adding bzip2 support in ruby

[biopieces.git] / bp_bin / bowtie_seq

diff --git a/bp_bin/bowtie_seq b/bp_bin/bowtie_seq
index aedf02ad3edce43918f4eb460846e9407060f98e..19368f84452db6ffb3def6777d1ac646e64c9d1e 100755 (executable)
--- a/bp_bin/bowtie_seq
+++ b/bp_bin/bowtie_seq
@@ -21,19 +21,25 @@
 
 # >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> DESCRIPTION <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
 
-# Use bowtie to map sequences in the stream against a specified genome.
+# Use bowtie to map sequences in the stream against a specified genome or index.
 
 # >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>><<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
 
 
 use warnings;
 use strict;
+use Data::Dumper;
 use Maasha::Biopieces;
 use Maasha::Common;
 use Maasha::Fastq;
 use Maasha::Fasta;
 use Maasha::Calc;
 
+use constant {
+    SEQ_NAME => 0,
+    SEQ      => 1,
+    SCORES   => 2,
+};
 
 # >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>><<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
 
@@ -42,18 +48,26 @@ my ( $options, $in, $out, $index, $tmp_dir, $tmp_in, $tmp_out, $fh_in, $fh_out,
 
 $options = Maasha::Biopieces::parse_options(
     [
-        { long => 'genome',      short => 'g', type => 'genome', mandatory => 'yes', default => undef, allowed => undef,     disallowed => undef },
-        { long => 'mismatches',  short => 'm', type => 'uint',   mandatory => 'no',  default => 0,     allowed => "0,1,2,3", disallowed => undef },
-        { long => 'max_hits',    short => 'h', type => 'uint',   mandatory => 'no',  default => undef, allowed => undef,     disallowed => 0     },
-        { long => 'seed_length', short => 's', type => 'uint',   mandatory => 'no',  default => 28,    allowed => undef,     disallowed => 0     },
-        { long => 'cpus',        short => 'c', type => 'uint',   mandatory => 'no',  default => 1,     allowed => undef,     disallowed => 0     },
+        { long => 'genome',      short => 'g', type => 'genome', mandatory => 'no', default => undef, allowed => undef,     disallowed => undef },
+        { long => 'index_name',  short => 'i', type => 'string', mandatory => 'no', default => undef, allowed => undef,     disallowed => undef },
+        { long => 'mismatches',  short => 'm', type => 'uint',   mandatory => 'no', default => 0,     allowed => "0,1,2,3", disallowed => undef },
+        { long => 'max_hits',    short => 'h', type => 'uint',   mandatory => 'no', default => undef, allowed => undef,     disallowed => 0     },
+        { long => 'seed_length', short => 's', type => 'uint',   mandatory => 'no', default => 28,    allowed => undef,     disallowed => 0     },
+        { long => 'cpus',        short => 'c', type => 'uint',   mandatory => 'no', default => 1,     allowed => undef,     disallowed => 0     },
     ]   
 );
 
+Maasha::Common::error( qq(both --index_name and --genome specified) ) if     $options->{ "genome" } and     $options->{ "index_name" };
+Maasha::Common::error( qq(no --index_name or --genome specified) )    if not $options->{ "genome" } and not $options->{ "index_name" };
+
 $in  = Maasha::Biopieces::read_stream( $options->{ "stream_in" } );
 $out = Maasha::Biopieces::write_stream( $options->{ "stream_out" } );
 
-$index = "$ENV{ 'BP_DATA' }/genomes/$options->{ 'genome' }/bowtie/$options->{ 'genome' }";
+if ( defined $options->{ 'genome' } ) {
+    $index = "$ENV{ 'BP_DATA' }/genomes/$options->{ 'genome' }/bowtie/$options->{ 'genome' }";
+} elsif (defined $options->{ 'index_name' } ) {
+    $index = $options->{ 'index_name' };
+}
 
 $tmp_dir = Maasha::Biopieces::get_tmpdir();
 $tmp_in  = "$tmp_dir/bowtie.seq";
@@ -66,6 +80,7 @@ while ( $record = Maasha::Biopieces::get_record( $in ) )
     if ( $entry = Maasha::Fastq::biopiece2fastq( $record ) )
     {
         Maasha::Common::error( "Mixed FASTA and FASTQ entries in stream" ) if defined $type and $type ne "FASTQ";
+        Maasha::Common::error( "Sequence longer than 1024 not allowed")    if length( $entry->[ SEQ ] ) > 1024;
         Maasha::Fastq::put_entry( $entry, $fh_out );
 
         $type = "FASTQ";
@@ -73,6 +88,7 @@ while ( $record = Maasha::Biopieces::get_record( $in ) )
     elsif ( $entry = Maasha::Fasta::biopiece2fasta( $record ) )
     {
         Maasha::Common::error( "Mixed FASTA and FASTQ entries in stream" ) if defined $type and $type ne "FASTA";
+        Maasha::Common::error( "Sequence longer than 1024 not allowed")    if length( $entry->[ SEQ ] ) > 1024;
         Maasha::Fasta::put_entry( $entry, $fh_out );
 
         $type = "FASTA";
@@ -84,7 +100,10 @@ while ( $record = Maasha::Biopieces::get_record( $in ) )
 close $fh_out;
 
 push @args, "-n $options->{ 'mismatches' }";
+push @args, "-v $options->{ 'mismatches' }";  # DANGER: using seed mismatches as alignment mismatches - may work, may not!
 push @args, "-f" if $type eq "FASTA";
+push @args, "-p $options->{ 'cpus' }";
+push @args, "--phred64-quals" unless $type eq "FASTA";
 
 if ( defined $options->{ 'max_hits' } ) {
     push @args, "-k $options->{ 'max_hits' }";
@@ -113,7 +132,6 @@ while ( $line = <$fh_in> )
     chomp $line;
 
     @fields = split /\t/, $line;
-
     $record = bowtie2biopiece( \@fields );
 
     Maasha::Biopieces::put_record( $record, $out );
@@ -141,21 +159,22 @@ sub bowtie2biopiece
 
     # Returns a hash.
 
-    my ( $record, @scores );
-
-    $record->{ 'Q_ID' }     = $entry->[ 0 ];
-    $record->{ 'STRAND' }   = $entry->[ 1 ];
-    $record->{ 'S_ID' }     = $entry->[ 2 ];
-    $record->{ 'S_BEG' }    = $entry->[ 3 ];
-    $record->{ 'SEQ' }      = $entry->[ 4 ];
-    $record->{ 'SCORES' }   = $entry->[ 5 ];
-    $record->{ 'MISMATCH' } = $entry->[ 6 ];
-
+    my ( $record, $s_id, $s_len, $hits );
+
+    $record->{ 'Q_ID' }       = $entry->[ 0 ];
+    $record->{ 'STRAND' }     = $entry->[ 1 ];
+    $record->{ 'S_ID' }       = $entry->[ 2 ];
+    $record->{ 'S_BEG' }      = $entry->[ 3 ];
+    $record->{ 'SEQ' }        = $entry->[ 4 ];
+    $record->{ 'SCORES' }     = $entry->[ 5 ];
+    $record->{ 'SCORE' }      = $entry->[ 6 ] + 1;
+    $record->{ 'ALIGN' }      = $entry->[ 7 ] || '.';
+    $record->{ 'S_LEN' }      = length $entry->[ 4 ];
     $record->{ 'SEQ_LEN' }    = length $entry->[ 4 ];
     $record->{ 'S_END' }      = $record->{ 'S_BEG' } + $record->{ 'SEQ_LEN' } - 1;
-    $record->{ 'SCORES' }     =~ s/(.)/ord( $1 ) - 33 . ";"/ge; # http://maq.sourceforge.net/fastq.shtml
-    $record->{ 'SCORE_MEAN' } = sprintf( "%.2f", Maasha::Calc::mean( [ split /;/, $record->{ 'SCORES' } ] ) );
+    $record->{ 'SCORES' }     =~ s/(.)/chr( ( ord( $1 ) - 33 ) + 64 )/ge; # convert phred scores to illumina scores
 
+    $record->{ 'HITS' }       = '.';
     $record->{ 'REC_TYPE' } = "BOWTIE";
 
     return wantarray ? %{ $record } : $record;