I'm distressed by the attitude towards 23andMe's research - isn't good science supposed to report all results, regardless of whether they support the desired hypothesis? Isn't it useful to discover the negatives as well as the positives?
I've been a customer of 23andme for 3 years, and they've always been very forthright about providing guidance on what you really could discern from the results they provide, including a measure of the strength of the scientific evidence that provided probabilities of you getting this disease or that one.
Personally, the highlights for me where when my Brother, Mother, Niece and Nephew tossed their results into 23andme, and it instantly identified them as relatives, and what their relationship to me was - without any last names or locations to go on. (Though admittedly, my niece and nephew were identified as "Niece/Aunt, Nephew/Uncle)
I've been happy with the sober, reasonable, and conservative approach that 23andme has taken. "Genewatch" on the other hand, sounds like they have an axe to grind - they certainly lost credibility with me based on the comments I read. Not sure what their issue is...
I think the issue is more that according to their research, the service can't use its genetic testing to give you a heads-up about upcoming disease in your physiology, which is one of the big sells of 23andMe.
Of course, this is one study and one disease, and maybe other stuff will come out that validates the approach. (I have no dog in this fight)
But you point out an interesting fact: perhaps the biggest value they're providing is simple introspection, using tools we didn't have in the past. For me, understanding how my ancestry fits into the world and its history would be more than enough incentive to use 23andMe, depending on how much detail the science and the data set can offer.
"the service can't use its genetic testing to give you a heads-up about upcoming disease in your physiology, which is one of the big sells of 23andMe"
Nonsense - I have a common variant of one blood clotting factor known to be responsible for a greatly increased risk of deep vein thrombosis and life threatening pulmonary embolisms. My mother has the same autosomal dominant gene.
She was hospitalized for DVT and PE following a knee surgery, was genetically tested and found positive.
23andme identified me as being a carrier of said mutation and followup genetic sequencing confirmed the diagnosis.
Now I know, with certainty, that I must periodically get up and move around during long car rides and plain flights or risk a life threatening event.
Health-related discoveries aside, there are countless things to be analyzed from 23andme data, including ancestry and data mining for correlation with biometrics (imagine being able to reconstruct facial images based on dna samples alone).
There are always hypercritical crankpots to speak out against any new technology. If only these people would imagine constructive ways to make use of this so-called "useless" data.
I was with you until your last paragraph. You're not a "crankpot" if you're pushing for rigorous scientific research and transparency. Yes, there are currently identifiable alleles that indicate specific diseases or predispositions, but there are a whole multitude of factors beyond the basic, linear assembly of nucleotides that contribute to the development of a disease – epigenetics being a big one at the moment.
The tech is definitely cool and a step forward in delivering a detailed and granular view of disease predisposition, but that view certainly does not give you the entire picture.
You're right of course. There are certainly other factors, and scientific rigor is a Good Thing (tm).
However, I've found that the older the scientist, the more his or her disposition to say "no" to new ideas - either because the idea is untested or because it flies in the face of their already preconceived biases they've been amassing over their lifetime.
It's often said that science advances funeral by funeral. I think it's more of an indictment about human nature, and the nature of aging, than any particular individual.
When I was an undergraduate, I mentioned 23andme in class - the class had 3 professors. The oldest professor was completely against it: "Why would anyone want their SNPs mapped? It would be useless, at best, and potentially dangerous." The youngest professor was enthusiastic and optimistic about the future: "How much and where do I sign up?", and the middle aged professor was somewhere in between: "The data probably isn't very useful right now, and I think I'll wait until it comes down in price."
Of course, without people pushing the boundaries, and amassing the data, there would be no discoveries. There is nothing wrong with scientific rigor, but I come down hard against those who allow rigor to get in the way of discovery. Sometimes anecdotes are the basis of major discoveries - not all studies need to start out double-blind with huge sample sizes, but the important ones should definitely end up that way.
The tech is definitely cool and a step forward in delivering a detailed and granular view of disease predisposition, but that view certainly does not give you the entire picture.
23andMe makes a very good job of notifying you about this per disease they have markers for.
"I'm distressed by the attitude towards 23andMe's research - isn't good science supposed to report all results, regardless of whether they support the desired hypothesis? Isn't it useful to discover the negatives as well as the positives?"
Where does the article say that 23andMe shouldn't have published their results? That has nothing to do with what this is about.
I'm now more inclined to buy from 23andMe than I was before because I know they are doing legitimate science and reporting the results without prejudice.
Questions:
[1] Wouldn't it be worth funding this company just to find this out about Parkinson's disease?
[2] Is this a reason for them to pivot or adjust their marketing to talk about what their service does do well?
I've always been impressed with how honest their data seems to be on diseases. They made it very clear up front that it's all trial-and-error and they're doing the best they can with the data they have, and that advances are made all the time.
I doubt I would ever have paid full price for the service, but I got in during a sale and have been quite pleased for the money.
Their business model for folks who got in on the $99 deal now seems be "Want to know your Alzheimer's risk? Pay more to get your DNA resequenced with our latest tools!"
Citation? I got in on the $99 + $5/mo deal at Christmas, and I have not once received an upsell request. I have, however, received updated results every month since then. Now, either they're holding results back in order to keep me on the hook, or they're actually running new tests as new papers are published.
I just checked a few of my recently-updated results, and all of them had new papers cited with recent publish dates. I spot-checked a few upstream from the most recent update, and it seems that half of the updated results had papers published in the last three months, and half of those were published in June.
I don't know where you got your information, but I've been quite happy with 23andme.
> Their business model for folks who got in on the $99 deal now seems be "Want to know your Alzheimer's risk? Pay more to get your DNA resequenced with our latest tools!"
> How much is more? Another $299.
Incorrect on all counts. Current cost of the service is $99 + $9/month subscription which you may cancel. You also have the option to "bank" your saliva/DNA so that when they come up with a new gene chip, you will have results from that without having to repay. You just have to keep up your subscription, obviously. Since I got my results back I think they've been pushing updates about once a month (probably not a coincidence, but I don't feel like checking the exact dates), both times having information on genes covering 5-10 new topics.
What you describe did exist back when they switched from the v1 to the v2 gene chip (and I think the price was $399 at the time, no subscription existed), but now (v3) they have figured things out. I got in on "DNA day" this past April and paid $0 + subscription (so $100 for a year) and I chose the option to bank it, etc.
> [1] Wouldn't it be worth funding this company just to find this out about Parkinson's disease?
Only if they plan to branch out from genotyping into other areas of science, as the article describes how their research shows the limits of this approach in Parkinson's disease.
Maybe they only do the genotyping and other people can handle other aspects of disease prediction and prevention. Knowing something doesn't predict something else is just as good as knowing it does. In either case you need more information, but you've learned something.
It should be noted that every paper 23andMe has curated to provide their risk calculations was originally published in a peer reviewed journal, often Science or Nature. It should also be emphasized that new methods (like the recent Visscher et al. manuscript) are starting to give considerable signal for predicting complex traits from genome sequence.
Take Genewatch with a grain of salt, they are activists rather than scientists and are infamous for trashing anything associated with genetics and genomics (whether GMO crops or complex trait research).
I've shared the link you've kindly shared with the scientist who posted the link I posted here to an email list including participants in a behavioral genetics seminar.
I suppose that part of what goes on in discussions like this is sorting out which findings are "significant" not just in the statistical sense but also in actual clinical practice.
is temporarily unavailable. (I have just emailed him about that.) That is usually a good source of links to current literature on behavioral genetics, the subject that prompts my interest in genetics literature.
Just as relevant, to me: they were honest about their findings. Compare this to all of the highly flawed studies that some companies have pushed into journals in the past two decades.
Logging in to see my results for Parkinson's, they show a very low confidence interval - 3 grey stars. That seems very honest to me.
Instead, I'm glad they could use all this data to advance the state of the art, and will more likely trust them if they report on some diseases that are more linked to genetics.
Good article, sensationalist headline. They demonstrated that existing genetic predictors of a single disease were not much better than random selection. 23andMe's business model does not solely rely on the prediction of Parksinson's Disease risk.
This paper should be published in the Journal In Support Of The Null Hypothesis. It's just a valid a conclusion and good job by 23andme. I am a customer and I will continue to be a customer of theirs for the foreseeable future.
> 23andMe’s paper includes a new estimate that the heritability of Parkinson’s Disease is 23% (suggesting that 77% of the variance will never be explained by genes, but that missing genes yet to be discovered account for a further 16%).
That's not correct. There is more to heritability than just genes, so it is not at all guaranteed that the extra 16% can be accounted for by sequencing the rest of the genome.
Also, the headline is rather sensationalist as none of this is particularly surprising. GWAS studies have produced underwhelming results for many years.
Haven't read the paper, yet, but if they've really established heritability of Parkinson's risk, that is a positive result in itself. The first hit for a Parkinson's Disease twin study suggests that genetic factors play a limited role in the disease's etiology, at least in late onset:
RESULTS: Of 268 twins with suspected parkinsonism and 250 presumed
unaffected twin brothers, 193 twins with PD were identified
(concordance-adjusted prevalence, 8.67/1000). In 71 MZ and 90 DZ pairs
with complete diagnoses, pairwise concordance was similar (0.129
overall, 0.155 MZ, 0.111 DZ; relative risk, 1.39; 95% confidence
interval, 0.63-3.1). In 16 pairs with diagnosis at or before age 50
years in at least 1 twin, MZ concordance was 1.0 (4 pairs), and DZ was
0.167 (relative risk, 6.0; 95% confidence interval, 1.69-21.26).
CONCLUSIONS: The similarity in concordance overall indicates that
genetic factors do not play a major role in causing typical PD. No
genetic component is evident when the disease begins after age 50
years. However, genetic factors appear to be important when disease
begins at or before age 50 years.
I've done 23andMe. I would never rely on its results for disease prediction. A lot of their results read like a fortune cookie anyway. But it was interesting to genetically verify my Ashkenazi Jewish heritage with a maternal haplotype that is traceable to a group from the middle east around the beginnings of Judeo-Christianity. This was awesome, considering I could only trace my roots back about four generations prior to 23andMe.
You shouldn't rely on the results, certainly, but I don't think they should be taken with a grain of salt. I would suggest the problem with most people (and I'm not claiming to know your background) is that they aren't familiar with the statistics used in reporting the risks. For 23andMe, these are basically just relative risk and absolute risk. But when you look at relative risk (I'm five times as likely to develop Type 1 diabetes), it's much more upsetting than the absolute risk (my risk is 5% over my lifetime instead of the usual 1%). I do think they try and mitigate it with their population pictures (100 figures and X many are colored in to show your likelihood) but I think that at the heart of it their user base (and the population at large) just doesn't really understand statistics well.
I've been a customer of 23andme for 3 years, and they've always been very forthright about providing guidance on what you really could discern from the results they provide, including a measure of the strength of the scientific evidence that provided probabilities of you getting this disease or that one.
Personally, the highlights for me where when my Brother, Mother, Niece and Nephew tossed their results into 23andme, and it instantly identified them as relatives, and what their relationship to me was - without any last names or locations to go on. (Though admittedly, my niece and nephew were identified as "Niece/Aunt, Nephew/Uncle)
I've been happy with the sober, reasonable, and conservative approach that 23andme has taken. "Genewatch" on the other hand, sounds like they have an axe to grind - they certainly lost credibility with me based on the comments I read. Not sure what their issue is...