Let's pretend I have $500K to spend

[DNAcowboy]

1st, thanxx for this forum. I found you guys right on time.
I'm not yet a user but I'm a bit familiar with the NextGen technologies from a concept perspective but for sure not from a user perspective.

However, I am in a new situation where I may have access to a financial support for a NextGen platform. But I want to understand if the work being made at our institution could really make use of a NextGen instrument. Then I may like to talk about the advantages of 454, Solexa and SOLID.

Because I'm in a Biomedical environment, most work being done here is resequencing, and is targeting a single gene, or a group of genes (disease associated...). To make it short, let's say that we may need to sequence 1 to 10 genes and small virus genomes (<100kb), something some of you may find small numbers, but for large cohort of patients. Even with Biomeks and CEquencers you know how much efforts it takes.

From what I have read in the 'industry' litterature, the NextGen instruments have not been defined for such projects. Am I right? They are more into large, large, large pieces sequencing for a single sample.

I understand the Roche PicoPlate can be divided into 16 runs (16 patients?) and I rapidly read this morning about their new Ligation Multiplex Identifier (MID) kit that may increase that number.

But are the Solexa and Solid approach compatible with such sequencing strategy for small number of genes? and for multiple sample runs?

Maybe I should stick with CEquencing?

[apfejes]

That's an interesting question. I only really have experience with the Solexa/Illumina platform, so I probably can't contribute much to a discussion of the other platforms.

I guess I'm not sure why you'd want to do Next-Gen sequencing if you're only doing 1-10 genes. It would be cheap and efficient to use traditional sequencing methods for that - primers and Polymerase are cheap, and you don't need to invest in the bioinformatics time/effort/hardware to use it.

Even assuming you have a large cohort, I'd think it would be more cost efficient to invest in a robot, or work with a sequencing centre with high throughput capabilities.

Btw, a single flowcell of data from an Illumina run is likely to cost you MUCH more than $500, anyhow, and wouldn't come close to providing the data you're looking for.

[DNAcowboy]

My appology, I had wrote $500K, not $500 (I made the change). I may even wrote €500K which is closer to what I can pretend. Of course, a nextgen run won't be even close to $500. So may I take the opportunity to share the feedback on economics I could retrieve in various places:

-Illumina GA- 1.3 Gb/run; 1TB data/run; $8950/run; $5,97/Mb; $5-600K/instrument.

-AB Solid- 3 Gb/run; data/run ?; $17447/run; $5,81/Mb; $500K/instrument

-Roche FLX- 0.5 Gb/run; 20GB data/run; $8439/run; $84,39/Mb; $500K/instrument

-Helicos- 1GB/hr; data/run ?; $18.000/run (or 50 channels for 50 patients and 150MB/patient @ $360/patient); $2.4/MB; $1.35M/instrument

Does everyone have similar economics?

Considering the number of genes to sequence which might be low, what if 4-5 different group could have the same throughput and combined sequencing assays on a same instrument? ie, during 1 run, analyzing 5x 5-10 genes? would it make more sens to use a nextgen?

[Malarky67]

The thing you are looking for is an efficient form of barcoding. A sequence that will be part of the end tags-- that will be sequenced as part of the reaction, identify the sample and then be ignored for further analysis.

There are barcode methods for 454-- and GATC have a proprietary method for Illumina

see for instance:

Error-correcting barcoded primers for pyrosequencing hundreds of samples in multiplex pp235 - 237
Micah Hamady, Jeffrey J Walker, J Kirk Harris, Nicholas J Gold & Rob Knight
Nature Methods - 5, 235 - 237 (2008)

This maybe useful for some applications: but you'll have to do the maths.

[DNAcowboy]

thanxx, I'll have a look at this paper.

[Chipper]

Hi,

the prices I have heard during seminars etc were more like ~3500$ for reagents for one flowcell for Illumina or $1000 per lane depending on who you ask....

For Helicos the price quoted on their homepage is $1.35 M for the instrument, not $2 M? And also $18K per run (2x50 lanes), where did you get $4K from? Sounds too good to be true...

For SOLiD it depends on the type of flowcell you use - a run with 8 lanes / spots is significantly more expensive than one with one spot only. I don't remember the exact price, but I think your quote is way to high. Is it for two slides? They do have a sytem for multiplexing that sounds promising in theory, but I do not think it is available yet. It would mean that you do not have to waste part of the sequence for the bar code.

Have you looked into the Polonator at all? It's only ~ $120k for the instrument so it would leave you with a nice budget for reagents... But like for the Helicos there is no actual data available from the final instrument I think?

What would the cost be per sample for this type of sequencing using CE?

[ECO]

Every service provider I've talked to quotes ~$4500 per lane on a Solexa flow cell...so $36k for a full 8 lane chip.

A ton of that has to be profit, even including the labor.

For a CE project, you can easily get <$2 reads from a big company.

[apfejes]

I talked to people here at the Genome Science Centre in Vancouver about pricing almost a year ago, and had much lower prices than those listed above. (somewhere around $6-8k CDN per flow cell.) Of course, that may have changed, and is probably different depending on the library construction required.

Still, I'm sure we're not charging $4,500 per lane, and we always use one lane for control, so you only get 7 lanes of data (not 8) per flow cell. (Do the other providers make you pay for the control lane as well, ECO?)

[ECO]

Originally Posted by apfejes

I talked to people here at the Genome Science Centre in Vancouver about pricing almost a year ago, and had much lower prices than those listed above. (somewhere around $6-8k CDN per flow cell.) Of course, that may have changed, and is probably different depending on the library construction required.

Still, I'm sure we're not charging $4,500 per lane, and we always use one lane for control, so you only get 7 lanes of data (not 8) per flow cell. (Do the other providers make you pay for the control lane as well, ECO?)

The providers I talked to (GATC Bio, SeqWright, Prognosys, BCGSC, Illumina...couple others) didn't mention additional control costs.

BCGSC was cheapest (still close to $4k USD IIRC), however my company didn't want to sign over all rights to the data, which was kind of a weird requirement for a service facility.

[apfejes]

Interesting - I don't claim to have any say in the policies or otherwise at the GSC (Nor am I trying to advertise for them) and I can totally see why you wouldn't want to sign over any rights. At any rate, I think they're targeting collaborations and academic partnerships, rather than commercial contracts, so that may be it.

Regardless of their goals, thanks for letting me know - I wasn't aware of that condition.

[DNAcowboy]

ok guys, i've modified helicos cost according a document called "Investor Presentation
JP Morgan 2008 Healthcare Conference
San Francisco, CA
January 10, 2008" that, if I remember well, I've retrieved from the Helicos web site. Sorry for the confusion with the $4k runs. Nextgen sequencing is a bit confusing not only from a technology stand point but also from an economics side.

[DNAcowboy]

Originally Posted by ECO

Every service provider I've talked to quotes ~$4500 per lane on a Solexa flow cell...so $36k for a full 8 lane chip.

I don't know yet how the GA works at the bench side, so does this means you can run individual lanes or do you need to run the full 8 lanes in one shot?

Originally Posted by ECO

For a CE project, you can easily get <$2 reads from a big company.

Do you mean $2 per 7-800b read? for both strands?
Price list, full reagent vol. with a BCI CEQ, I have $3/DNA extraction (automated), $2/PCR+purif., $14$/seq. reac.+purif. Total= $19 for 700bp 2 strands. ($20k/1Mb 2 strands).

At this point I want to understand this: what is the equivalence between 1Mb obtained from a CEquencer (which is, for me, equivalent to the sequencing of 1Mb in both directions) and 1Mb obtained from a NGS?

[apfejes]

Originally Posted by DNAcowboy

I don't know yet how the GA works at the bench side, so does this means you can run individual lanes or do you need to run the full 8 lanes in one shot?

Generally, they run all 8 lanes in one shot - it would be a waste of a flow cell to run it half empty, but because each lane can be loaded with a separate sample, most providers will let you buy single lanes of an 8 lane run.

As I mentioned before, some places will use the 8th lane for a control lane, meaning only 7 are usable per run.

EDIT: I just realized that ECO has posted a picture of the flow cell on seqanswers, which might help you get a visual image of how they're used. Here.

[DNAcowboy]

Originally Posted by apfejes

Generally, they run all 8 lanes in one shot - it would be a waste of a flow cell to run it half empty

This means you cannot "re-use" a flow cell. that would have run just a few lanes?

[apfejes]

That is an interesting question. I don't know for sure, but I can't really think of a great reason to do that, either. If you're running that few lanes of data at a time, it's probably more cost-efficient to contract the work out and not buy your own cell/reagents/GA. Since running the lanes "few at a time" would also require more control lanes (I assume you'd want a control lane with each run), you'd be wasting a lot of the capacity of the cell on the controls.

Of course, I've never even asked if the GA can run lanes separately. (Can the cluster forming station do its chemistry on single lanes at at time?) If the opportunity comes up, I'll ask the people here who are using it, but I'm pretty sure that the answer will be no.

[bioinfosm]

Originally Posted by DNAcowboy

This means you cannot "re-use" a flow cell. that would have run just a few lanes?

From what I know ... you waste the lanes that were left empty. no going back to use them later

[sci_guy]

DNAcowboy, if your main application is sequencing 1 - 10 genes in multiple patients then traditional Sanger sequencing would be best, unless you're wishing to sequence the same genes in scores of people, then a custom resequencing array would be more advisable. But, if you wish to do deep sequencing of tumour DNA for rare somatic mutations in your genes of interest then the read depth of next-gen seq is needed.

Applied biosystems has the VariantSeqR primer library - a good place start with primer design. If your gene of interest is reasonably sized it will probably take about 20 - 25 primer pairs to amplify the exons and 1kb upstream of TSS. For 10 genes lets say: 200 - 250 reactions/person. A liquid handler robot and Invitrogen E-gels will make your life easy here. If you do the math on PCR/Seq costs it will probably be somewhat cheaper than a 454/Solexa lane plus you can buy very mature and nice GUI based software to do your analysis (e.g Sequencher). It is still the Wild West when it comes on software pipelines with next-gen. If you are a Linux shell geek then you'll be fine there.

If you're sequencing a large pool of people the hefty Affymetrix resequencing array design cost and minimum chip purchase requirements is a non-issue and this option is worth looking into.

Next-gen seq of 1 - 10 genes would require up-front selection of your DNA from the genome, so you'll need short PCR exon amplification, LR-PCR tiling or Nimblegen chip-seq as a "front-end"... a cost of a few thousand per person here (unless you recycle your Nimblegen chips and buy a Maui hyb station, itself a $20K piece of equipment).

The viral genome sequencing application seems built for next-gen sequencing. If these jobs are few and far between then next-gen service sequencing would be a good option.

Hope this helps.

[DNAcowboy]

It sure does help, thanxx.

People around me, for sure are working on a small number of genes but you'll find funny how this number exponentially increases when you mention the nextgen sequencers capacity. We indeed work on rare mutations for various diseases and time has come to go further, over the well known mutations.

We also are onto small viruses genomes and that's part of the reason I would like to understand the math of running more than 1 sample on a NGS.

I actually own a 8 capillary-CEquencer and the math I have for 1Mb @ $20K should be indeed less on a 96 capillary system (with ABI chemistry) but presumably still over $10K. Will you agree on that kind of cost for home-CEquencing?

[sci_guy]

Originally Posted by DNAcowboy

I actually own a 8 capillary-CEquencer and the math I have for 1Mb @ $20K should be indeed less on a 96 capillary system (with ABI chemistry) but presumably still over $10K. Will you agree on that kind of cost for home-CEquencing?

Don't know about CEquencing pricing. The pricing I had calculated was for sending purified DNA in 96 well plates to a ABI 3730xl service sequencing centre.

If I plug your numbers into my budgeting spreadsheet I get:
Cost for one person $5K/person
Cost for ten people $2.5K/person

Costs include oligos, PCR, PCR cleanup and service sequencing. The reason for the large price drop per person when going from 1 to 10 people sequenced is recycling PCR oligos. BTW, Invitrogen offers oligo synthesis of full 96-well plates at a VERY competitive price.

If you have access to Affymetrix array hardware their resequencing chips are good in terms of value and throughput if you have numbers of patients and numbers of bases to sequence that much their wafer size. In the table below Array Format translates to the number of arrays etched into one wafer of silicon and Maximum Capacity is the max number of bases sequenced in both directions on dsDNA. Format-49 costs are a once only free of around $24K for chip design then $600 odd per chip.

Array Format Maximum Capacity
49 303,366
100 117,254
169 47,974

If you use next-gen-seq you will need to price in the "front-end" enrichment technologies (LR-PCR/PCR/Chip-Seq). This can really change the equation.

Originally Posted by DNAcowboy

We also are onto small viruses genomes and that's part of the reason I would like to understand the math of running more than 1 sample on a NGS.

Read some webinar presentations on the Illumina website. Good place to start.

[horigen]

Originally Posted by ECO

The providers I talked to (GATC Bio, SeqWright, Prognosys, BCGSC, Illumina...couple others) didn't mention additional control costs.

BCGSC was cheapest (still close to $4k USD IIRC), however my company didn't want to sign over all rights to the data, which was kind of a weird requirement for a service facility.

This site quotes a price of "about 800 dollars/lane for 18 cycles to 1,200 dollars for 60 cycles/lane".

http://www.genomecenter.ucdavis.edu/...equencing.html