Use Bowtie Index to get sequences using locations

[shawn.mek]

We have the fasta files (obviously) for the hg19 genome, we used them to create a big Bowtie index.

I was hoping not to have to keep the fasta file. Instead just look up sequences in the Bowtie index when I get chromosome locations.

I know when the alignment comes back it tells me where the alignment occurs and which fasta record (header) that it came from. So all the info is there, but I can't figure out how to pull out a sequence given a location.

Does anyone know if this is possible, or know much about the index format (perhaps I could write a little program to fish out a sequence)?


Thanks

[winsettz]

We have the fasta files (obviously) for the hg19 genome, we used them to create a big Bowtie index.

I was hoping not to have to keep the fasta file. Instead just look up sequences in the Bowtie index when I get chromosome locations.

I know when the alignment comes back it tells me where the alignment occurs and which fasta record (header) that it came from. So all the info is there, but I can't figure out how to pull out a sequence given a location.

Does anyone know if this is possible, or know much about the index format (perhaps I could write a little program to fish out a sequence)?


Thanks

You should be able to extract that information from the sam output. I've not used bowtie2-inspect before, but it could be what you are looking for.

bowtie2-inspect
No index name given!
Bowtie 2 version 2.1.0 by Ben Langmead ([email protected], www.cs.jhu.edu/~langmea)
Usage: bowtie2-inspect [options]* <bt2_base>
  <bt2_base>         bt2 filename minus trailing .1.bt2/.2.bt2

  By default, prints FASTA records of the indexed nucleotide sequences to
  standard out.  With -n, just prints names.  With -s, just prints a summary of
  the index parameters and sequences.  With -e, preserves colors if applicable.

Options:
  -a/--across <int>  Number of characters across in FASTA output (default: 60)
  -n/--names         Print reference sequence names only
  -s/--summary       Print summary incl. ref names, lengths, index properties
  -e/--bt2-ref      Reconstruct reference from .bt2 (slow, preserves colors)
  -v/--verbose       Verbose output (for debugging)
  -h/--help          print detailed description of tool and its options
  --help             print this usage message

[shawn.mek]

Just to clarify, I mean using the index - giving it a chromosome name (fasta header) and location numbers, and getting back a sequence.

I don't want to run an alignment, just pull out the sequence. So no SAM output.

For this I'm using bowtie, not bowtie2. But of bowtie2 can do this...

Thanks

[shawn.mek]

The bowtie-inspect thing does get all the info out, but thats 3gb of info since I can't select a location

[lh3]

Although bowtie index essentially keeps the genome, I doubt it is optimized or designed for your purpose. Use faidx if you only want to retrieve a few regions.

[shawn.mek]

I want to retrieve lots of regions efficiently, but thanks for pointing me to faidx, I'll see how it works.

[dpryan]

If you really have a LOT of positions, then it's best to read the genome into memory. samtools faidx is great for a smallish number of sites, but it grabs the sequence from disk, making it a bit slow for a large number of queries.

[shawn.mek]

yeah, I'm torn on holding it in memory or not. Toy with different workflows

[gringer]

Although bowtie index essentially keeps the genome, I doubt it is optimized or designed for your purpose.

The bowtie index is optimised for searching, but it's an overkill (and inefficient) for getting subsequences. If you want compressed indexed storage for just DNA sequence retrieval, then the 2bit format is probably best:

http://genome.ucsc.edu/FAQ/FAQformat.html#format7

The code points to a way to retrieve ranges:

http://genome-source.cse.ucsc.edu/gitweb/?p=kent.git;a=blob_plain;f=src/inc/twoBit.h;hb=HEAD

/* Parse a .2bit file and sequence spec into an object.
 * The spec is a string in the form:
 *
 *    file/path/input.2bit[:seqSpec1][,seqSpec2,...]
 *
 * where seqSpec is either
 *     seqName
 *  or
 *     seqName:start-end

So there's probably a program somewhere for getting subsequences out of that file using seqName:start-end notation.

edit: indeed, BLAT has such functions included. See here for a bit of discussion about 2bit retrieval using Perl:

http://www.perlmonks.org/?node_id=672251

[ctseto]

http://seqanswers.com/forums/showthread.php?t=34027&highlight=2bit

If you just need to retrieve known regions.

[tatianaorli]

You should be able to extract that information from the sam output. I've not used bowtie2-inspect before, but it could be what you are looking for.

bowtie2-inspect
No index name given!
Bowtie 2 version 2.1.0 by Ben Langmead ([email protected], www.cs.jhu.edu/~langmea)
Usage: bowtie2-inspect [options]* <bt2_base>
<bt2_base> bt2 filename minus trailing .1.bt2/.2.bt2

By default, prints FASTA records of the indexed nucleotide sequences to
standard out. With -n, just prints names. With -s, just prints a summary of
the index parameters and sequences. With -e, preserves colors if applicable.

Options:
-a/--across <int> Number of characters across in FASTA output (default: 60)
-n/--names Print reference sequence names only
-s/--summary Print summary incl. ref names, lengths, index properties
-e/--bt2-ref Reconstruct reference from .bt2 (slow, preserves colors)
-v/--verbose Verbose output (for debugging)
-h/--help print detailed description of tool and its options
--help print this usage message

This was extremely useful to me and my lab mates today, thank you so much for this comment back in 2013! I had a bowtie2 index file I got from a collaborator and extracted the original fasta sequences out of it using that program from the bowtie2 package. I would have never thought about doing it if I had not seen your answer here. I