DNA profiling reliable?

[Deetee]

Seems that the FBI quoted statistics on the likelihood of finding a matching DNA profile may be a bit suspect (http://www.latimes.com/news/local/la-me-dna20-2008jul20,0,1506170,full.story).

The original work on the Arizona data base was done back in 2001.
There have been some analyses (http://dna-view.com/ArizonaMatch.htm) of the mechanics behind this I found thru googling.
Has any of this been used to overturn a conviction or get cases thrown out?

ETA - perhaps I would have better using the title "DNA profiling interpretation reliable?", since there is nothing inherently wrong with the methodology.

[I Ratant]

The Arizona data base found two people with the same birthday in essence.. but throw in the birthdate, and the choice must be more precise.
As one of the analyses pointed out, and any math class can,
finding two people with the same birthday requires as few as 23.
Two on April 22, say.
But two on April 22, 1989, really lowers the odds.
The same with a DNA match.
It depends on how many of the other mis-matches you want to ignore.

[blutoski]

Seems that the FBI quoted statistics on the likelihood of finding a matching DNA profile may be a bit suspect (http://www.latimes.com/news/local/la-me-dna20-2008jul20,0,1506170,full.story).

The original work on the Arizona data base was done back in 2001.
There have been some analyses (http://dna-view.com/ArizonaMatch.htm) of the mechanics behind this I found thru googling.
Has any of this been used to overturn a conviction or get cases thrown out?

ETA - perhaps I would have better using the title "DNA profiling interpretation reliable?", since there is nothing inherently wrong with the methodology.

I've always felt that the technique is very reliable for exonerating, but less so for providing evidence toward conviction, because of the chance of false positives.

The thing to keep in mind is that we don't sequence the suspect's entire genome - just a handfull of variable genes.

The chance of an accidental match will vary with the alleles the suspect actually has: if he has very common alleles, there may be millions of people in the country in question who match; if he has very rare alleles, there may only be a few other matches in the country.

[Wildy]

But I thought that was the point of using variable number tandem repeats?

Isn't the whole concept of VNTRs that they are different enough in each person to make a match unique?

[bokonon]

I suppose once we reach a point where all felony arrestees have their DNA added to a database, the chance of false positives may rise. In most current cases, where you have to find a suspect and get a new sample from the suspect to match, the odds that the suspect will also be a false positive are remote. If you have no suspects, and you get a match in the database for someone who can't supply an alibi, I'd be a little more cautious about assuming it isn't a false positive.

[blutoski]

VNTRs practically eliminate the chance of false positives. But not all prosecutors use it (they want a positive match at any cost).

The most common cause of false positives remains accidental contamination.