Pooling samples in Sanger sequencing?

[Alex Coventry]

This is not really a "next gen" question, but I am not sure where else to ask. I would welcome suggestions for alternative venues.

I am considering using ABI 3730 machines to Sanger sequence a fairly large number of samples in approximately a 10kb locus. The question I am investigating is haplotype-oriented, and I am thinking about pooling the samples by haplotypes estimated from affy 6 genotypes. The idea is that the variants I am seeking are fairly common on these haplotypes, with a frequency of maybe 25%. The haplotypes will be rare and mostly occur on only one chromosome in each sample, so about one eighth of the chromosomes would carry the variants I'm targeting. I could tolerate a relatively high error rate in calling the variants, because the calls will be used in a kind of rare-variant burden test.

Is this just a crazy idea, because the Sanger data is analog? I'd be grateful for references to papers describing this kind of design, if any, or even just "this is crazy," if it is.

The trouble with sequencing the pools with nextgen technology is that the locus is relatively small, so structuring the samples efficiently would require something like a ludicrously complicated barcoding arrangement. However, if there is a sensible way to query such a small locus with nextgen tech, I would love to hear about that, too.

[CTGF]

Hi Alex,

I am not sure whether you reviewed Sanger sequencing reads. I reviewed a lot before. The major problem of Sanger sequencing calling is the noise under the peaks. So if you pool samples together, and see a small second peak under a big peak, it really difficult to tell whether it is a true variant or noise. Actually I think it is impossible to call. You can test it by sequencing several samples, and decide.

[Alex Coventry]

Thanks, CTGF. If you know of any papers which establish that smaller peaks are that uninformative, I would be grateful to hear about them.

[CTGF]

Thanks, CTGF. If you know of any papers which establish that smaller peaks are that uninformative, I would be grateful to hear about them.

I don't know about the paper, but I according to my own experience, it's difficult to tell whether the small peaks are noise or signal. I suggest you to pool several people with known genotypes of the region you are interested in, and Sanger sequence them. See whether you can call those SNPs.

[tbanks]

I haven't come across papers discussing the quantitative floor of Sanger sequencing but a number of papers on COLD-PCR do discuss the allele detection limits of Sanger. It is possible to detect a 10% allele dosage using Sanger but the trace needs to be perfect. In my experience you need to interrogate a position with both forward and reverse reads to have confidence in any allele occurring at <25%. It's also important to look at samples known not to have the allele as much of the noise in Sanger data is context dependent and you'll find some noise to be identical across different samples. You can try to reduce noise by sequencing from template generated from nested-PCR runs.

Good luck and let us know how it goes.

[sklages]

PolyPhred might be interesting .. http://droog.gs.washington.edu/polyphred/

[pmiguel]

You don't want to do this. It will not be robust. Do not pool them. Sequence them separately. Too expensive? Sounds like you are doing PCR reactions anyway, so 454 with fusion primers should work great for you.

--
Phillip