I've been wondering if any agency in the US government has been using DNA databases to do "genetic triangulation"--and if not, when they'll start.
I'm in 23andme, and I get messages saying "someone who is your 4th/5th/6th cousin wants to connect" all the time. I figure if I was given a bit of DNA from a crime scene, by cross-referencing all of the 4th,5th, and 6th cousins, the number of potential people matching that DNA has got to be tiny.
Science fiction dystopias used to hypothesize a complete DNA database but I'm pretty sure even the spotty coverage we have now is pretty powerful.
DNA storage is a very tricky area ethically. You're dead right that even spotty coverage is sufficient to make all sort of inference, while on the other hand we have much more DNA from crime scenes than we have the resources to process, and when we do process it it can lead to the exoneration of people who were serving long sentences.
Lacko f administrative will ont he part of police departments seems to be the largest part of the problem, if stories of unprocessed rape kits by the hundreds are anything to go by. Thankfully some states have committed to clearing the backlog and not allowing it to build up again.
Computational power probably isn't the limiting resource in DNA analysis (I've known plenty of people who do it in non-forensic applications in the biotech field, and I've heard of various things limiting throughput in any given effort, and computing resources were never on the list.)
ISIS has been using government biometric databases at checkpoints to ID regime agents and execute them on the spot. They had seized facial recognition databases that were largely put together by Americans. When you wind up in these biometric databases you really can't be sure who is going to use it for what. It's not like the data is ever secure.
In the months before Osama bin Laden was killed, the Central Intelligence Agency ran a phony vaccination program in Abbottabad, Pakistan, as a ruse to obtain DNA evidence from members of Bin Laden’s family thought to be holed up in an expansive compound there, according to an American official.
BTW, it appears the CIA has committed to not doing that sort of thing again, as a result of complaints from the public health community about this operation:
A newly disclosed CIA policy mandates that it won't use vaccination programs as part of its operations, according to the Obama administration.
The directive by CIA Director John Brennan, made nine months ago but only coming to light now, followed concerns raised by leaders of a dozen U.S. public health schools in a letter to President Barack Obama.
They spoke out following revelations the CIA had enlisted a doctor to oversee a false immunization campaign in Pakistan ahead of the 2011 raid that killed al Qaeda leader Osama bin Laden in Abbottabad.
So they say. The CIA are liars by trade (like the NSA) and I would give their pronouncements and oaths as much attention as I give their pathetic apologies for their past and ongoing inhumanity. They will break their "word" tomorrow if they feel it is in their best interests.
I'm an American and I'm repulsed by what is often done in my name by my "intelligence" agencies.
> BTW, it appears the CIA has committed to not doing that sort of thing again
I don't think the CIA could ever find another doctor to help it with a similar scheme. Not after what happened to Dr. Afridi. [1]
This doctor has been put thru a living hell. Prison in Pakistan is probably no Sunday picnic in the best of circumstances. But in this case there are allegations of torture involving cigarette burns and electric shocks.
So if the CIA ever approaches anyone who's ever heard of The Information Superhighway, aka The Internet, maybe that new guy will be smart enough to google for whatever happened to the guy who help us catch Bin Laden.
Also, if you believe the claims by Pakistan that the doctor is being prosecuted for crimes other than helping the USA, I'll leave you with how Seth Meyers put it on SNL Weekend Update after OJ Simpson was convicted of robbery in Las Vegas:
Today, a jury in Las Vegas found OJ Simpson
“guilty” on all seven counts against him,
including kidnapping, aggravated assault,
and theft. But really... Murder. [2]
What they have promised not to do is to masquerade as a vaccination program (thus harming the reputation of all who attempt to implement vaccination programs). They have NOT promised to cease using DNA surveillance of relatives to locate a person of interest.
> What they have promised not to do is to masquerade as a vaccination program (thus harming the reputation of all who attempt to implement vaccination programs).
The damage is done. What the CIA promises afterwards is not going to carry any weight where this matters.
> Genetic material obtained through a fake door-to-door hepatitis B vaccination programme reportedly helped the CIA confirm Bin Laden's whereabouts in the Pakistani city of Abbottabad.
There is a team in Denmark, working on building body and face models from persons genetic code. Stuff like age, height range, skin, hair, eye color, big/small/regular nose, wide/narrow face.
That’s probably not a very reliable proxy for age estimation though (for reasons alluded to by sampo, and simply because telomere length varies across individuals regardless of age). It’s probably much easier and more reliable to measure the accumulation of single-base mutations (SNPs): DNA repair is faulty and such mutations accumulate with age. By comparing the genetic material of several cells across a sample this can be estimated.
However, telomeres are repetitive code, and not amenable to standard PCR and sequencing methods (as far as I know). I don't know if the amounts of DNA or chromosomes typical in forensic samples are enough for the techniques used to measure telomere length. Well maybe they are, if the Danish group is working on this.
Well I think that if the length of the fragment is larger than the repeated fragment, you can get a good idea of the number of repetitions.
This is not a case like CNV (copy number variation) where a whole gene can appear a number of times, so I think this is somewhat possible by PCR (though probably not very precise).
The problem in finding repetitions using PCR is as follows: In order to perform PCR you splice the DNA in chunks randomly, the length of the chunks depends on the particular machine but it's of the order of 100-1000 bases. Then you need algorithmically find superpositions to reconstruct the original sequence.
If the repeated sequence is longer than the chunk you have no hope of knowing how many times it repeats, if on the other hand it's much shorter (I think it's 6 bases for telomeres) you can see a good number of repetitions per chunk.
If you want to know something more detailed about sequencing and CNV I can send you a couple of sections from my PhD dissertation.
Is it possible to do 23andMe pseudonymously? Mail them a money order and use a temp email address? I would never volunteer a dna sample like that clearly tied to my SSN.
I have always assumed that the airport scanners are also collecting biometrics on everyone for video identification purposes and for that reason I try to avoid them.
Funny comment. You want to stay anonymous and at the same time mail someone your hardest to fake piece of ID. There is something really wrong in your reasoning, your DNA is your identity and if any other family members of yours use 23andme as well then your relationship to them could very well be uncovered.
I believe he has some point. Humans generally have multiple (mostly highly-overlapping) [sub-]identities, one per clique. And we usually don't like mixing them too much. We're one person at work, but a bit different one at a close friends' party.
Our genetic data is a part of the picture but not a whole picture. That's somehow in a same manner your photo identifies you, but when you have a photo of random person on a street you're unlikely to easily figure out their name.
That is because a photo can not be linked to your family relationships as a DNA strand can.
So photos and DNA have very different properties when it comes to identity, and barring some stupendously unlikely coincidence (or an identical twin or clone) your DNA is much more unique than your photo will ever be (witness the number of celebrity doubles).
Indeed. However, I think, a wish to not provide a link between your DNA and your name is somehow understandable. Anonymizing a subset of social graph isn't something completely wrong, although, indeed, I agree, it's unlikely to work well, because, it's very likely that only a single node would go without a name.
(However, if all his near relatives would do the same, all one could figure out that John Doe is probably a relative of a Jane Doe. And they're somewhat related to other people with real names, but at least deanonymizing requires more work and other databases.)
A really old skit by a Dutch comedian went something like this:
In old days we had only 1 digit phone numbers in our village. 0 was the operator, 1 was Dr. Jones, 2 was the parish priest, 3 was the butcher, 5 was the greengrocer, 6 was the local contractor, 7 the constable, 8 mrs. Smith, 9 the Jacksons and we had a secret number.
I'm in 23andme, and I get messages saying "someone who is your 4th/5th/6th cousin wants to connect" all the time. I figure if I was given a bit of DNA from a crime scene, by cross-referencing all of the 4th,5th, and 6th cousins, the number of potential people matching that DNA has got to be tiny.
Science fiction dystopias used to hypothesize a complete DNA database but I'm pretty sure even the spotty coverage we have now is pretty powerful.