gattaca- waiting for my test results

[mellowtigger]

Saturday morning, I mailed off a tube of my spit.  It will take about 8 weeks to get back my results.  When the results are posted online, I will be able to view a vast number of specific points of my DNA to see how they compare against known differences in the human genome.  I'll be using the tag "gattaca" here on LiveJournal to follow my results of this personal journey.  I suspect that I will be learning a great deal about the human genome starting about 8 weeks from now.  *laugh*

"Gattaca", for those who don't know, was a great movie about the dangers of predicting human life outcomes based on the results of genetic testing.  The name of the movie is built using the same letters as the 4 chemical parts used to build a dna molecule: guanine, adenine, thymine, cytosine.

I first learned about 23andme.com a few months ago.  (Hello, lordalfredhenry .)  I did not feel any overwhelming urge to run my own DNA through their system, but I confess that pure unadulterated curiosity finally got the better of me.  I added $400 to my credit card debt, and now here I am waiting for results.  I've already created my community profile there, although I think you have to have paid your money and claimed your special test kit code in order to participate.
https://www.23andme.com/user/?community_profile=92caa46f527ff773

Using their sample result data set as an example, I will get about 633,000 data points from my DNA.  These data points are called SNPs ("snips"), Single Nucleotide Polymorphisms.  These are places in the long strand of DNA code that a single difference is found.  So instead of getting my complete genetic code back, I'll be getting the "deltas", the differences.  They are using as a reference template, the published "human genome, build 36".
http://www.ncbi.nlm.nih.gov/projects/mapview/map_search.cgi?taxid=9606&build=36

8 weeks until I see results.   That's 56 days.  Or 1,344 hours.  Or even 80,640 minutes.

*twiddle*

Comments

[lordalfredhenry.livejournal.com]

LOL. I posted accidentally to the thread below this one. That's cool! I've just shared. I'm curious to compare results with you.

[mellowtigger]

Me too. I've marked to share also. (I think. Hard to tell since I don't have data yet.)

No worries. I'll go delete the other post to avoid confusing anyone. :)

[hilltop.livejournal.com]

Very cool, it should be interesting to see what comes out of this.
And, on a related note, this is something I'd love to do.
http://www.yourdnasong.com/

[mellowtigger]

Hrmmm. Interesting concept, but I have my doubts about the implementation.

DNA has waaaaaaay too many data points to turn it into a song, and those data points have only 4 possible values, so that'd be a really boring song anyway. I don't see any obvious way to convert any portion of DNA into music. I think he's taking "artistic license" to an extreme. :(

Still, neat concept. :)

[mellowtigger]

*more pondering*

I suppose it's 8 data values, if you take methylation (epigenetics) into account. Still not enough for an octave based on 12ths though.

*still more pondering*

Or... instead of base pairs, suppose we took them in triplets, the way nature does. 4³=64, so 4 base pairs in triplets make 64 codons. That works, plenty of room for notes that way. Instead of using the 4 base pairs to make notes, use the amino acids (triplets).
http://en.wikipedia.org/wiki/Genetic_code

[hilltop.livejournal.com]

I've exchanged quite a few emails back and forth with him on this, and there is a bit of leeway, but for the most part, he uses the data as it comes out. It's just a mighty cool idea.

[snousle.livejournal.com]

Several of my colleagues from Perlegen were "poached" by 23andme and now work in their bioinformatics group. We had done quite a bit of work on inferring ancestry from high-density maps, so I'm intimately familiar with the statistics involved, and presumably they are using a very similar technique in their analysis. I'm not sure exactly what they're providing to customers, but in principle you can get QUITE a lot of interesting ancestral information from that number of markers. Useful health information, not so much - the risks attributable to genetic causes aren't usually very meaningful when it comes to medical choices for otherwise healthy individuals.

I know a guy here in Ukiah that knows nothing of his ethnic origins, and when I told him about this company he was keenly interested. I'm sure he'll want to hear about how it works for you, since you're the first person I know who has used their service.

[mellowtigger]

I've talked to one adopted person who has been denied information about his biological father (from people who did divulge information about his biological mother). He's very interested in a statistical report too.

I'm more interested in doing research than making predictions. I want to gather a large group of autistics, classify ourselves into different groups (much more specific than current psychiatry definitions) and see if any patterns emerge from our DNA about ourselves.

There are already public databases sharing SNP information:
http://www.snpedia.com/index.php/SNPedia

Admittedly, my theory is that epigenetics will have more to do with explaining autism than genetics. 23andme does not test for that dna information, unfortunately. There are already public databases for that information too, which we've discussed before:
http://mellowtigger.livejournal.com/30858.html

[snousle.livejournal.com]

I'm more interested in doing research than making predictions. I want to gather a large group of autistics, classify ourselves into different groups (much more specific than current psychiatry definitions) and see if any patterns emerge from our DNA about ourselves.

Unfortunately it is extremely difficult to perform case/control studies of this type. The main reason is not the availability of the genome scanning technology, it is subject recruitment. I'm not sure exactly what restrictions (if any) would be placed on a self-funded individual or group performing such research, but even without the incredible bureaucratic hurdles placed on, say, university researchers, even a few simple PCR-based tests I arranged for some people interested in the CCR5 HIV-resistance gene convinced me that amateur genetics with human subjects is a BAD idea. Way too much potential for complication, particularly since it is very difficult to communicate the meaning of the results to people who don't already know a lot about genetics and probability.

Once you get the data the statistical analysis is also quite complicated, mostly because of the need to correct for the differences in ancestry that arise in all case/control groupings. Taking the data at face value, it's easy to find lots of "statistically significant" correlations between various markers and the phenotype of interest, the trouble is you find so many of them that you can tell right away that they aren't credible. Hence the need for mathematical back-flips to sift out any real associations that might exist. I personally developed a simple method of eliminating particular subjects from the study to achieve "balanced ancestry" in the case and control groups, but that was soon surpassed by some considerably smarter statisticians using much more sophisticated methods that could incorporate all the data. Unfortunately the details of that are now a bit beyond my expertise. ;-)

Nevertheless, one of our former collaborators has already performed a fairly extensive autism study and if you are interested the data are publicly available. I haven't looked at it directly and I don't know exactly what phenotype information is available, but here's a lead on it:

http://www.hopkinsmedicine.org/Press_releases/2007/10_22a_07.html

[mellowtigger]

Ooh, neat. I still worry, though, that these studies that talk about "autism" are missing important details. If they were more clear about "autisms", then I'd be more likely to trust their findings.

Some talk (CONSTANTly!), some never verbalize at all. Some appear to be ruled by rage, and some appear inert. There are lots of ways to categorize autistics, and those categories affect the kind of intervention/training that they could use. It seems to me that studies that rely on "autism" would be as misleading as a study that relies on "blindness". Different causes, different strategies, different genetic patterns to identify (I think).

I would be interested in learning if families with multiple members with autism diagnoses all display the same "kind" of autism. The only thing I've read along those lines is that a parent will be diagnosed with Asperger's after they learn that their child is diagnosed with Kanner's (classic) autism.