ABI Color Space to Bases

[sgupta]

Hi,

I am trying to convert color space sequences generated by ABI SOLiD sequencer to actually bases using the following color space data "matrix":

AA=0
AC=1
AG=2
AT=3
CC=0
CA=1
CT=2
CG=3
GG=0
GT=1
GA=2
GC=3
TT=0
TG=1
TC=2
TA=3

So, this
>44_35_267_F3
T20220213203000111000122223221121222

gets converted to

>44_35_267_F3
CCTCCTGCTTAAAACACCCCAGAGATCTGTCAGAG

I want to do this to be able to use alignment programs that cannot work with ABI color space data. But so far I think I am doing something wrong because my alignment rates are less than 5% using published data (allowing upto 2 mismatches, mouse genome).

Any insights would be really appreciated.

I may just go ahead and use MAQ to do this in color space but I am not sure why this does not work the way I am currently trying to doing it. I am very new to SOLiD data so I maybe missing some piece of information here.

Thanks in advance.

[lgoff]

SOLiD Alignment

I have found that it is much better to do any analysis that you can in colorspace before you make the transition to DNA space. We are currently using the SHRiMP (U Toronto) alignment algorithm for fast and accurate alignment in colorspace. But even still 5% seems pretty low for DNA-space alignments of SOLiD data.

[lgoff]

Script available

To answer your original question, just send me an email and I will provide you a python script that will convert .csfasta to .fasta as needed.

Loyal
lgoff(at)broad.mit.edu

[Chipper]

Lgoff,
what advantage do you see with SHRiMP compared to the ABI tools? It is said to be very slow?

sgupta,
Direct conversion is not possible for reads that have any sequencing error since it will change all following bases in base space. Your coversion look correct though, but it is very common that sequences have at least one cs error. SOCS and ZOOM! are supposed to do colorspace alignments, perhaps worth a try.

[lgoff]

SOLiD

Originally, I was very put off by the SOLiD pipeline. It was initially very closed and there wasn't much I could do outside of the genome resequencing for which it was originally designed. The matching is relatively fast with SOLiD, but I do like the k-mer+Smith-waterman approach of SHRiMP. While the SOLiD pipeline has become much more robust. When we received our original machine, with the original cluster, it was underpowered for anything human. We had to re-develop our own pipeline for the specific applications we were using SOLiD for (smRNAs at the time). So we went with SHRiMP, and I have stuck with it since. Since we are lucky enough to be able to parallelize everything very nicely, the speed is not terribly an issue for us. I haven't tried the SOLiD pipeline in the past few months. Am I missing any dramatic improvements?

[ECO]

Hi Loyal,

Can you share your strategy for parallel processing with SHRiMP?

[Torst]

Can you share your strategy for parallel processing with SHRiMP?

Parallelizing SHRiMP is as simple as splitting your input fasta/fastq file into smaller ones, running SHRiMP on each, then merging the hits output file.

Nesoni (open source) does this automatically for you: http://www.vicbioinformatics.com/software.nesoni.shtml

[nilshomer]

Hi,

I am trying to convert color space sequences generated by ABI SOLiD sequencer to actually bases using the following color space data "matrix":

AA=0
AC=1
AG=2
AT=3
CC=0
CA=1
CT=2
CG=3
GG=0
GT=1
GA=2
GC=3
TT=0
TG=1
TC=2
TA=3

So, this
>44_35_267_F3
T20220213203000111000122223221121222

gets converted to

>44_35_267_F3
CCTCCTGCTTAAAACACCCCAGAGATCTGTCAGAG

I want to do this to be able to use alignment programs that cannot work with ABI color space data. But so far I think I am doing something wrong because my alignment rates are less than 5% using published data (allowing upto 2 mismatches, mouse genome).

Any insights would be really appreciated.

I may just go ahead and use MAQ to do this in color space but I am not sure why this does not work the way I am currently trying to doing it. I am very new to SOLiD data so I maybe missing some piece of information here.

Thanks in advance.

You can also do the conversion directly on our web server:
http://genome.ucla.edu/bfast-server/. Click on the left tab that says CS2NT/NT2CS and enjoy!

I would recommend aligning in color space since one color error will cause all bases after the color error to be translated incorrectly. Many great color space aware mapping tools exist.

Nils