very short deletion messes up SAMtools SNP calling

[eyalbd]

Hi,

I used Bowtie to align a solid read to the human genome, and now I'm trying to use samtools to call snps. When I use the pileup function, I get tons of snps, most due to a polymorphism (poly-c) in which there is a difference between my read and the reference. As samtools doesn't recognize this I get tons of snps from that point onwards.

I would appreciate any help on the matter.

Thanks,

Eyal

[lh3]

Use an aligner that is capable of gapped alignment. This is ESSENTIAL. No variant caller can work well with an ungapped aligner.

[eyalbd]

Thanks a lot for the reply!

Which aligners are capable of gapped alignment? I understand MAQ is, but I couldn't get it to run as I do not have access to a cluster, so I need a software that can run on my core i7 with 12gb (so 10gb max for alignment).

Many thanks,

Eyal

[Thomas Doktor]

Since Li Heng is too polite to suggest BWA I will recommend it, it's comparable to Bowtie in terms of speed and supports gapped alignments: http://bio-bwa.sourceforge.net/index.shtml

[eyalbd]

Thanks. I'll try it.

[eyalbd]

I tried using BWA, I used the supplied solid2fastq.pl file to create a gzip of my reads in fastq. Used the default settings for BWA, and later pileup, got me very bad alignment, with no connection at all between the reference genome (I'm checking only the mitochondria) and the consensus call.
What coud I be doing wrong?

[Thomas Doktor]

I'm not really sure since I don't work with SOLiD reads, but I think that BWA actually just uses the fastq format to store the colorspace reads in and uses ACGT as color representations. If you then try to align these fastq files normally, you will get many errors because of the nature of the colorspace encoding.

What you should do is generate a colorspace reference of your genome of interest and then align against that. The command looks like this for a human sized genome:

bwa index -a bwtsw -c genome.fa

You then align your reads in colorspace:

bwa aln -c genome.fa reads.fastq > alignment.sai

In any case, you should definetely always align colorspace reads in colorspace.

[eyalbd]

These steps are exactly the ones I followed. When I look at the SAM file now, I see many N's in the reads, in similar places in the sequence, for instance:

NGGNGNNNTAGGGNANNNANGCCNGNTNGNGNTNGNNNGATNGNCNNNN
NTCNTNNNAGTGCNANNNGNGTGGGNGNGNTNANCGNNGCGCGNANNNN

etc...

[Thomas Doktor]

Looks odd, how do the "raw" fastq files look?