281–290 of 458 posts
Earlier quoted context omitted.
I seem to remember reading that even when they delete your data, they keep your sample. So really all it does is prevent you from getting it.
This is not correct and in fact is the exact opposite. By default, they destroy the sample after analyzing it.
Earlier quoted context omitted.
> Your exome changes all the time. I work in DNA analysis (not for 23andme). Citation needed. > Your gut biome changes all the time. Your gut biome isn't being tested by a genetics company collecting your saliva or epithelial cells.
How complete are theses tests? If I wait another 10, 15 years to get sequenced, would I get more accurate, reliable information? Can advances in sequencing technology make it worth people to get resampled? Or is it just an issue of better analysis?
I can't answer that.
> If I wait another 10, 15 years to get sequenced, would I get more accurate, reliable information?
If you're asking about the DNA data itself? Yes, absolutely. The underlying sequencing technology is advancing every year.
If you're asking about the value-added informational inferences sold to consumers? I wouldn't put much faith in that, but I'm also a cynic when it comes to computers and business relationships with consumers.
@mathnerd314's comment [0] is correct, current DNA analysis products frequently produce conflicting results [1]. I've seen the software side for why that can happen. The technology is extremely complicated and sometimes convoluted. And there are many "standards" (xkcd-style [2]) for reporting even more ad-hoc features. Big companies can't even get basic web services to work right 100% of the time, what makes you think DNA analysis software can do that?
There's a lot of math which goes into consumer DNA products. The math involves a lot of statistics and population information. The math can be biased towards or against particular people. And it's usually hidden behind closed-source algorithms and proprietary information. Do you trust that?
I strongly think this is an area where international standards organizations should come into play and where laws should be put in place to govern data correctness, responsibility, and culpability. The NIH is doing some work towards that end (eg, precisionFDA challenges) but it's doing a lot more science-related work instead of consumer-related work IMO. As-is, consumer sequencing analysis products (especially in the United States) are in the wild west.
Without strong standards and laws I don't think you're guaranteed to get more accurate or reliable information in 10 to 15 years. You'll get the same kind of information you can get today, but perhaps cheaper.
> Can advances in sequencing technology make it worth people to get resampled?
Yes. The sequencing products that I have worked on have gone through multiple generations during my (so far) six years of employment in the industry. Each product's generation changes the sequencing technology itself, the analysis product, or sometimes even both.
> Or is it just an issue of better analysis?
Current technology has a lot of trouble sequencing completely through a full strand of DNA. If you can solve that then analysis will instantly get better. To get around that, there are many different types of sequencing available.
The micro-array that's used by most consumer companies focuses on hundreds of thousands of single-basepair locations and do so quite affordably at industrial scale. If you have any sort of unique variation at one of the targeted locations then micro-array sequencing technology will likely have trouble getting data about that location. But en mass, one data point isn't enough to throw off your results.
Another type of sequencing works with chunks of DNA - partial strands - but sequencing quality goes down with the length of the chunk. One chromosome will have millions of basepairs split into thousands of chunks but the maximum length of the chunks for usable data is measured in hundreds or maybe thousands of basepairs. To help solve that, the sequencing technology will make many hundreds of duplicates of those chunks and sequence all of them simultaneously in a massive parallel operation. Even though the sequencing itself is slow, the parallel part makes it high throughput. Then the low quality of each individual read is offset by high numbers of copies and reassembly is just a (very) computationally expensive task.
If you can improve the quality delta across a DNA fragment, particularly to the millions of basepair lengths, then that would be a game-changing breakthrough, especially if you can keep throughput high. I intuitively know it's possible: your own cells copy your DNA every time cell division occurs, and they do it extremely fast compared to current sequencing technology. I'm just a software dude though.
There are other sequencing technologies which I've not worked on and can't comment about.
[0] https://news.ycombinator.com/item?id=22130359
[1] https://www.healthista.com/dna-genetic-test-review-three-dif...
It may be that there aren’t any new products that are either useful or that people want. But still, it highlights the type of company they want to be, a kind of genomics facebook.
> Wojcicki has theories, but she doesn’t have clear proof for why consumers are shying away from getting tests Because the ones that want them have gotten them and there's no reason to ever buy the product again once you've done it once. I mean, smartphones are great, but the market for them would have hit a wall pretty quickly if no one ever had a reason to replace them after buying their first.
Earlier quoted context omitted.
If the government encounters one of their kid's DNA, that's enough to identify them as a son/daughter of the person with that DNA profile. If a medical insurance company purchases their DNA, analyzes it, and anticipates that a child of theirs is likely to have medical costs 5x the average, those costs will be passed on to those kids, absent legislation to the contrary. This applies to everyone you're related to, as w…
> If the government encounters one of their kid's DNA, that's enough to identify them as a son/daughter of the person with that DNA profile. I don't believe that's ever been tested in court (so there isn't precedence) and the ways in which DNA are complex enough - and the mutation rate is high enough - that it might be like fingerprints... if you're the only person with DNA that might be like the kind found and you'v…
Earlier quoted context omitted.
Honestly, I would also say that a partial reason why these would've dropped is that everyone who can do it has done it. You only need these kinds of results once because your DNA doesn't change. I knew people who were giving or had gotten these as gifts, but those are one-off purchases. And if you've heard of people doing it then you'll do it too but that's also a one-off. I don't think I'd personally benefit because…
I did it 6/7 years ago because the novelty, and enthusiastically recommended it to family and friends. Now I regret doing it considering how Google and Facebook has abused consumer data. I definitely would not submit my kids to the same fate.
Earlier quoted context omitted.
If the government encounters one of their kid's DNA, that's enough to identify them as a son/daughter of the person with that DNA profile. If a medical insurance company purchases their DNA, analyzes it, and anticipates that a child of theirs is likely to have medical costs 5x the average, those costs will be passed on to those kids, absent legislation to the contrary. This applies to everyone you're related to, as w…
> If the government encounters one of their kid's DNA, that's enough to identify them as a son/daughter of the person with that DNA profile. I don't believe that's ever been tested in court (so there isn't precedence) and the ways in which DNA are complex enough - and the mutation rate is high enough - that it might be like fingerprints... if you're the only person with DNA that might be like the kind found and you'v…
They sold 10M kits, and they primarily market in the US. DNA never changes, so you have a lifetime max of 1 purchase of 23andMe per lifetime. Their competitors have likely sold 15-30M as well, so maybe 30-40M of these kits have been sold. That's 10% of US market. What's the max conceivable penetration here? 20%?
It occurs to me that they could have followed the software trend where everything is a subscription service. As more is learned from the genome, they can deliver more value to the customers over time.
Lots of layoffs happening recently: 23andme, stubhub, tripadvisor etc. Early indicators of a tech bubble?