Sequencing and genotyping one gene

[gavin.oliver]

Can anyone recommend how best to sequence a single gene (25KB) for SNP/Indel-calling and genotyping purposes.

Sanger is fine for sequencing the gene once but multiple copies are required for the application and it quickly becomes very expensive.

Most sequencing centres seem to tend towards their HiSeqs which would mean 1500 x coverage or something similarly crazy (and that is with multiplexing).

Can anyone suggest an alternative and/or a provider?

Thanks in advance.

[krobison]

Probably design PCR primers (which will require qualification) and amplify it.

First decision is whether to design long PCR amplicons which require fragmentation or short amplicons which might go directly on a sequencer with sufficient read length. Also, your platform choice may be driven by your turnaround time needs. Another key question is how many different samples are you going to try to sequence simultaneously, and for what purpose?

For example, suppose you expect to get samples infrequently but want to analyze them quickly. That is going to steer you towards a short amplicon strategy. Similarly, speed and many samples will probably put you in that neighborhood, as you can barcode the amplicons with fusion primers.

On the other hand, if you will have a large number of such samples and don't care about speed, then perhaps long PCR with fragmentation is the way to go.

You might also look at the Olink selector technology; it may not pay off though unless you have a lot of samples.

In all of these cases, you should be able to find a service provider or core lab that can either do the run on a "benchtop" sequencer (Ion Torrent or 454jr now; MiSeq in future) or run it as a lane mixed with other projects. This, of course, needs to be designed in from the start. In any case, you shouldn't be paying more than $5K per sample, and should be able to make it much less (depends on platform, strategy, etc).

[gavin.oliver]

Approximately 50 samples for the purpose of SNP/Indel analysis and genotyping. I had assumed 454 technology would be no good here because of the Indels. So it's OK to use traditional PCR approaches (instead of an enrichment kit like Sureselect) and then sequence with NGS?

[rocksd]

Approximately 50 samples for the purpose of SNP/Indel analysis and genotyping. I had assumed 454 technology would be no good here because of the Indels. So it's OK to use traditional PCR approaches (instead of an enrichment kit like Sureselect) and then sequence with NGS?

I just sent a personal email to you so if you wish you can contact me as I am dealing with a similar request.

Rocksd

[HeidiJTP]

I realize this is an older thread, but I have a similar situation and was wondering which approach you used. I want to genotype a full gene and include 5' regulatory sequence as well for approximately 10 kb. Using long-PCR and PacBio sequencing would still require fragmentation. How can you reconstruct full alleles when genotyping large regions?

In other words, if one individual is heterozygous in both the 5' proximal promoter as well as in exon 5, but these are sequenced as separate fragments, is it possible to determine the complete allele 1 and allele 2? Would it be better to amplify shorter, overlapping fragments and hope there is enough polymorphism to allow assembly of each allele?

Thanks for any advice you can offer!

[lkral]

I am patiently waiting for the MinION to come out. I also want to sequence ~10 kb PCR products and obtain full allele sequences.

[HeidiJTP]

MinION does sound amazing, but I don't think I can wait that patiently

What about the methods described by Lindsay, Bonfield, and Hurles (2005) "Shotgun haplotyping: a novel method for surveying allelic sequence variation" Nuc Acids Res 33(18)

They suggest that haplotypes for long PCR products can be phased using paired-end reads and high coverage. If so, PacBio would even side-step the need for paired-end reads, as polymorphisms should show up within their ~3 kb reads and allow for assembly of each allele. Anyone more familiar with error rates and such think this is plausible?