Sunday, July 20, 2008

Playing games with DNA

This is some complex stuff, DNA, and it gets even more complex when you try to bamboozle juries with what it means - actually, maybe it makes it less complex, but unraveling the bamboozlement gets complex. Here's what I'm talking about.

The Los Angeles Times has reported on a phenomena that those of us who have gone through more intensive DNA trainings know about, something called the Arizona Database situation. Typically, in a DNA case, the prosecution will contend that the chances of a random person having a profile that matches the suspect who left behind his DNA is one in something like 5 quadrillion (that's a one with 15 zeros after it). I've seen it get as high as quintillion (add another 3 zeros, but who's counting at this point, there's only 5 billion people on the plant, and something like 10 billion people who have ever even been in existence, now and all the way in the past). In other words, your client, the guy sitting next to you, must be the person who did it, because his DNA "matches," and the chances of that happening coincidentally are, well, impossible.

But, it turns out, it's not. This is where it gets complex, and deals with bizarre statistical things, which is why most lawyers don't really understand it, and try not to get DNA cases where there is no other evidence pointing to a suspect's guilt. You see, the DNA that the prosecution is matching is not your whole DNA profile, but just a small little part, up to 15 spots out of billions of spots, areas that are otherwise known as "junk DNA" (ie - it does not have anything to do with any traits such as blond hair, brown eyes, height, etc... It's all just random junk). It used to be that 9 matches was one in some huge number of billions, and that was enough, then it became 11, then 13, now 15. But, 9 is still enough, because it's still one in however many billion.

Until a government analyst in Arizona ran all the profiles in the 44,000 person Arizona database against each other, and found 122 people matched at 9 of those locations (or loci, as they're known in the field), something that was only supposed to happen one a billion times or so. And most of these people were completely unrelated.

Now, one would expect the FBI and other law enforcement organizations - who of course are in a search for the truth and justice - to find out why this has happened, to see if, in fact, there are cases where 2 people do have the same DNA and aren't related, to see how often this happens in other situations.

Done laughing yet? Have you read any of my previous posts about not trusting government and law enforcement? Here's another reason. The FBI did not do this. They stopped states from trying to do this. Not even through legal means, but through lies. They said things like: "tell the Court you'll lose your accreditation if forced to do this search," followed, sotto voce with "don't worry, you won't really lose it." Or how about: "tell the Court that it will tie up the system for days and can corrupt the whole system, rendering it useless." And guess what, these lies have worked most of the time. In California, the Courts have roundly refused to run the whole system against each other to see if anyone matches.

Except, in 2 states, they actually have done so. Don't worry, those of you sitting on the edge of your seats worried that they may have lost their accreditation or that the system shut down, corrupted - it all turned out fine. Oh, except they found a bunch of matches where they shouldn't have. One time they even had a match in 13 out of 13 loci, or more than a one in a quadrillion chance. The Maryland crime lab argued that those are probably, in fact, the same person, put in their twice under different names. Or maybe twins. They're so convinced it's something innocuous like that, they haven't bothered to even check.

Here's the thing: police science people (see, I can't even call them scientists they're such jokers) aren't willing to consider this possibility. Perhaps this "junk DNA" that people are comparing against aren't really junk at all. What if these code for character traits like, say, propensity for violence. While the statisticians are saying the chance of any two people having the same DNA at one of these spots is, say 1 in 10, maybe it's only one in 10 for non-violent people, but more violent people tend to have DNA that groups in the same area frequently. Then it wouldn't be so rare.

Now, I know that's a little complex, and I'm not a scientist (I sucked at math and science, that's why I became a lawyer), but it comes down to something like this. Say a Hispanic person does a murder, and leaves their DNA behind. The DNA gets compared to another Hispanic person, and they declare it's a match at 9 loci, and the chances of that are 1 in 5 billion (ie - there's no one else on the planet with the same DNA). They say this because each of those loci are completely random, there is no chance that one person has one profile over another at any of those loci.

However, what if it is found out that 1 of those loci really tends to show eye color (this is very basic, and it probably wouldn't show something that basic), another shows hair color, another shows skin tone, and another shows height. Now, the person who did the murder was 5'6", black hair, brown eyes, and slightly darker skin. If the suspect is about the same, you've described, in the Los Angeles area, probably about a million Hispanic men. Now if these loci were random, maybe it's one in a billion, but if they're not random, and you'd expect a bunch of Hispanic men to have this profile, then maybe the odds are only 1 in 30,000 Hispanic men may have this DNA profile (not all will have the same DNA, even if they have those traits, as there are a bunch of DNA spots that have to do with hair, eyes, skin and height, it just makes it more likely they'd have similar DNA). But, suddenly, you're looking at over 30 men in Los Angeles who may match that DNA.

This doesn't sound like a match to me (unless you're law enforcement, and it's close enough for government work - regardless, you're getting some Hispanic guy with a record off the street, if he didn't do this, he probably did something else anyways).

No wonder why the FBI is trying to bottle this stuff up, imagine DNA wasn't the picture perfect thing they've portrayed it to be.

Now, I know this is thick, and I probably described it poorly. If enough people write me and say "huh?" I'll try and re-do in a better way, maybe even consulting with someone who knows this better than me. If not, I'm interested in whether people understand my point here (besides the obvious, which is, another law enforcement cover-up).

Read the article for more info, they describe it better than I do.

Special thanks to reader MT, who pointed this article out to me this morning before I had a chance to look at the paper.


Anonymous said...

I've been scouring the web since the LA Times came out with the online version of the article. It literally blew up all over on various blogs, message boards, etc. across multiple disciplines.

Your entry is hands down the most accurate and "original" that I've come across. Thank you for taking the time to ferret through the article and explain some things that might not be so obvious to the average attorney.

Lil Spicy

A Voice of Sanity said...

See Exploring the wolves in dogs' clothing for some interesting material. It seems that a minute change in DNA can 'ripple' through the being causing many, many changes.