Is Sanger better than Next-Gen for Anything?

[ejphd2012]

Do you think there are any applications for which Sanger sequencing (i.e. capillary electrophoresis, CE) still holds a competitive advantage over Next-Gen technologies?

I thought that as of a few years ago,
1. Sanger was still needed for sequencing gaps between the short reads of NGS, and
2. NGS was more error-prone than Sanger, and in particular NGS didn't do well with PCR errors e.g. for sequences containing short tandem repeats

But I don't know if that's still true, and I hear that MiSeq has nailed the lid on the coffin for Sanger.

Are there applications for which researchers still need Sanger sequencing as a follow-up to their NGS results?

For users who still maintain their CE machines (e.g. ABI 3730, etc), what do you use them for? I also saw some reports from a few years ago that CE machines could be re-purposed for glycosylation profiling instead of DNA sequencing. Any other creative ideas for soliciting samples to keep the machine running?

Thanks for sharing your ideas!

[dpryan]

Sanger sequencing is still often a bit longer and less error prone than NGS, so when you're looking to confirm mutations (rather than screen the whole genome for them), this is still a good tool. Likewise, in molecular biology labs, outsourcing very small amounts of sequencing (from subcloning or targeted mutagenesis) and then using a MiSeq is often overkill. In my PhD lab, we had a couple older ABI machines, which were handy since I could just sequence a fraction of a plate as needed. Actually dealing with the chromatograms takes a lot less effort (in the context of small amounts of data) than trying to use the standard NGS tools. So, for many labs, NGS may prove to be more work.

[GenoMax]

Not every application/project needs a ton of sequence. At under $3-4 dollars a sample Sanger sequencing will always remain the cheapest way to check on constructs/PCR products etc with a quick turn around. It still is the definitive way to confirm findings for clinical applications.

So no I do not think Sanger sequencing is going to disappear any time soon.

[JamieHeather]

Exactly what GenoMax says - if I just want to double check a single plasmid, there's no need for anything more than 1 sequencing reaction.

Just because we invented printers we didn't throw away the pens.

[microgirl123]

Our facility still runs several hundred Sanger sequences every week. If you're just checking a PCR reaction or a clone, you don't need a thousand dollars worth of next-gen sequencing done!

[westerman]

Ditto what microgirl123 said. Our core facility runs hundreds if not thousands of Sanger sequences every week (5000+ during Aug 2013). Our customers would not be doing so much sequencing if they did not find it valuable.

[krobison]

Originally Posted by GenoMax

Not every application/project needs a ton of sequence. At under $3-4 dollars a sample Sanger sequencing will always remain the cheapest way to check on constructs/PCR products etc with a quick turn around.

Always is a very dangerous word to use. While there isn't an obvious candidate on the horizon, there is always a possibility of another technology delivering into this space.

It still is the definitive way to confirm findings for clinical applications.

At some point the clinical folks will cotton on that this is often a belt-and-suspenders measure.

[bios90]

So, is the cost per sample really only $3-4? Is there any other costs, like, shipping, handling, storage etc., that are significant? In my mind I was thinking customers might pay ~$100 to get a single amplicon sequenced in one day (maybe because I am still catching up on the sequencing field). Is that number way off? I guess what I am trying to learn is, how much would customers pay to get single-few amplicons sequenced, if they can do it quickly, like in real time.

Thanks.

[chadn737]

It depends upon your institution/business. At the Universities that I have worked at, there are core facilities and no shipping costs. It takes a day or so to get results, but per sample the cost is at most $6. I am never in such a rush that I would pay much more than that to have a single amplicon sequenced.

[dpryan]

When I was doing my PhD, we paid $1 per well of a 96 well plate. The company picked it up from one of our 4 degree fridges and we could typically download the sequencing data the next afternoon. The only other costs are some supplies (tubes and such) and whatever you need for the PCRs (primers, taq, etc.). If you're doing a fair bit of just one area, e.g. looking for mutations in one gene in many people or checking constructs, then Sanger sequencing is still quite economical and quick.

[westerman]

The significant costs are two fold. BigDye and personnel. Our (university core facility) prices range from less than $0.50 USD (you do all of the work and just give us a plate to run on the sequencer) to $9.20 USD (we do all the work and repeat as needed until you are satisfied with the results). This is University pricing. Outside work is higher. Some internal work is lower.

Of course this doesn't count in shipping. I'd say that 99.8% of our customers just walk in their samples.

While I agree with Robison that it never pays to say "always" it is going to be hard to beat 3730 pricing. I don't see it occurring in the near future.

[Jeremy]

Sanger was never really designed for genome scale applications (that's why the first human genome took so long and cost so much money), so the development of next gen systems barely affected it. Sanger is still used for what it was designed for, sequencing target regions, checking inserts, checking for plasmids, sequencing for mutations etc. etc.
Sure next gen can sequence amplicons, but if you only need to check one locus in 50 samples for example, it is cheaper and faster to just do Sanger sequencing.

[krobison]

Originally Posted by bios90

So, is the cost per sample really only $3-4? Is there any other costs, like, shipping, handling, storage etc., that are significant? In my mind I was thinking customers might pay ~$100 to get a single amplicon sequenced in one day (maybe because I am still catching up on the sequencing field). Is that number way off?

Yes. There are a number of services that will sequence essentially overnight for <$10/sample. FedEx shouldn't cost you another $80 (dry ice really isn't necessary for a clean DNA sample).

[relaswar]

I agree with most of the posts above. Additionally, I would like to emphasise that the main 'off-putting' aspect of NGS is still data analysis: both cost-wise and time-wise. While with Sanger sequencing data, even a biology graduate could grasp it very quickly and churn out the results very quickly, NGS demands specialist knowledge and trained bioinformaticians to analysis the mountains of data (and loads of computing power!). This incurs good amount of cost and time. So, unless the issue of data analysis is solved, NGS would still have applicability for high throughput projects only. The other issue is lack of proper standardization for data analysis methods. There is a need to develop robust QC metrics for data comparison between platforms and between various software outputs. I hope a day will come when we run a sequencer overnight and capture and analyse the data next morning, ready for presentation in the afternoon. Or am I being very naive?

[westerman]

Originally Posted by relaswar

I hope a day will come when we run a sequencer overnight and capture and analyse the data next morning, ready for presentation in the afternoon. Or am I being very naive?

It depends on what you want. For a miSeq run (a "small" amount of data ... at least compared to a hiSeq) we can run the sequencer, if not overnight, at least within a day. Then demultiplex, filter, map to an known genome, get SNPs, etc. by the afternoon. But all of that is routine and can be automated. Anything requiring thought or extensive comparative analysis will take more time. But that is true with Sanger sequencing as well. Our core facility will generally take our customers' Sanger samples up to noon, prep them, put them on the sequencer, get results overnight and provide raw sequences, traces, blast results, restriction map, simple sequence comparison by the next morning. What our customers do with this information, and how long they take to get around to doing more in-depth analysis, is not for us to know.

If we were in a clinical setting with a more pre-determined workflow (instead of our typical 'sequence an unknown plant/animal in a variety of ways') then even more automation could be done.

So, the point of this long and rambling message is that, yes, NGS sequencing and analysis can be done rapidly especially with the benchtop sequencers but you have to know beforehand what type of analysis you want.

[krobison]

Originally Posted by relaswar

I agree with most of the posts above. Additionally, I would like to emphasise that the main 'off-putting' aspect of NGS is still data analysis: both cost-wise and time-wise. While with Sanger sequencing data, even a biology graduate could grasp it very quickly and churn out the results very quickly, NGS demands specialist knowledge and trained bioinformaticians to analysis the mountains of data (and loads of computing power!). This incurs good amount of cost and time. So, unless the issue of data analysis is solved, NGS would still have applicability for high throughput projects only. The other issue is lack of proper standardization for data analysis methods. There is a need to develop robust QC metrics for data comparison between platforms and between various software outputs. I hope a day will come when we run a sequencer overnight and capture and analyse the data next morning, ready for presentation in the afternoon. Or am I being very naive?

But this really isn't the fault of NGS, but rather the scale of the problem you are addressing. When Sanger sequencing was done at scale to sequence ESTs, BACs, genomes etc., the same issues you lament were present.

New entrants such as GnuBio have emphasized that their boxes will spit out interpretable results, not raw data. But if the questions you are asking are complex, then expecting instant results->intelligent presentation still won't be realistic.

[bios90]

I wonder what technology can displace Sanger, and if that technology could open more markets. For instance, if Illumina or Oxford or GnuBio came up with a sequencer that matches Sanger spec-for-spec (in readlength, error profile, and throughput), but provide real-time answers and be portable - Would this be disruptive, or only incremental over Sanger's stranglehold on low-plex applications? Reading the comments above, there doesn't seem much urgency to find a better version of Sanger in the markets it serves.

Alternately, is there a market for a portable amplicon sequencer (1 to 96 amplicons) with similar price and performance specs as Sanger? What other attributes does it need to have to be considered disruptive or groundbreaking?

[ajthomas]

Originally Posted by bios90

Alternately, is there a market for a portable amplicon sequencer (1 to 96 amplicons) with similar price and performance specs as Sanger? What other attributes does it need to have to be considered disruptive or groundbreaking?

The only thing I can forsee displacing traditional Sanger sequencers (e.g. ABI 3730) is a small scale instrument that any lab could buy for perhaps <$15,000 and run perhaps 1-12 samples at a time and yield similar results. Pretty much anyone can send a DNA sample to their institution's core facility or a commercial company and have sequence back within 24-48 hours for <$10. The only thing that could make that better would be if you could do it in your own lab so you could have the data within a few hours for the same or less cost.

It's already very cheap, so getting even cheaper wouldn't make much difference. There is room for improvement in time, but not much room.

[gringer]

I feel like doing some vapourware calculations. What would it take for a MinION to get there? Taking ball-park figures, I think they can sequence 500 nanopores at something like 1 base per second for 6 hours at a cost of ~$1000. Say you need 10 copies for a sanger-accurate sequence:

500 pores / 10 = 50 "good-quality" bases per second * 6 hours ~= 1Mbp of good-quality sequence per run, so ~$1 per kbp at $1000 per run

That sounds reasonable. Of course, the cost-per-run is always a killer.

[gringer]

Originally Posted by gringer

500 pores / 10 = 50 "good-quality" bases per second * 6 hours ~= 1Mbp of good-quality sequence per run, so ~$1 per kbp at $1000 per run

Updating this with genlyai's numbers. There are two significant differences from my original estimate:

• high-quality sequencing speed (10bp/s vs 1bp/s)
• run time (20h vs 6h)

400 pores * 0.01 kbp/s = 4 kbp/s ~= 0.4 "good-quality" kbp/s
0.4 kbp/s * 20h = 28800 kbp of good-quality sequence per run
$1000 / 28800 kbp = $0.034 per kbp at $1000 per run

... which is quite a lot less.