cache = cacheManagement($cache, $cachePath, 0) || '.';
135

136
137
# Remove old files from the cache directory everytime ProtoGene is running
cacheManagement('old', $cachePath, 1);
138
139


140
141

################## Check external software in the PATH
142
checkExternalSoftware($webblast_exe, $exonerate_exe);
143
144
if ( $local ){
    checkExternalSoftware($blast_exe);
Sebastien Moretti's avatar
Sebastien Moretti committed
145
146
147
148
}



149
150
151
################## Check input files
# Open and check template file
my ($template_NT, $template_AA, $template_SEQ);
152
#TODO: template is fasta only now
153
154
($template_NT, $template_AA, $template_SEQ) = checkTemplate($template) if ( $template );

Sebastien Moretti's avatar
Sebastien Moretti committed
155

Sebastien Moretti's avatar
Sebastien Moretti committed
156
157
158
159
160
# 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
161

162
my $fasta_checker = -1;
163
my (@original_names, @original_seq); #Use 2 lists in parallel : original_names, seq
Sebastien Moretti's avatar
Sebastien Moretti committed
164
165
open( my $MSA, '<', "$msa" ) or die "\n\tCannot open your MSA file\n\n";
READ_MSA:
166
while(<$MSA>){
Sebastien Moretti's avatar
Sebastien Moretti committed
167
168
    next READ_MSA if ( $_ =~ /^#/ );

169
    if ( $_ =~ /^>/ ){
170
171
        $fasta_checker++;
        my $name = $_;
Sebastien Moretti's avatar
Sebastien Moretti committed
172
173
174
        $name    =~ s{\r\n}{}g; #Remove return lines from windows OS '^M'
        $name    =~ s{\r}{\n}g;
        $name    =~ s{^>[ \t]+}{>};
175
        chomp($name);
Sebastien Moretti's avatar
Sebastien Moretti committed
176
177

        push @original_names, $name;
Sebastien Moretti's avatar
Sebastien Moretti committed
178
    }
Sebastien Moretti's avatar
Sebastien Moretti committed
179
180
181
182
183
184
185
186
187
188
    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
189
        #fasta sequence on 1 line
190
        if ( !exists($original_seq[$fasta_checker]) ){
Sebastien Moretti's avatar
Sebastien Moretti committed
191
            push @original_seq, $sequence;
192
193
        }
        else {
Sebastien Moretti's avatar
Sebastien Moretti committed
194
            $original_seq[$fasta_checker] .= $sequence;
195
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
196
197
    }
}
198
close $MSA;
Sebastien Moretti's avatar
Sebastien Moretti committed
199
200

#Are there FASTA sequences ?
201
if ( $fasta_checker == -1 ){
Sebastien Moretti's avatar
Sebastien Moretti committed
202
    failure();
Sebastien Moretti's avatar
Sebastien Moretti committed
203
204
    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";
205
    exit(1);
Sebastien Moretti's avatar
Sebastien Moretti committed
206
}
207
elsif ( $lim >0 && $fasta_checker > $lim ){
Sebastien Moretti's avatar
Sebastien Moretti committed
208
    failure();
209
    print {*STDERR} "\tThe FASTA file is too large, try with less than $lim sequences\n\tor split your file\n\n";
210
211
    exit(1);
}
212
elsif ( exists( $original_seq[0] ) && $original_seq[0] =~ /[acgtu]/i ){
213
    my $first_seq = $original_seq[0];
Sebastien Moretti's avatar
Sebastien Moretti committed
214
215
216
217
218
219
    #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;
220
    if ( ($a+$c+$g+$t) >= (($a+$c+$g+$t+$non)*80/100) ){
Sebastien Moretti's avatar
Sebastien Moretti committed
221
        failure();
Sebastien Moretti's avatar
Sebastien Moretti committed
222
223
        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";
224
225
        exit(1);
    }
Sebastien Moretti's avatar
Sebastien Moretti committed
226
227
228
229
}
undef $fasta_checker;


Sebastien Moretti's avatar
Sebastien Moretti committed
230
# Check GigaBlaster status if used
231
if ( $giga==1 ){
232
    $local = 0;
Sebastien Moretti's avatar
Sebastien Moretti committed
233
    $local = 2  if ( head('http://www.igs.cnrs-mrs.fr/public_g/remoteblast.cgi') );
234
235
}

Sebastien Moretti's avatar
Sebastien Moretti committed
236

Sebastien Moretti's avatar
Sebastien Moretti committed
237
# Temporary file extension
238
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
239
240


241

Sebastien Moretti's avatar
Sebastien Moretti committed
242
#Start main program with version # of programs and list original queries
243
print "\n\t   Protogene\t$VERSION\t$date\n\n";
244

245
open(my $EXONERATEISHERE, "$exonerate_exe --version |");
246
my $IsExonerateHere = <$EXONERATEISHERE>;
247
close $EXONERATEISHERE;
Sebastien Moretti's avatar
Sebastien Moretti committed
248
249
print "\t   $IsExonerateHere\n";
undef $IsExonerateHere;
250
undef $VERSION;
251

Sebastien Moretti's avatar
Sebastien Moretti committed
252
for(my $m=0; $m<=$#original_names; $m++){
253
    printf("\t>%-3s  is  %s\n", $m, $original_names[$m]);
Sebastien Moretti's avatar
Sebastien Moretti committed
254
}
Sebastien Moretti's avatar
Sebastien Moretti committed
255
$date .= "_$$"; #Add PID to be more uniq
Sebastien Moretti's avatar
Sebastien Moretti committed
256
257
258
259



# run_name implementation for web server
260
$originalMSA = $run_name if ( $run_name ne '' );
Sebastien Moretti's avatar
Sebastien Moretti committed
261

262
263
unlink("$originalMSA.cds", "$originalMSA.cdsP", "$originalMSA.cdsP.html",
       "$originalMSA.out", "$originalMSA.boj",  "$originalMSA.log");
Sebastien Moretti's avatar
Sebastien Moretti committed
264
265


266
# Build the sequences, from the alignment, to perform blast search
267
EACH_SEQ:
268
269
270
271
272
273
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
274

Sebastien Moretti's avatar
Sebastien Moretti committed
275
    my @equivalent_blast_hits; #Multiple BLAST hits with the same highest e-value/score
276
277

    if ( exists($template_NT->{$short_name}) && $template_NT->{$short_name} ne '' ){
Sebastien Moretti's avatar
Sebastien Moretti committed
278
279
280
        # First try nucleotide template
        push @equivalent_blast_hits, 'My_own_seq';
        print "\n\n$r done\n\nNucleotide template for $r:\t$template_NT->{$short_name}\n\n*******************************************************************\n\n";
Sebastien Moretti's avatar
Sebastien Moretti committed
281
    }
282
    elsif ( exists($template_AA->{$short_name}) && $template_AA->{$short_name} ne '' ){
Sebastien Moretti's avatar
Sebastien Moretti committed
283
284
285
        # 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
286
287
    }
    else{
288
    # else send them to webblast
Sebastien Moretti's avatar
Sebastien Moretti committed
289
290
291
292
293
294
295
        # 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;

296
297
298
299
300
        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
301

302
303
304

        # Get the BLASTp hit(s)  acc number
        open(my $BLASTP, '<', "$cache/$date.blastp.$r");
305
        while(<$BLASTP>){
306
            push @equivalent_blast_hits, $1  if ( $_ =~ /^>\s*\d+@?\w?\w?\w?__([^\s\|\.]+).*$/ ); #Warning: double '_'
307
308
        }
        close $BLASTP;
309
        unlink("$cache/$date.blastp.$r", "$cache/$date.seq2blast.fas.$r") if ( $tmp == 0 || exists($equivalent_blast_hits[0]) );
Sebastien Moretti's avatar
Sebastien Moretti committed
310
        if ( !exists($equivalent_blast_hits[0]) ){
311
            print "\n$r ...\tNo blast result found above thresholds for $fasta_header\n\n";
312
            buildFailureOutputFiles($r, 'No_BLASTp_Result', 'Not_searched');
313
314
315
316
            undef $original_seq[$r];
            undef $original_names[$r];
            next EACH_SEQ;
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
317
318
319
320
    }



321
    #When multiple equivalent blast hits  ==> use, and add, every hits
322
    my ($resultPOS, $resultBOJ) = ('', '');
323
    my @failureStatus = (''); # To always have $failureStatus[0] defined
324
    my @intronStatus  = ('');
325
    HIT_LINK:
Sebastien Moretti's avatar
Sebastien Moretti committed
326
327
    for(my $qq=0; $qq<=$#equivalent_blast_hits; $qq++){
        print "\n$fasta_header  -->  [$equivalent_blast_hits[$qq]]\n";
328
329
330
331


        my @nt_GIs;
        my $intronStep = 0;
332
333
334
        # Get Nt sequence from template
        if ( exists($template_NT->{$short_name}) && $template_NT->{$short_name} ne '' ){
            if ( $template_NT->{$short_name} !~ /^My_Seq$/i ){
335
                download_seq($cache, '', '', $template_NT->{$short_name});
336
337
338
339
340
341
342
343
                @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-$$";
            }
344
        }
345
        # Get Nt sequence through prot acc -> prot gi -> nt linked gi -> nt acc
346
        else{
Sebastien Moretti's avatar
Sebastien Moretti committed
347
            # Prot ACC -> PUID
Sebastien Moretti's avatar
Sebastien Moretti committed
348
            my $protGI = blastPAcc2PGI($equivalent_blast_hits[$qq]);
349
            if ( $protGI eq '' ){
Sebastien Moretti's avatar
Sebastien Moretti committed
350
351
                print "\tNo protein (prot GI) link found for $equivalent_blast_hits[$qq] in $fasta_header\n\n";
                @failureStatus = ('PUI_unavailable', $equivalent_blast_hits[$qq]);
352
                next HIT_LINK;
353
            }
Sebastien Moretti's avatar
Sebastien Moretti committed
354

Sebastien Moretti's avatar
Sebastien Moretti committed
355
            # PUID -> nt GIs & GeneID
Sebastien Moretti's avatar
Sebastien Moretti committed
356
            my ($ntGIs, $geneID) = protGI2NTGIs($protGI, $equivalent_blast_hits[$qq]);
357
            if ( $ntGIs eq '' && $geneID eq '' ){
Sebastien Moretti's avatar
Sebastien Moretti committed
358
359
                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]);
360
                next HIT_LINK;
361
362
363
364
            }
            print " linked with [$ntGIs $geneID]\n";


Sebastien Moretti's avatar
Sebastien Moretti committed
365
            # Get transcript and contig seq
Sebastien Moretti's avatar
Sebastien Moretti committed
366
            if ( $ntGIs ne '' ){
Sebastien Moretti's avatar
Sebastien Moretti committed
367
                @nt_GIs = split(/,/, $ntGIs);
Sebastien Moretti's avatar
Sebastien Moretti committed
368
369
                downloadSeqFromGIs($cache, '', '', @nt_GIs);
            }
370

Sebastien Moretti's avatar
Sebastien Moretti committed
371
            # GeneID -> Chr
372
373
374
375
376
377
378
            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
379
                    # Get Chr seq
380
381
                    if ( $chr ne '' ){
                        download_seq($cache, $amont, $aval, $chr);
Sebastien Moretti's avatar
Sebastien Moretti committed
382
                        $intronStep = 1;
383

384
                        unshift(@nt_GIs, "$chr--$amont-$aval");
385
                    }
386
387
388
389
390
                }
            }
        }


Sebastien Moretti's avatar
Sebastien Moretti committed
391
392
        # Align Nt seq with our protein query seq
        my ($POS, $BOJ) = runExonerate($cache, $date, $r, $original_seq[$r], $fasta_header, @nt_GIs);
393
        next HIT_LINK if ( !defined $POS );
394

Sebastien Moretti's avatar
Sebastien Moretti committed
395
        my ($POSresult, $bestNucleotide) = prepareResults4CDS($POS, $equivalent_blast_hits[$qq], $original_seq[$r], $fasta_header);
396
397
        $resultPOS .= $POSresult if ( $bestNucleotide ne '' && $POSresult ne '' && $resultPOS !~ /_G_$bestNucleotide/ );

398
399
        @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]_/ );
400
401
402

        my ($intronLess, $nameLess) = ('', '');
        $intronLess = $POS->{0}       if ( %$BOJ eq 0 && exists($POS->{0}) && $POS->{0} ne '' );
403
        $nameLess   = $bestNucleotide if ( %$BOJ eq 0 && $intronStep==0 && $bestNucleotide =~ /^NA[CTWZGS]_/ );
Sebastien Moretti's avatar
Sebastien Moretti committed
404
        my ($BOJresult, $bestGenomic) = prepareResults4BOJ($BOJ, $equivalent_blast_hits[$qq], $original_seq[$r], $fasta_header, $intronLess, $nameLess);
405
        $resultBOJ .= $BOJresult      if ( $bestGenomic ne '' && $BOJresult ne '' && $resultBOJ !~ /_G_$bestGenomic[ \-]/ );
Sebastien Moretti's avatar
Sebastien Moretti committed
406
407
408
    }


409
    if ( $resultPOS eq '' ){
410
        if ( $failureStatus[0] ne '' ){
411
            buildFailureOutputFiles($r, $failureStatus[1],         $failureStatus[0]);
412
413
        }
        else {
414
            buildFailureOutputFiles($r, $equivalent_blast_hits[0], 'Alignment_failure');
415
        }
416
        next EACH_SEQ;
Sebastien Moretti's avatar
Sebastien Moretti committed
417
    }
418
419
    elsif ( $resultBOJ eq '' ){
        if ( $failureStatus[0] eq 'PUI_unavailable' || $failureStatus[0] eq 'No_nt_link' ){
420
            buildFailureBOJOutputFile($failureStatus[1], $failureStatus[0], $fasta_header, $original_seq[$r]);
421
422
        }
        elsif ( $failureStatus[0] eq '' && $intronStatus[0] ne '' ){
423
            buildIntronlessBOJOutputFile($intronStatus[1], $intronStatus[0], $intronStatus[2], $fasta_header, $original_seq[$r]);
424
425
        }
        else{
Sebastien Moretti's avatar
Sebastien Moretti committed
426
            buildFailureBOJOutputFile($equivalent_blast_hits[0], 'Unavailable', $fasta_header, $original_seq[$r]);
427
        }
428
        createCDSOutputFile($resultPOS, $original_seq[$r], $fasta_header);
Sebastien Moretti's avatar
Sebastien Moretti committed
429
430
    }
    else {
431
        createCDSOutputFile($resultPOS, $original_seq[$r], $fasta_header);
432
        createBOJOutputFile($resultBOJ);
Sebastien Moretti's avatar
Sebastien Moretti committed
433
434
435
436
437
438
439
440
    }


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


Sebastien Moretti's avatar
Sebastien Moretti committed
441
checkAndCleanStderrFiles("$cache/$date.ExonerateError") if ( -e "$cache/$date.ExonerateError" );
Sebastien Moretti's avatar
Sebastien Moretti committed
442
checkOutputFiles();
Sebastien Moretti's avatar
Sebastien Moretti committed
443
unlink("$msa") if ( $converted==1 );
444
unlink 'blast_result.txt';
Sebastien Moretti's avatar
Sebastien Moretti committed
445
my @tmpFiles = glob($cache."/$date*");
Sebastien Moretti's avatar
Sebastien Moretti committed
446
if ( exists($tmpFiles[0]) ){
Sebastien Moretti's avatar
Sebastien Moretti committed
447
448
    mkdir("$cache/$date.TMP");
    move("$_", "$cache/$date.TMP") while(<@tmpFiles>);
Sebastien Moretti's avatar
Sebastien Moretti committed
449
450
451
452
453
454
455
}


#Ouput Checker


#STDOUT Log output for our server
Sebastien Moretti's avatar
Sebastien Moretti committed
456
if ( -s "$originalMSA.cds" || -s "$originalMSA.out" ){
Sebastien Moretti's avatar
Sebastien Moretti committed
457
    print "\n\nTERMINATION STATUS: SUCCESS\n";
458
    print 'OUTPUT RESULTS', "\n";
459
    print "    #### File Type=        MSA Format= fasta_CDS Name= $originalMSA.cds\n"
460
        if ( -s "$originalMSA.cds" );
461
    print "    #### File Type=        MSA Format= fasta_CDS+Query Name= $originalMSA.cdsP\n"
462
        if ( -s "$originalMSA.cdsP" );
463
    print "    #### File Type=        MSA Format= colored_fasta_CDS Name= $originalMSA.cdsP.html\n"
464
        if ( Views::Html("$originalMSA.cdsP") && -s "$originalMSA.cdsP.html" );
465
    print "    #### File Type=        MSA Format= Rejected_seq Name= $originalMSA.out\n"
466
        if ( -s "$originalMSA.out" );
467
    print "    #### File Type=        MSA Format= fasta_Exon-Boundaries Name= $originalMSA.boj\n"
468
        if ( -s "$originalMSA.boj" && -s "$originalMSA.cds" && $hideBOJ==0 );
Sebastien Moretti's avatar
Sebastien Moretti committed
469
470

    unlink "$cache/web_tempo.result";
Sebastien Moretti's avatar
Sebastien Moretti committed
471
472
}
else{
Sebastien Moretti's avatar
Sebastien Moretti committed
473
    failure();
474
    exit(1);
Sebastien Moretti's avatar
Sebastien Moretti committed
475
476
}

477
exit(0);
Sebastien Moretti's avatar
Sebastien Moretti committed
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
515
516
517
518
519
520
521


=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

522
=item empty: empty completely the cache directory
Sebastien Moretti's avatar
Sebastien Moretti committed
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543

=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 !

544
=item Standard Perl modules B<lib>, B<strict>,  B<warnings>,  B<diagnostics>, B<Carp> are required
Sebastien Moretti's avatar
Sebastien Moretti committed
545

546
=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
547
548
549
550
551
552
553
554
555
556
557
558
559

=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

Sebastien Moretti's avatar
Sebastien Moretti committed
560
=item version 4.2.0
Sebastien Moretti's avatar
Sebastien Moretti committed
561

Sebastien Moretti's avatar
Sebastien Moretti committed
562
=item on Aug 08th, 2013
Sebastien Moretti's avatar
Sebastien Moretti committed
563
564
565
566
567
568
569
570
571

=back

=head1 AUTHORS

=over 8

=item Sebastien MORETTI

Sebastien Moretti's avatar
Sebastien Moretti committed
572
=item moretti.sebastien [AT] gmail.com
Sebastien Moretti's avatar
Sebastien Moretti committed
573

574
=item Vital-IT computing center
Sebastien Moretti's avatar
Sebastien Moretti committed
575

576
=item Swiss Institute of Bioinformatics
Sebastien Moretti's avatar
Sebastien Moretti committed
577

578
=item Lausanne, Switzerland
Sebastien Moretti's avatar
Sebastien Moretti committed
579

580
=item http://www.vital-it.ch/
Sebastien Moretti's avatar
Sebastien Moretti committed
581
582
583
584
585
586
587
588

=back

=cut


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

589
# Failure declaration
Sebastien Moretti's avatar
Sebastien Moretti committed
590
sub failure {
591
592
593
    print {*STDERR} "\n\nFATAL\n\nTERMINATION STATUS: FAILURE\n\n";
    return;
}
Sebastien Moretti's avatar
Sebastien Moretti committed
594

Sebastien Moretti's avatar
Sebastien Moretti committed
595
# Check external programs presence in the PATH
596
sub checkExternalSoftware {
Sebastien Moretti's avatar
Sebastien Moretti committed
597
    for my $exe ( @_ ){
598
        if ( ! which($exe) && !-x $exe ){
Sebastien Moretti's avatar
Sebastien Moretti committed
599
600
601
602
            failure();
            print {*STDERR} "\t'$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
603
    }
Sebastien Moretti's avatar
Sebastien Moretti committed
604
    return;
Sebastien Moretti's avatar
Sebastien Moretti committed
605
606
}

607
608
# Check cache directory accessibility: Test and change rights, if required, of the cache directory
sub checkCacheAccessibility {
609
    my ($cacheDir) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
610

611
    if ( -d "$cacheDir/" && -w "$cacheDir/" && -x "$cacheDir/" && -r "$cacheDir/" ){
612
        # Good
Sebastien Moretti's avatar
Sebastien Moretti committed
613
    }
614
615
616
    elsif ( -d "$cacheDir/" ){
        if ( -o "$cacheDir/" ){
            chmod 0754, "$cacheDir/";
617
        }
618
        else{
619
620
621
            failure();
            print {*STDERR} "\tPermissions do NOT seem to be valid for the '$cacheDir' cache directory\n\n";
            exit(2);
622
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
623
624
    }
    else{
625
        mkdir "$cacheDir/"; # FIXME: mkdir is not recursive !
626
        if ( -o "$cacheDir/" ){
627
            chmod 0754, "$cacheDir/"; #0754 is in octal, so 754 or '0754' are not good !
628
629
        }
        else{
630
631
632
            failure();
            print {*STDERR} "\tPermissions do NOT seem to be valid for the '$cacheDir' cache directory\n\n";
            exit(2);
633
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
634
    }
635

636
    return;
Sebastien Moretti's avatar
Sebastien Moretti committed
637
638
}

639
# Cache Behavior Management
Sebastien Moretti's avatar
Sebastien Moretti committed
640
sub cacheManagement{
641
    my ($cache, $cacheDir, $inScript) = @_;
642

643
    if ( $cache eq 'empty' ){
644
        unlink <$cacheDir/*.fas>;
Sebastien Moretti's avatar
Sebastien Moretti committed
645
        unlink <$cacheDir/*.ExonerateError>;
646
647
        unlink("$cacheDir/webblast.log", "$cacheDir/web_tempo.result", "$cacheDir/debug.tempo");
        print "\n\tcache directory [$cacheDir] is empty\n\n";
648
        exit 0;
Sebastien Moretti's avatar
Sebastien Moretti committed
649
    }
650
    elsif ( $cache eq 'old' ){
651
652
653
654
655
        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;
656
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
657
    }
658
    elsif ( $cache eq 'update' || $cache eq 'use' ){
659
660
        # DEFAULT: Limit cached files to less than $cacheStorageTime days old ones for 'update'
        $cache = $cacheDir;
Sebastien Moretti's avatar
Sebastien Moretti committed
661
    }
662
    elsif ( $cache eq 'none' ){
663
        # Do not use cache
Sebastien Moretti's avatar
Sebastien Moretti committed
664
    }
665
666
667
668
    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 $cache );
Sebastien Moretti's avatar
Sebastien Moretti committed
669
670
    }
    else {
671
672
673
        failure();
        print {*STDERR} "\n\tWrong cache argument, or your cache folder '$cacheDir' doesn't seem to exist\n\n";
        exit(3);
Sebastien Moretti's avatar
Sebastien Moretti committed
674
675
676
677
678
679
680
681
682
683
    }

    return($cache);
}

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


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

Sebastien Moretti's avatar
Sebastien Moretti committed
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
# 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';
            $converted = 1;
699
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
700
701
    }
    else {
Sebastien Moretti's avatar
Sebastien Moretti committed
702
        failure();
703
704
        print {*STDERR} "\tThe file \'$msa\' doesn't not seem to be reachable\n\n";
        exit(1);
Sebastien Moretti's avatar
Sebastien Moretti committed
705
706
    }

Sebastien Moretti's avatar
Sebastien Moretti committed
707
    return($msa, $converted);
Sebastien Moretti's avatar
Sebastien Moretti committed
708
709
710
711
}

#Prepare Fasta Header of the query name to incorporate our annotations
sub fastaHeaders4ProtoGene{
712
    my ($QueryName) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
713

714
    $QueryName =~ s{  +}{ }g;
715
    my ($acc, $desc) = ($QueryName, $QueryName);
716
717
    $acc  =~ s{^ *([^ ]*) .*$}{$1};
    $acc  =~ s{\s+$}{}g;
718
719
    $acc  =  $acc.'_G_@@';

720
721
722
    $desc =~ s{^ *[^ ]* *(.*)$}{$1};
    $desc =~ s{^[\| \.]+}{};
    $desc =~ s{[\|\. ]+$}{};
723
724

    return($acc.$desc);
Sebastien Moretti's avatar
Sebastien Moretti committed
725
726
727
728
}

#Prepare Fasta Header to put it on the fasta header of the result outputs
sub getFastaHeaderAnnot{
729
    my ($bestTarget) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
730
731

    #Get annotation from nucleotide file header
Sebastien Moretti's avatar
Sebastien Moretti committed
732
733
    open(my $BEST, '<', "$cache/$bestTarget.fas");
    my $annot = <$BEST> || '';
734
    close $BEST;
Sebastien Moretti's avatar
Sebastien Moretti committed
735

Sebastien Moretti's avatar
Sebastien Moretti committed
736
    chomp($annot);
737
738
739
740
    $annot =~ s{^ *[^ ]* }{};
    $annot =~ s{[\. ]*$}{};
    $annot =~ s{^ *}{};
    $annot =~ s{  +}{ }g;
Sebastien Moretti's avatar
Sebastien Moretti committed
741
742
743
744
745
746
747
748
749
750
751

    return($annot);
}

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


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

#Direct Reverse translation
sub reverse_trad{
752
    my ($aa) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
753

754
755
756
757
758
759
    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
760

761
    my $triplet = 'nnn';
762
    while( my ($x, $y) = each(%reverse_code) ){
763
764
765
766
        if ( uc($aa) eq $x ){
            $triplet = $y;
            last;
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
767
768
769
770
771
772
773
774
775
    }

    return($triplet);
}

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


####################### webblast #######################
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
807
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
808
        print "run BLAST.. against $blast_param->{'db2'} because there was no acceptable hit against $blast_param->{'db1'}\n"
809
            if ( $whatBlast != 1 && $db ne 'nr' );
Sebastien Moretti's avatar
Sebastien Moretti committed
810
        print "run BLAST.. again, with decreased thresholds, because there was no acceptable hit against $db\n"
811
            if ( $whatBlast == 1 || $db eq 'nr' );
812
813

        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
814
815
    }

816
817
818
819
820
821
822
    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
823

824
    return;
Sebastien Moretti's avatar
Sebastien Moretti committed
825
826
827
828
829
830
831
}


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


##################### NCBI requests #####################
832
833
834
sub fetch {
    my ($url) = @_;

835
    XML:
836
837
    for (my $tries=0; $tries <20; $tries++ ){
        my $content = get($url);
838
        print {*STDERR} "[$content]\n\n" if ($debug);
839
840
841
842
        if ( defined $content ){
            next XML if ( $content =~ /<ERROR>/i );
            return $content;
        }
843
    }
Sebastien Moretti's avatar
Sebastien Moretti committed
844
    print {*STDERR} "\n\tERROR: Problem with NCBI eutils, please try again later\n";
845
    print {*STDOUT} "\n\tERROR: Problem with NCBI eutils, please try again later\n";
Sebastien Moretti's avatar
Sebastien Moretti committed
846
    exit(4);
847
}
848

849
850
851
sub fetch_fasta {
    my ($url, $outfile) = @_;

852
    return 1 if ( -s "$outfile" && $cache eq 'update' && -M "$outfile" <= $cacheStorageTime );
853

854
855
856
857
858
859
860
861
862
863
864
    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>){
865
#                $lines++              if ( $_ !~ /^>/ ); #Allow download of entries with only fasta header and no sequence associated
866
867
868
869
870
                $counter = $counter+2 if ( $counter==1 && $_ !~ /^\w/ && $lines==1 && $_ !~ /^>/ );
                $counter++            if ( $_ =~ /^>/ );
                $counter = $counter+2 if ( $_ !~ /^>/ && ($_ =~ /Error:/ || $_ =~ /[<>]/) );
            }
            close $FASTA;
871

872
873
874
            return 1 if ( $counter == 1 );
        }
    }
875
    print {*STDERR} "[[$url]]\n" if ( $debug );
876
    unlink("$outfile");
Sebastien Moretti's avatar
Sebastien Moretti committed
877
    print {*STDERR} "\n\tERROR: Problem with NCBI eutils, please try again later.\n";
878
    print {*STDOUT} "\n\tERROR: Problem with NCBI eutils, please try again later.\n";
Sebastien Moretti's avatar
Sebastien Moretti committed
879
    exit(5);
880
881
}

Sebastien Moretti's avatar
Sebastien Moretti committed
882
883
884

#Prot ACC -> PUID
sub blastPAcc2PGI{
885
    my ($blastHit) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
886

887
    my $protGI = '';
888
    #FIXME: should be ${blastHit}[pacc] but something is broken at NCBI
889
    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");
890
891
    if ( $content =~ /<Id>(\d+)<\/Id>/ ){
        $protGI = $1;
Sebastien Moretti's avatar
Sebastien Moretti committed
892
893
894
895
896
897
898
    }

    return($protGI);
}

#PUID -> nt GIs
sub protGI2NTGIs{
899
    my ($protGI, $blastHit) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
900

901
902
    my $ntGIs  = '';
    my $geneID = '';
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
    my $content = fetch("http://eutils.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=protein&db=nuccore,nucleotide,gene&id=$protGI&retmode=xml&tool=ProtoGene&email=smoretti\@unil.ch");

    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,";
921
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
922
923
    }

924
    #Remove redundancy if any
925
    chop $ntGIs; #Remove last ','
926
927
928
929
930
931
    my %hash_NT = split(',', $ntGIs);
    $ntGIs      = join(',',  keys(%hash_NT) );
    chop $geneID;
    my %hash_GE = split(',', $geneID);
    $geneID     = join(',',  keys(%hash_GE) );

932
    return($ntGIs, $geneID);
Sebastien Moretti's avatar
Sebastien Moretti committed
933
934
935
}

sub geneID2Chr{
936
    my ($geneID, $blastHit) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
937

938
939
    my $chr   = '';
    my ($amont, $aval) = ('', '');
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
#   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;
961
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
962
    }
963

964
965
    if ( $amont eq '' || $aval eq '' ){
        $amont = '';
966
        $aval  = '';
Sebastien Moretti's avatar
Sebastien Moretti committed
967
    }
968
969
    elsif ( $amont ne '' && $aval ne '' ){
        if ( $amont > $aval ){
970
971
            $amont = $amont + 5000;
            $aval  = $aval  - 5000;
972
973
        }
        else{
974
975
            $amont = $amont - 5000;
            $aval  = $aval  + 5000;
976
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
977
    }
978
    $amont = 1 if ( $amont ne '' && $amont <= 0 );
979
    $aval  = 1 if ( $aval  ne '' && $aval  <= 0 );
Sebastien Moretti's avatar
Sebastien Moretti committed
980

981
    return($chr, $amont, $aval);
Sebastien Moretti's avatar
Sebastien Moretti committed
982
983
984
}

sub downloadSeqFromGIs{
985
    my ($cache, $amont, $aval, @acc) = @_;
Sebastien Moretti's avatar
Sebastien Moretti committed
986

987
    GET_SEQ:
Sebastien Moretti's avatar
Sebastien Moretti committed
988
    for(my $a=0; $a<=$#acc; $a++){
989
        if ( $amont =~ /^\d+$/ && $aval =~ /^\d+$/ ){
990
            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") ;
991
992
        }
        else{
993
            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");
994
        }
Sebastien Moretti's avatar
Sebastien Moretti committed
995
    }
Sebastien Moretti's avatar