Sanger sequencing relation to MP sequencing

[TJK-OHSU]

Greetings,
I'm trying to come up with a brief but fair description of the relationship of Sanger (CE) sequencing to Massively Parallel sequencing.
Would you give me your take on this topic?
Here's what I have so far:
Briefly, a rule of thumb is that if the sequencing region of interest is < 60,000 bp than it is still cheaper to use the CE Sanger sequencing approach. Full "exome" capture and sequencing is currently ~$2,200 at the MPS Core. For that money you could sequence 88,000 bp at 5X coverage. You would have 800 nucleotide read lengths. So fairly easy assembly of the data to a reference sequence and very close to 99.99% accuracy. MPSequencing on the other hand gets ~75 nt read lengths, so assembly and mapping to a reference sequence is more difficult and accuracy, last I've read with >20X coverage is ~99.9%. So 1 error in 1000 bases instead of 1 error in 10,000 bases.
It is for the latter reason that Sanger sequencing is used to validate sequence variants discovered with MPsequencing.
In addition closing gaps between contigs (assembled shorter reads into longer contiguous regions) is usually accomplished with Sanger sequencing.
And because of the higher error rate for individual MP reads, the reads that are not "seen" multiple times are filtered out. This can make it more difficult to find variations that occur at a low frequency if not all cells contain the variation. E.g. looking for causative mutations in cancer research.
Your help with this would be greatly appreciated.

[westerman]

I'd say that your 60,000 bp estimate is correct. We can do Sanger sequencing at about $1 per read. The cheapest NGS sequencing we can do run about $400 (a lot of that is library construction cost). Thus spending $400 on 400 Sanger reads times 800 base pairs divided by 5x coverage is 60,000 bp. The NGS sequencing at $400 would provide, oh, around 500M bases which, even at 50x coverage, would provide 10M bases. 60K versus 10000K ... such a hard choice!

On the other hand there are all sorts of caveats to those figures. For one thing in order to get that $1/read number we require a 384 well 2-primer project; thus your minimum cost is closer to $768. As for the NGS number, that would be a barcoded sample on 1/12th of a lane. We would then need to find another 11 projects to fill up that lane before we could run your sample. The NGS assembly (even at PE-100 reads) is likely to be more spotty than the Sanger assembly at 800-bp reads.

[krobison]

Greetings,
I'm trying to come up with a brief but fair description of the relationship of Sanger (CE) sequencing to Massively Parallel sequencing.
Would you give me your take on this topic?
Here's what I have so far:
Briefly, a rule of thumb is that if the sequencing region of interest is < 60,000 bp than it is still cheaper to use the CE Sanger sequencing approach. Full "exome" capture and sequencing is currently ~$2,200 at the MPS Core. For that money you could sequence 88,000 bp at 5X coverage. You would have 800 nucleotide read lengths. So fairly easy assembly of the data to a reference sequence and very close to 99.99% accuracy. MPSequencing on the other hand gets ~75 nt read lengths, so assembly and mapping to a reference sequence is more difficult and accuracy, last I've read with >20X coverage is ~99.9%. So 1 error in 1000 bases instead of 1 error in 10,000 bases.
It is for the latter reason that Sanger sequencing is used to validate sequence variants discovered with MPsequencing.
In addition closing gaps between contigs (assembled shorter reads into longer contiguous regions) is usually accomplished with Sanger sequencing.

While I think you are generally on the right track, you are muddying the waters by comparing exon capture + NGS to Sanger sequencing; in the first you seem to assume you don't yet have your target region but in the second you apparently have it cloned for Sanger. If you need to get at the target by PCR, then the cost of oligos & PCR must be included. You must also plan for PCR failures.

For NGS, some other comparative numbers would be $1500-$2500 to do a single Ion Torrent run at a provider. If you had an instrument, the now $99 314 chip would be gross overkill for a 60Kbp fragment, and even with $200-400 or so in library prep costs would have a total cost below Sanger. There are also a number of outfits claiming to offer exomes down in the $1K range, though I suspect you need a large order for that.

And because of the higher error rate for individual MP reads, the reads that are not "seen" multiple times are filtered out. This can make it more difficult to find variations that occur at a low frequency if not all cells contain the variation. E.g. looking for causative mutations in cancer research.

If you have a mixed population of cells, as with cancer samples, Sanger is simply less sensitive. Unless your sample is 80+% tumor, your odds of finding heterozygous somatic mutations are quite poor with Sanger, but can be quite good with MPS. This is because the Sanger sequencing, unless you clone out individual products (more cost), gives a signal based on the population average -- if an allele is only 10% of the population it generates 10% signal, which will generally be missed with Sanger. Conversely, MPS techniques count individual molecules, and so the ability to detect rare mutations is a function of read depth and error rates. Detection of <1% mutations has been reported multiple times in the literature using 454 or Illumina.

[pmiguel]

Actually I think Sanger sequencing is a waste of money on any project that can be deployed on a next gen sequencer.

Current price for a 25,000 rxn kit BigDye Sequencing (Sanger sequencing) Kit from Applied Biosystems is right around $197,000. Now assume you are doing 1/32nd reactions. That makes a 25K kit capable of doing 800,000 reactions. Roughly $0.25/read -- just in Big Dye reagent. We ignore polymer, buffer, array, service contract -- all other costs. Lets give you 800 bases/read. That works out to $312.50 per megabase of sequence, just in Big Dye.

Illumina gives sequence at around $0.10/megabase in reagents. (By the way, probably most price effective to do 2x100 paired end reads on ~400 bp fragment libraries.) A GS-FLX/GS-FLX+ around $10/megabase.

The real problem is that it takes a lot of effort to convert Sanger sequencing methods over to Next Gen methods. Especially when you need to add in bar coding to pack enough projects into a single lane/region to make the buy in price for the project reasonable.

--
Phillip

[TJK-OHSU]

Good points.The fact that you can increase the read depth using MP, essentially as high as you would like, does overcome the inherently higher error rate.

Back to the relation of the two technologies though.
The cost issue becomes calculation of the current cost per lane times the number of lanes needed to answer your question. ~60Kb, for now, seems to be a reasonable rule of thumb. Though as krobison points out, the PGM (Ion Torrent) may soon change that.
When groups need to confirm variations found by MPS, they are working on a smaller region of interest, so the cost analysis favors Sanger sequencing, for now.

[TJK-OHSU]

Actually I think Sanger sequencing is a waste of money on any project that can be deployed on a next gen sequencer.

The real problem is that it takes a lot of effort to convert Sanger sequencing methods over to Next Gen methods. Especially when you need to add in bar coding to pack enough projects into a single lane/region to make the buy in price for the project reasonable.
--
Phillip

There are also the informatics and time per run factors to consider.

[niceday]

Hi Pmiguel,

You mentioned you were doing 1/32nd reactions. I'm assuming you are either diluting down, reducing the reaction volume or both.
Are you able to give me any information on this?
I have a bit of work to do and need to drive the cost down as much as possible.
Thanks.

[pmiguel]

For some applications we use 0.25 ul of Big Dye in a slightly less than 1/2 volume reaction. (7.5ul final volume). That puts us at 1/32nd reaction. I think we do about 120 cycles of themal cycling.

ABI supplies their mysteriously named "5X buffer" -- you have to make up what you don't add in Big Dye with that. As of version 3.1 of Big Dye, just using a Tris and MgCl2 homemade buffer does not seem to work. Also, adding 5% DMSO, is no longer helpful -- in fact it appears to have negative effects.

Okay, there are lots of other factors in play though. But basically it comes down to Big Dye being much "weaker" at 1/32nd reactions than at lesser dilutions. That is, contaminants that don't impact results at lesser dilutions are show stoppers at 1/32nd reactions. Generally these are the usual culprits: SDS, ethanol, etc. But at the higher dilutions, stuff like acetate (as in the sodium acetate you use a co-precipitant) are inhibiting the polymerase -- especially for long reads.

So signal droop becomes your main enemy. That is, signal is strong at the beginning of the reaction, but drops to nothing a few hundred bases out. This is especially an issue, because you probably want to resuspend your reactions in water, not formamide, because water gives about 5x the signal of formamide. But water also favors shorter products during electrokinetic injection-- especially if there is any salt around.

To counteract signal droop we had to reduce the amount of acetate coming in from the DNA prep by lowering the alcohol concentration for the DNA precipitation out of the lysate. That isn't enough, though. We also had to double the normal extension times for the sequencing reactions.

As far as cost savings go, you need to have a sense of proportion. If you are spending 10% of you budget on the sequencing reaction and 90% on DNA preps and clean up reagents, then focus on lower the costs of DNA preps and reaction clean up reagents. There will be more "fat" there to be cut.

Also, keep in mind that you want to push towards simplification, if possible. The simpler your protocol is, the fewer places there will be to make an error.

Finally, if you are trying to save money: don't do Sanger sequencing. Find a way to use a next gen sequencer to accomplish your goals.

--
Phillip

[niceday]

Thanks Phillip. I've been doing nextgen for a few years but have been asked to look at an established capillary setup to reduce costs.
My first call is to seqanswers before I start spending money to save money.

[bertgold]

not right imho

Briefly, I want to say alot of the comparisons above are incorrect but I can't provide all details at this moment as I am in 'manuscript in preparation'. I am using 2 microliter reactions for big dye in a 384 well plate and am simultaneously giving these for 454. Final answers are not all in, but, the barcoding is not a problem, the big dye is working fine, and the real costs are the AmpPure, CleanSeq (on the Big Dye) and various kits needed for 454. I would happily to to MiSeq if anyone wants to offer me a machine. No, not send out. I have discovered in years of service that without absolute control of every parameter, perfecting new methods is a waste of time.

PS- Anything written here is my personal 2 cents. ID info is provided for just that, ID only:

Bert Gold, Ph.D., FACMG
Staff Scientist
Center for Cancer Research
National Cancer Institute
Boyles Street; Box B
Fort Detrick
Frederick, Maryland 21702
Phone: 301-846-5098
Fax: 301-846-7042
E-Mail: [email protected]
Website: http://ccr.cancer.gov/Staff/Staff.asp?profileid=7351

[pmiguel]

Briefly, I want to say alot of the comparisons above are incorrect but I can't provide all details at this moment as I am in 'manuscript in preparation'. I am using 2 microliter reactions for big dye in a 384 well plate and am simultaneously giving these for 454. Final answers are not all in, but, the barcoding is not a problem, the big dye is working fine, and the real costs are the AmpPure, CleanSeq (on the Big Dye) and various kits needed for 454. I would happily to to MiSeq if anyone wants to offer me a machine. No, not send out. I have discovered in years of service that without absolute control of every parameter, perfecting new methods is a waste of time.

Hi Bert,
We just use ethanol to clean up Big Dye reactions. Is the CleanSeq really more expensive per reaction than the Big Dye itself? I mean with 2 ul sequencing reactions are you not pushing $2/reaction just in Big Dye costs?

--
Phillip

[bertgold]

I am using 0.05 microliters big dye per well in 384 well plate. According to our numbers here that is less than 4 cents per well. I would happily use Ethanol instead of CleanSeq if it will really work. Beckman/Agencourt would hate that of course. Do you have a protocol I could adapt to an FX robot?

[bertgold]

It occurs to me that you probably centrifuge and aspirate the Ethanol. That will not work for the throughputs we are using.

Yes, the CleanSeq at $ 20,000 per liter (appoximately) is far more expensive than the big dye per reaction.

[bertgold]

Incidentally, I resent this web site's assertion that I am a 'junior' member. I am 57 years old with more than 30 years of experience. I would hardly call myself 'junior'.

[pmiguel]

Incidentally, I resent this web site's assertion that I am a 'junior' member. I am 57 years old with more than 30 years of experience. I would hardly call myself 'junior'.

That is just the default. You should be able to edit your profile and put whatever you want in there. Go to the left top corner of a page and click "User CP". This stands for "User Control Panel". Then click "Edit Your Details". If you edit the "Custom User Title" box, you can change the "Junior Member" title that is the source of your resentment.

--
Phillip