When sequencing makes genotyping obsolete (soon) | Hacker News Reader

When sequencing makes genotyping obsolete (soon)

unz.com

18 pointscinquemb22 comments

The author makes the comparison between the cost of 23andme's genotyping arrays and the cost of whole genome sequencing (WGS), but the comparison with whole exome sequencing (WES) is probably more appropriate. For the unfamiliar, exome sequencing specifically sequences the exons (which are the components of genes). I'm not sure how many SNPs 23andme genotypes outside of exomes. Since the exome is 1-2% of the whole genome. You can imagine the savings when doing WES vs WGS (although the prep for WES is more costly). WES can be done for under $300 at this point (I'm not sure of the actual price point on the latest Illumina systems, but my guess is it's lower than this).

Anything that is interrogating risk for complex disease must look outside of exons. Most genome wide association study (GWAS) index SNPs, in diseases such as cancer[1], autoimmune disorders and psychiatric disorders, fall in introns of genes and the large intergenic regions. This means that it's hard to pinpoint the gene that is being effected, and if you're only looking at the exons of genes, you're missing almost the whole picture of inherited risk.

[1] from my paper in 2012: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3467035/figure/g...

I agree with you completely, but comparing from a consumer standpoint the exome gives you a lot more information than the SNP chips for a potentially similar consumer price point.

That's fair, but if you don't already present with a disease, I don't know how much WES can tell you. At least with a SNP chip, and the correct analysis, you can start speculating about what may happen down the road.

I agree completely. The nice thing about the exome is that the knowledge is more extensible than the SNP chip. If we discover new variants within the exome you won't need to be genotyped a second time. That logic probably argues more for WGS though. It is also the problem with WGS and WES. You get a ton of variants of unknown significance that the patient/customer will be curious about but be able to obtain no answers.

The terminology is not accurate. Sequencing can be a form of genotyping. What the title refers to is that sequencing makes array-based genotyping obsolete.

Yes - with Sequence-based Genotyping or RADSeq you use sequencing machines for genotyping

The upsides of using genomic or RAD reads compared to array-based is that you don't need to make an array first, and sequencing reads is still more expensive than using an array; but with an array you get a more complete set of called genotypes, and you don't get problems with any eventual bad-quality read library.

Nevertheless, at some point we'll drop array-based genotypings.

They are correct in calling out Illumina's de-facto monopoly rents they are extracting on the market, but sadly I don't share their wildly optimistic view that we are eminent for technological disruption that will re-start the price plummeting of whole genome sequencing.

Nanopores are no where near the throughput and accuracy of Illumina's sequencing by synthesis tech, and if there is a pathway to challenge Illumina's position, it will be extremely complex, iterative and _long_.

Meanwhile Illumina is amassing a billion dollar war chest and is adding its own complex and iterative improvements to its platform (two-color detection, longer and longer reads, higher cluster density), maintaining its market lead.

As much as the analogy to microprocessor manufacturing and Moore's law is alluring, the messy stuff of biology and single molecule chemical manipulation and sensor detection is unlikely to obediently follow the same innovation curve.

For a few years, DNA sequencing significantly outpaced Moore's law. Then, Illumina beat the competition so badly that they folded.

I've got something for nanopore sequencing coming out in the next few month. Stay tuned.

Do you use a different nanopore than α-hemolysin or MspA?

solid-state

Read some of your works, if its related, sounds pretty cool! Would love to hear more.

I think the real differentiator for a company in this field would be to let you pay full price for your genome, but with the guarantee that they will not retain any copies of it once you have received it.

Almost all sequencing centers will not hold your data much longer than a few months (and won't put it to any use in that time). The storage requirements are just too high for them to put them on tape and archive them long term. Companies like 23andme are obviously a different story.

EDIT: I'm talking about sequencing core labs that do sequencing as a service. They also have no need to keep the data because they don't necessarily have any metadata about the samples that they are sequencing.

"The storage requirements are just too high" Yeah? Can you elaborate on how big we're talking here?

100GB per person for the raw data from 30X WGS

So right now, one can buy 5 TB hard-rive for about $130 (lets ignore bulk buying because I have no knowledge of such).

If you theoretically could get 30 Million people at 30X WGS, that would be about 3M terabytes (600k drives) and you would need at least $78M, which seems a bit expensive, but it seems feasible if you were planning on capitalizing for at least a order of magnitude more, no?

Yes, in fact the UK is doing the first real population sequencing project right now

http://www.genomicsengland.co.uk/

This all isn't impossible. HLI is also sequencing well over 10000 genomes per year at this point. The service requested (having your data delivered and then deleted) already exists, but is not the most consumer friendly option.

BGI can theoretically sequence 100 human genomes per day now, with that rate only increasing: http://www.genomics.cn/en/news/show_news?nid=98952

So 100GB*100=10000GB=10TB per day, that's quite a lot of data but not unheard of in the age of Google. And from there you'd get the additional data from read-alignments, SNP-calling, genome re-annotation etc. pp.

Buying those 600,000 drives is probably only 10% of your costs, you need to run a data center if you want to actually use that data.

Also analysis pipelines generate intermediate results that are also large, and data is stored with 2x or 3x redundancy. So that 100GB might be more like 300GB once you factor in robust storage and bioinformatics.

Still, multiply that by 3, then 1.5 for the overhead… that's still about 5 months of QE :P

A post doc at who worked at ATLAS came to visit our lab a couple weeks a go, and just listening to the workflow they have for the events coming off the detectors… and all the BS they have to deal with working in academia… yeah… I'll take that money, erect a CSP system up in Spain and ask them to come do this with me, just to start off with a fun side project :P

Reply on news.ycombinator.com