IUPAC coded reference sequences

[data]

I'm just getting started, so hopefully this isn't a silly post.

Mosaik and some other aligners are able to use IUPAC codes in their alignment process.

I was wondering if there are any publicly available whole genome reference sequences available that contain the IUPAC codes for known SNPs. If not, I was curious if there are any applications that can apply these codes via data from dbSNP, etc., or if I would be better off coding something to handle this.

Thanks!

[nilshomer]

I'm just getting started, so hopefully this isn't a silly post.

Mosaik and some other aligners are able to use IUPAC codes in their alignment process.

I was wondering if there are any publicly available whole genome reference sequences available that contain the IUPAC codes for known SNPs. If not, I was curious if there are any applications that can apply these codes via data from dbSNP, etc., or if I would be better off coding something to handle this.

Thanks!

A quick perl script from your local bioinformatician should be able to handle the generation of a IUPAC reference from dbsnp. How will you deal with indels in dbsnp?

[data]

I've been looking into scripting my own solution, but I ran across what appears to be a good place to start:

http://hgdownload.cse.ucsc.edu/goldenPath/hg18/snp130Mask/

I don't have a good answer about how to represent indels in the reference sequence.

[batbeast]

Data,

SNPmask should work fairly well. Obviously this is a moving target with new dbSNP/1000g releases.

I've been thinking of establishing some modified IUPAC reference sequences and posting them eventually.

You may want to also see: "An extended IUPAC nomenclature code for polymorphic nucleic acids" Bioinformatics 2010 26(10):1386-1389; doi:10.1093/bioinformatics/btq098

[Jabu]

The use of a reference genome that includes IUPAC codes for known relatively frequent variants in an alignment is an elegant idea.

I am wondering if the currently available aligners (BWA, Bowtie, etc.) can handle this type of genome build as a reference?

It would be nice to use the 1000 genomes data for this purpose, instead of dbSNP which contains a large fraction of non-frequency validated SNPs.

Does a human reference build of this type currently exist?

Also, the question still remains on how to deal with indels?

[nilshomer]

Try Mosaik for the IUPAC codes.

[JimC]

Try Mosaik for the IUPAC codes.

Nils,
I can't find any clear statement about whether BWA or Bowtie handle IUPAC codes in reference sequences (or possible errors when using reference sequences with them included).

Can you point me to any documentation or explanation about BWA or Bowtie and IUPAC ?

I'm working on a bacterial genome which contains some small number of IUPAC characters in the reference.

Thanks

Jim

[nilshomer]

Neither support them directly (I had to check the source code for BWA). BWA randomly converts them to a DNA base, not sure what Bowtie does.

[ohofmann]

NovoAlign also handles IUPAC reference sequences. There is a difference in how these are being handled -- Mosaik considers any base aligned to an IUPAC position a partial mismatch, NovoAlign does not penalize them.

(Note: Might be the other way around, can't check the documentation right now)

[nilshomer]

NovoAlign also handles IUPAC reference sequences. There is a difference in how these are being handled -- Mosaik considers any base aligned to an IUPAC position a partial mismatch, NovoAlign does not penalize them.

(Note: Might be the other way around, can't check the documentation right now)

TMAP considers them as a mismatch during seeding but as a match for the final alignment. IUPAC positions are difficult to handle when most aligners assume the input is a simple reference string rather than a regex or graph structure.