1000genomes SRR107002.filt.fastq.gz bad format?

[wlangdon]

The first few lines of ftp://ftp.1000genomes.ebi.ac.uk/vol1....filt.fastq.gz are:
@SRR010375.1 VAB_S0083_BCM_20080731_1_Pilot_2_YRI_Fragment2199_1705_1499 length=35
T01000333201121302130101020001110011
+
!5-,'324)%,72%91/<9'50*+&%$(%$6&(.)1

I was expecting the second line to contain something like:
TGGGGCGCTTGTCATTCATTGGTTCCCCACAAACA

When fed into Bowtie it complains "Reads file contained a pattern with more than 1024 quality values."

Does anyone recognise the T01000333201121302130101020001110011
format?
Thank you
Bill

[Bukowski]

It's a fastq file from a SOLiD sequencer - so the base encodings are not in 'base space' but in 'color space'. It's 2 base encoding.

Take a look at:

http://marketing.appliedbiosystems.com/images/Product_Microsites/Solid_Knowledge_MS/pdf/CSHL_Fu.pdf

[wlangdon]

Dear Bukowski,
Many thanks for your rapid and helpful reply:-)
(I must admit I did not follow CSHL_Fu.pdf but I guess thats not necessary to use
the data.)

I am now using bowtie with --color. However I guess it will take 4 or 5 hours for
bowtie-build to create me a colorspace index.

BTW is there any reason why bowtie does not read-convert colorspace files
and use them with its usual indexes?
[I guess a more helpful error message would not go amiss either.]

Alternatively does anyone have a colorsequence to fasta conversion tool.

Many thanks
Bill

[Bukowski]

Dear Bukowski,
Many thanks for your rapid and helpful reply:-)
(I must admit I did not follow CSHL_Fu.pdf but I guess thats not necessary to use
the data.)

I am now using bowtie with --color. However I guess it will take 4 or 5 hours for
bowtie-build to create me a colorspace index.

BTW is there any reason why bowtie does not read-convert colorspace files
and use them with its usual indexes?
[I guess a more helpful error message would not go amiss either.]

Alternatively does anyone have a colorsequence to fasta conversion tool.

Many thanks
Bill

I haven't done any work with SOLiD data for a couple of years, but the di-base encoding means that if you have an error in a base when you do your color space > base space conversion (and there are tools for this, but I never used them) then all subsequent bases in the read are wrong (as each colour encodes the transition between bases).

There's a good (brief) overview of some of the considerations in the SHRiMP paper:

http://www.ploscompbiol.org/article/info:doi/10.1371/journal.pcbi.1000386

Consequently it is better to do the alignment in a color space aware tool, and if I worked with SOLiD data anymore that is what I would do. However some tools (such as BWA) have already dropped color space support.

[wlangdon]

Dear Bukowski,
Once again thank you for your very helpful reply.

It took just under 3 hours for bowtie-build to create a colorspace index
for the human genome (NCBI 37.5 ASM). It seems to be working well.

Thanks again
Bill

http://www.cs.ucl.ac.uk/staff/W.Langdon

[wlangdon]

Dear Bukowski,
Once again thank you for your very helpful reply.

It took just under 3 hours for bowtie-build to create a colorspace index
for the human genome (NCBI 37.5 ASM). It seems to be working well.

Thanks again
Bill

http://www.cs.ucl.ac.uk/staff/W.Langdon