genomics-based match-making

[gusl]

At SciBarCamp, I heard some buzz about an online dating service that matches people based on their DNA.

What are your thoughts on this?

ScientificMatch offers such a service based on immune markers. They charge $1000, which probably means a very small dating pool.

I would want any such site to be interoperable with OkCupid, which seems to be quite good at matching people based on personality (despite recent "match inflation": how come I have so many 95% matches all of a sudden??)

Also, 23AndMe's full package costs $399. I wonder if their SNPs correspond to immune-markers... or if anyone is mining their data, looking for patterns in couples who have been sequenced (of course, one needs to account for known patterns of assortative mating)... in which case, you could tap into this larger dating pool (and pay smaller price, and get millions of free SNPs).

As a scientist, I want my matches to empirically based, and to report my experiences as further data for their model.

Comments

[afanesvoltaje.livejournal.com]

my understanding was it's hla's

[gustavolacerda.livejournal.com]

thanks!

[glapaloopscap.livejournal.com]

Sounds like a very expensive scam. How do they know what makes a good couple?

[gustavolacerda.livejournal.com]

Several studies show positive indicators about couples whose HLAs complement each other.

ScientificMatch cites sources.

[gustavolacerda.livejournal.com]

Wikipedia talks about it too

[glapaloopscap.livejournal.com]

Sure, I've read about that too, but has anyone ever measured how much difference that makes to a relationship compared to, say, having similar musical taste or political affiliation? I've had "sexual spark" for plenty of people who I seriously doubt would have made good mates for me.

[gfish.livejournal.com]

Sounds like a machine learning business model. Get couples/romantic units to submit the DNA sequences along with a relationship report.

[rdore.livejournal.com]

I would be concerned that ScientificMatch simply does not have enough users:

http://siteanalytics.compete.com/scientificmatch.com/

[gustavolacerda.livejournal.com]

How can that analytics site count hits to ScientificMatch? The only way I know would involve hits to SM always sending a request for a file on the analytics site (usually a 1x1 image).

... or are these numbers estimates, based on IP requests?

[rdore.livejournal.com]

I don't know. But other sites that do similar kinds of rankings (Alexa and Quantcast) didn't even feel they had enough data to give info about ScientificMatch.

[afanesvoltaje.livejournal.com]

The frustrating thing about this company is they're stealing my backup scam. I've been telling myself for years now that if I get tired of working my real job, I can always switch to a scam where I tell people I have an MD and a PhD in "phereomone studies" (true enough), and that I'm going to retrain their olfactory systems to help them fall in love with people who will love them back (and I'll offer to make them skinny while I'm at it...). Then I'll provide some surprisingly efficacious combo of normal interpersonal skills counseling + placebo effect (eg, if people *expect* to now be interested in different people, they're more likely to actually do it), and they'll be happy and I'll be rich.

Without working so many #(*%# hours as I am now.

[flamingnerd.livejournal.com]

my thoughts are that both are not worth it at all. genetics just ain't there yet. we can sequence a bunch of shit, but the meanings of differences are still not usually obvious.

[gustavolacerda.livejournal.com]

Tangentially, I've been told to hold off a couple of years, until I can sequence my entire genome for cheap.

[the-locster.livejournal.com]

I worked on the netflix prize for some time and I strongly suspect that the user vectors from matrix factorisation of movie ratings would be a very good means of quantifying compatibility between people. I'm not sure what the best distance metric is - Manhattan(L1), Euclidean(L2) or some other metric, but that's an easy bit of experimenting to do.

Films cover such a wide array of subjects, feelings, belief systems, types of humour, etc. that I'm not sure any personality questionaire could ever represent the nuances that can be distilled from a matrix factorisation of 100 or so movie ratings. Unfortunately it only works for people who like films ;) but that is at least the majority of people.

[marymcglo.livejournal.com]

Didn't the Nazis have something like that?



...The thread had 13 comments already-- I had to say it.