Exome Sequence Offer

[TomBAvoider]

I received this email offer from Promethease. I'm thinking of taking them up on it.

 

"As someone who has received Promethease DNA (genome) reports from us, we would like to tell you about a new DNA sequencing service and a discounted price available only for a limited time to our users.

Over ten years ago we established SNPedia as a database of scientific literature about human DNA variants, and we created Promethease to help you learn about the variants you carry. Since then we've seen SNPedia's content double almost every year, and Promethease's interface has improved to enable better searching, filtering and exporting options. If you haven't seen us lately, here's a typical report: http://files.snpedia.com/reports/promethease_data/Genos_example1_ui2.html 

Regardless of the growth in SNPedia, a Promethease report can only say something about the variants listed in a person's "raw" DNA data file. Most users have DNA data generated using microarray platforms (DNA chips), and typically that means only a minority of the interesting DNA variants were actually tested. 

We're pleased to announce that a new company, Genos, is now producing a reasonably-priced sequence-based genome service that tests the majority of the DNA variants in all of your genes, and they have worked with us to ensure Promethease compatibility. 

Their current price for sequencing your whole exome is $399, but until Dec. 12, as a Promethease user you can purchase as many as you like for $359 (a $40 discount), and for each exome we will run a new Promethease report at no charge.[...]

 

[...]

 

While we benefit from having more exomes to analyze, which helps us curate more variants for SNPedia, you benefit from having more comprehensive data. Additional advantages to having exome sequence are detailed here. 

We remain as committed as ever to helping genome information make a difference in your life, and until inexpensive full genome sequencing becomes available, we see exome sequencing as the best choice to complement and expand the DNA chip-based data you already have for yourself and family members. Please feel free to contact us with any questions, and to also check our Genos FAQ page. 

Sincerely, 

Mike Cariaso & Greg Lennon 

info@SNPedia.com "

 

Does anyone have any thoughts about this?

[TomBAvoider]

I decided to go ahead. I'm in CA, so with taxes it's almost $400. I received a confirmation email. I'll report here on the experience - the whole process is supposed to take about 3 months, and as soon as I get the data, I'll run it through Promethease and compare with the 23andme report that comes through Promethease.

[TomBAvoider]

Well, I now have the results, and as promised I'm filing a report on my experiences.

 

My sample was received by Genos 12/12/16 and my sequenced data was available 03/02/17. 

 

Genos makes the raw data of your sequence available as a download in several formats: VCF (variant call format) ~6MB, the BAM (Binary Alignment Map) ~6GB, and FASTQ (available in two parts ~3GB each for a total of ~6GB). The larger formats BAM and FASTQ are only available for download for a limited period of time (approx. 2 weeks) as they claim they don't have enough space to store such large files indefinitely. I downloaded all formats (after some minor trouble).

 

They also provide conveniently a file format compatible with Promethease. 

 

I immediately uploaded that file to Promethease and obtained a report ($5), and *separately* uploaded my 23andme file for another report ($5). I thus had two separate reports done during the same time frame, thus drawing upon the same SNPedia database. 

 

I then compared the two reports side by side. The results were not what I expected. Keep in mind, that the 23andme chip used to sequence my sample was an ancient one (I subscribed to 23andme the moment they went live), several years old, and since then there's been tremendous technological progress in sequencing, therefore I anticipated it to be a lopsided contest in favor of Genos. After all, 23andme report yields ~24,000 genotypes and ~2000 annotated in ClinVar, whereas for Genos the numbers are ~42,000 and ~34,000 ClinVar. As said, Genos sequences substantially your entire exome. 

 

Yet, the side by side reports were a real eye opener. I should have known something was odd when the report file downloaded from Promethease for 23andme was larger than the one for Genos. In any case, a side by side comparison showed the 23andme report to be vastly superior. It had a larger share of higher magnitude rsID results by far (magnitude is a measure of verified results based on multiple studies, and therefore the confidence levels are higher for higher rated magnitude). So, by the numbers:

 

Magnitude 5 - 23andme: 1 ; Genos: 0

Magnitude 4 - 23andme: 2 ; Genos: 2

Magnitude 3.5 - 23andme: 1 ; Genos: 1

Magnitude 3.2 - 23andme: 1 ; Genos: 0

Magnitude 3 - 23andme: 17 ; Genos: 9

 

The magnitudes below 3 are of less interest as they are too uncertain, speculative, unverified and marginal-to-date, but in any case, there were still more of them from 23andme vs Genos (reflecting the larger size of the 23andme report).

 

There was obviously some overlap between the two reports, but what I found interesting, is that of the high magnitude results, (up to and including magnitude 3), almost all of the results from Genos (except for one) were also present in the 23andme report, whereas the 23andme report had twice the number of results compared to Genos. In other words, the Genos results are a subset of the 23results - and not a very large subset (less than 50%).

 

Bottom line: if you have your 23andme data report from Promethease, there is virtually no added value in getting a Promethease report based on the Genos data (as provided by Genos) - and conversely, if you were to pick one report, you would absolutely want to get the 23andme report, as it's over twice as fruitful (as measured by the number of 3 and above magnitude results).

 

Then we get to the respective websites. Hands down the 23andme website provides much superior user functionality. First and foremost, 23andme provides a raw data browsing functionality - input any rsID and you get your own variant back. This is extremely important. By contrast, Genos provides NO such functionality, instead, the only rsID you can search for is the tiny subset of ClinVar results that Genos deems of interest (effectively the same dataset as in the inferior report from Promethease based on the Genos file). In other words, that aspect of Genos is a major FAIL.

 

It becomes a fair question to ask: if the Promethease report from Genos is inferior to the 23andme, and there is no option of browsing raw data from Genos as there is from 23andme, what does Genos bring to the table? The answer is: nothing really, except for the option of your being able to download your entire exome file (which obviosly does not obtain for 23andme).

 

Separately, 23andme can draw upon its extensive customer database to generate reports which they then alert their users to - as they can data crunch their customer data and derive interesting results from this. Obviously, as Genos is a young company just starting out, with a relatively tiny customer base, they cannot match such functionality.

 

To sum up: at the present time - March 2017 - 23andme is a vastly superior value proposition compared to Genos, no ifs buts or maybes. And that's quite apart from the $100-$150 cost of 23andme sequencing vs the much more comprehensive and superior sequencing from Genos at $400-$500. 

 

Yet, this is not the entire story. As mentioned Genos sequences substantially your entire exome. This means, that while today belongs to 23andme, the future belongs to Genos, providing Genos plays their cards right. While you will be able to continue to get new results from your 23andme sequencing in the future (as SNPedia continues rapidly expanding), they will not be able to match the vastly greater number of potential results from Genos, as Genos can draw upon your entire exome.

 

In order to pull ahead, Genos needs to do two things, urgently. First and foremost, they *must* provide the capability for customers to browse the raw data in its *entirely* - without this functionality, they are dead in the water as far as a value proposition. Second, they must do a much, *much* better job of evaluating what variants are of interest so that when one browses their chosen results or generates a report through Promethease (and SNPedia), they are competitive and preferably superior to what you can get from the ancient 23andme chips - that's the least they can do.

 

One note - as I had some questions about their website, I contacted Genos, and they reached out to me and we had a very good exchange of information. They are quite responsive and interested in the needs of their customers. That is a big plus, as you are not likely to get such a hands-on direct interaction from a place like 23andme.

 

Finally, at the end of the day, for $400-$500 you can download your entire sequenced exome. That is a tremendous value, unlikely to be matched by 23andme anytime soon.

 

I remain optimistic about Genos potential, and I think they are a viable option for those who are interested in more than the limited data available through 23andme as we our knowledge base continues to expand. As far as I can see, the $400 represents a good investment for the future. 

[Mechanism]

Edited

[TomBAvoider]

Right, I got an email from Genos yesterday, announcing this acquisition. Hard to tell what this means for existing clients of Genos - for now they have access to their data and can of course download it, but Genos is not taking on new clients, and in fact if one has sent in a sample recently, they'll refund your purchase. What this means is they are no longer in the retail sequencing business. I suspect that longer term their integration with NantOmics means they're out of that space and from now on will only be one of those companies from which doctors order specific tests, which means they are of no use to anyone on this board. Oh well, that was certainly a short ride. Ultimately what I got for my $500 is my exome data, which is pretty basic, but hey, I've spent more than that on supplements which later transpired to be useless or worse. Win some, lose some.

[BrianMDelaney]

 

Bottom line: if you have your 23andme data report from Promethease, there is virtually no added value in getting a Promethease report based on the Genos data (as provided by Genos) [...]

 

Tom, Thanks for the extremely useful summary!

 

But I'm puzzled. An exome scan should have contained a gargantuan number of SNP calls – far, far more than the 500–900K offered by 23andMe.

 

Perhaps "as provided by Genos" is key: is it possible that they do not modify all the sequenced data into a Promethease-friendly format, but rather curate it in some way, selecting the SNPs that they think would be interesting?

 

I thought virtually all the SNPs tested by 23andMe were exomic, which would necessarily make it a subset of the results of a Genos scan. Not so?

 

Brian

[TomBAvoider]

Yes, the Genos sequencing resolves far more genotypes and ClinVar hits. The only reason I can identify as to why the Genos as filtered through Promethease data is so unimpressive is that the file Genos provides Promethease is something that they curate, and it is in fact tiny (compared to 23andme). When I spoke to Genos, I highlighted this issue, and they asked me to send them the respective Promethease files, which I did. I also went through the side by side comparison we communicated through gotomeeting.com which has the capability to display desktops, and that's how I was able to show the reports side by side. After our conversation, I sent them the files. They got back to me just a few days ago, and did not provide a satisfactory explanation for this discrepancy, merely provided me with a powerpoint graph showing how you get more Maximum Magnitude results over time from Genos compared to 23andme. As to why that was not the case for me, they had no explanation other than, and I quote "About Magnitude, it's a personalized score. Some people will have more, some people will have few." - which is nonsensical in this context, because we are comparing *the same* individual's (mine) data through two different services, so this has nothing to do with me having possibly more/less MM results compared to Bob ShitTheRagman. I give this aspect of their service a grade of F. It's all academic now, however, since they've been taken over and provide no sequencing service to the public anymore. It is however a good reminder that you shouldn't just rely on one company for your sequencing needs. Had I only had access to Genos, I'd think "wow, I got my whole exome here, it' got to give me more info than 23andme, so glad I went with Genos" - and been 100% wrong. It's only because I had access to both that I was able to see that the Genos data is severely lacking (as curated by Genos and filtered through Promethease). For that matter, while I do have my complete exome data file downloaded from Genos, I don't have a similar *complete* exome data downloaded from another service (23andme is just a small subset) - how do I know that the Genos complete exome is in fact as complete as they claim..? I have nothing to compare it to, so the fact is: I don't know. I'm assuming it's complete, but hey, after this experience of Promethease comparison, I'm somewhat cautious with any definitive statements or certitudes. 

 

Bottom line: getting your data from only ONE source or one company is probably suboptimal, and you're definitely taking a gamble. Caveat Emptor. 

[BrianMDelaney]

To be honest I'm not focusing on the number of high magnitude results, since that isn't the source of my puzzlement. The puzzle is: 23andMe sequences a small subset of the exome. The full raw data from Genos should yield Promethease results with more SNPs. (Most will have a low assigned magnitude, but that doesn't matter.)

 

There are only two explanations: 23andMe sequences lots of non-exomic SNPs that also happen to be in SNPedia (the database used for a Promethease report), or the Promthease-friendly version of the Genos file is stripped of lots of SNPs.

 

Explain that to Genos, as see what they say. I would love to know.

 

I've emailed the SNPedia people, but they usually ignore my emails, since I've sent so many....

 

Brian

[TomBAvoider]

Well, I can ask, but can't say if they'll answer - they're in the middle of being acquired, and the business model is changing completely... old customer retention is not likely to be a high priority. I suspect customer relations will be the first to go. But whatevs, I'll send in a query, see what happens.

[Sthira]

...MM results compared to Bob ShitTheRagman....

I can't add much to Maximum Magnitudes; but I am concerned about Bob ShitTheRagman.

 

When I search, I find Joe Shit the Ragman:

 

"noun phrase (Variations: Snuffy or Tentpeg may replace Shit the Ragman) An ordinary soldier; buck private; gi (1970s+ Army)"

 

Maybe Bob ShitTheRagman is a homeless war veteran?

[BrianMDelaney]

Tom, thanks. Would be useful to know.

 

I suspect it would not be hard to take the full dataset from Genos and automatically modify it to be in the Promethease format.

 

Brian

[TomBAvoider]

Ha, ha, Shtira, a co-worker from back in the early 80's would refer to randos as Joe Shit the Ragman. He was a Vietnam vet. I never knew the provenance of this expression, thanks for looking it up :)

 

Brian, I'll shoot 'em an email tomorrow night. As I mentioned in another post, the data from Genos is formatted as VCF, BAM, and FASTQ. Honestly, I have not looked into what it would take to format any of those files for Promethease, but what I would be even more interested in, would be to turn the info from one of those files into a searchable database that I can query for rsids. However, I'm not a database software maven, so I have not looked into it. I might ask around, see if anyone has any ideas. Maybe at the very worst, it could be in some giant pile where I could look for SNPs with a simple "find" search function like in a google or msft word doc format. That would be super clumsy, so maybe there is some procedure to have a database automatically parse and ingest one of those file formats, because it's obviously not practical to input by hand.

[Mechanism]

Edited