How do you analyze old DNA?

February 5, 2007 3:40 PM

How does DNA testing on old samples work?

I've been reading about Larry Youngblood's exoneration from rape charges in 2000 -- apparently new testing of old degraded samples proved it wasn't him. Can anyone explain in simple terms what new technology allowed his exoneration?

posted by footnote to science & nature (16 comments total)

Polymerase Chain Reaction. It permits a testing lab to take a vanishingly small sample of old DNA and to replicate it enough so that there's an adequate quantity for testing.
posted by Steven C. Den Beste at 4:05 PM on February 5, 2007

The Neanderthal Genome Project is sequencing 38,000 year old DNA, so <20 years is nothing.
posted by grouse at 4:20 PM on February 5, 2007

So when it's said that the DNA was "degraded," does that just mean there wasn't enough of it to test in pre-PCR days?
posted by footnote at 4:25 PM on February 5, 2007

I could be wrong, but I don't think any kind of forensic DNA technology was commonly used in 1984. When they said it was too degraded, they meant for serological testing.
posted by grouse at 4:29 PM on February 5, 2007

There are two changes between 1984 and now. As Grouse mentions, the first was the development of DNA testing as such.

The second change, in the late 1990's, was the appearance of relatively cheap and reliable mass-produced PCR machines. Currently you can set yourself up to do PCR for less than $5000, getting you both a machine and an adequate supply of relevant reagents. Prices have been coming down, but that reached the level of "widespread" and "reasonably cheap" in the late 1990's.

That wasn't a stair step, more like a cross over. As more and more machines were created and as economy of scale kicked in and prices dropped, they began to get used for more and more things, including criminology.
posted by Steven C. Den Beste at 4:46 PM on February 5, 2007

I don't want to derail, but I'm curious: how do you make sure it's really the original sample you're replicating, and not some other DNA that wandered in (say, from a HeLa cell, or some other lab pest)?
posted by nat at 4:49 PM on February 5, 2007

how do you make sure it's really the original sample you're replicating, and not some other DNA that wandered in (say, from a HeLa cell, or some other lab pest

You're right that this isn't a trivial issue. They start by using steril techniques to reduce contamination. Then, if it's human DNA you're replicating, they'll align the results to the consensus human genome to filter out any foreign DNA. Anything that doesn't map reasonably well will be discarded.
posted by chrisamiller at 4:53 PM on February 5, 2007

Contamination isn't as much of a problem as you might think. In the case mentioned in the OP, the test excluded the accused.

What that meant was that they found and tested a lot of DNA, and none of it was his. It doesn't really matter where the DNA did come from; what's important is that it didn't come from him, and if he'd been guilty his DNA should have been found where they were looking.

To make the test really valid they do need to show that the DNA they were testing was from a human, but that doesn't require totally eliminating all other forms of DNA.

Yet it's the case that the procedures used do tend to exclude irrelevant DNA. Partly that's a function of the primer DNA they use (which is chosen to be human-specific) and partly it's a function of the cutting enzymes used in the second part of the testing process, which are also human-specific. (I don't know what the official jargon is for those.)
posted by Steven C. Den Beste at 5:17 PM on February 5, 2007

...It doesn't even require that they eliminate all other human DNA. It's possible to test a mix of DNA from several humans, and as long as there's some specific DNA feature of the accused that is not present in the DNA of any of the people who were in the sample (even as contaminants) then it's still possible to exclude the accused.

In the Duke non-rape case they did genetic testing on samples taken from the alleged victim (anal, vaginal, and oral swabs as well as swabs of her underwear), and in that case what they were specifically looking for was Y-chromosome DNA. The testing lab found Y chromosome DNA and was able to show that it came from several different men. But the test result also showed conclusively that none of it came from any of the men who were at the party.

[Part of the scandal in that case is that Nifong hid that result from defense attorneys for several months despite a North Carolina "full discovery" law compelling its early disclosure. He also lied to a judge about it more than once. Once it came out, Nifong was forced to drop the rape charge, and a month later he handed the case over the North Carolina AG. There's every reason to believe that all charges against the three young men will soon be dropped. Nifong is in deep trouble and is probably going to be disbarred. He will probably also be sued and he might well face criminal charges himself.]

In the OP case, it was probably similar. They probably found DNA from several humans, but were able to show that none of that DNA came from the defendant.
posted by Steven C. Den Beste at 5:33 PM on February 5, 2007 [1 favorite]

Of course, part of what made nat's question so great is that HeLa cells are human. But HeLa cells also have a known DNA fingerprint, so their contribution can be easily identified.

what's important is that it didn't come from him, and if he'd been guilty his DNA should have been found where they were looking.

And even more impressively, Walter Cruise's DNA was there.

cutting enzymes

You mean restriction enzymes? Those have mainly gone by the wayside in DNA testing—variability is now measured by the size of short tandem repeats.
posted by grouse at 5:36 PM on February 5, 2007

Before the advent of PCR, comparison of sample DNA (crime scene vs. accused, for instance) was done using a technique called Restriction Fragment Length Polymorphisms (RFLP). This assay utilizes the "cutting enzymes" mentioned above. Here's a very basic web page on the topic. As mentioned before, newer techniques like PCR can amplify tiny amounts of DNA which allow detection with a smaller sample than was possible with RFLP.
posted by dendrite at 1:14 AM on February 6, 2007

Thanks for this info! If anyone's still reading, here's another follow-up question: how much is this analysis dependent on the skill of the technician these days? Should I have confidence in a statement like "In early 2001, officials got a hit, matching the profile of Walter Cruise" without knowing who those "officials" were?
posted by footnote at 6:31 AM on February 6, 2007

The second change, in the late 1990's, was the appearance of relatively cheap and reliable mass-produced PCR machines.

Earlier than that. It was already common when I started grad school in 1993. (Admittedly, I wasn't in forensics, but PCR has many many applications outside of just forensics.)
posted by DevilsAdvocate at 9:01 AM on February 6, 2007

how much is this analysis dependent on the skill of the technician these days?

I think it's a threshold issue. Beyond a certain basic level of competence, everyone's going to get the same answer. If the tech is incompetent, he can mess it up.
posted by Steven C. Den Beste at 9:31 AM on February 6, 2007

footnote: here's some more information about DNA forensics from the gummint.

In STR analysis, a computer will decide whether a set of measurements could possibly have come from the suspect or not, and then tell you the probability that this match would have occurred by chance (although be careful to avoid the Prosecutor's and defense attorney's fallacies). Fairly foolproof at that end. Only gross incompetence in handling the samples (such as analyzing the wrong sample) would lead to an unequivocal match. Negligence in allowing contamination would also lead to a multiple match, but this would be obvious from a computer report.

Unfortunately, gross incompetence happens.
posted by grouse at 1:38 PM on February 6, 2007

I marked Grouse as best answer because I went back and read the cases, which make clear that in the 80s the issue was serological testing.
posted by footnote at 1:58 PM on February 6, 2007