Statistical basis of DNA matches? Does DNA ID work?

Factual Questions
1

Inspired by this article:
http://www.latimes.com/news/local/la-me-dna20-2008jul20,0,1506170,full.story

where in summary, if one compares every DNA sample in a database of criminal records with every other sample, one finds many (tens) of partial matches. Far more than the usual concept implied by the odds of “113 billion to 1”. This information caused enough concern at the FBI that they used techniques that would clearly be illegal for private citizens to use to suppress this information (an FBI official urged their state counterparts to lie to a court).

The question is, what is the statistical basis of the use of DNA matches? Can someone explain how these matches can show up (according to the article they have been found in numerous independent databases) and not invalidate DNA matches?

2

The way the FBI is behaving, maybe it does invalidate DNA matches. Or at least force the courts to re-evaluate how meaningful a DNA match really is. The worst-case scenario, as far as the Feds are concerned, would be if they had to retry every case that was ever won based on DNA evidence.

3

Reading that article it still does not sound like DNA matching is made worthless. Even going with the notion that it is not 1 in 1 trillion the chances of a mistaken hit are slim. By their own number on 65,000 records searching on 9 loci they get 122 hits. That is a 0.18% chance of a mistaken hit. After that you combine it with other evidence (a duplicate hit may live far away or been on vacation or have an iron clad alibi and so on). Also, if they get a hit seems they should be able to check for a closer hit on 10, 11, 12 and 13 loci which get progressively even more reliable.

4

I guess I hyperbolized a bit, but it sounds like we might have to change our thinking regarding what a DNA match means. It might be that a DNA match doesn’t mean “We know he was there,” but rather “He was probably there.” A very significant difference. A DNA match is still very compelling evidence, but it might not the be the end-all slam-dunk that we’re used to thinking of it as.