Sunday, February 28, 2010


I found an interesting article about how the public's ignorance about how DNA matching works (and the FBI's work to ensure that it stays that way) when DNA is used to generate leads in cold cases by being compared to large databases.

The general gist goes like this. DNA matching works on markers - the more markers, the lower the probability that I've hit a coincidental match. The numbers that most people think of when they think about DNA matching (the FBI's standard numbers), say that the chances of a coincidental match are vanishingly remote - perhaps one in trillions. But that relies on 13 markets. Start bringing that number down, and the chances of a match start climbing. Got only five markers? Well, now you're in the one and about 1.1 million range. Sure that seems pretty small. But, if I'm applying that to a database of DNA samples and looking at any hits that come back - if there are one million samples in the database, I'm nearly assured of getting one hit - even if the actual perp isn't in the database at all. So in cases where there is no other evidence, and only a few markers, DNA isn't all that it's been cracked up to be. But, as the saying goes: They don't want you to know that. And therein lies the story.

Of course, the point could be made that if someone has their DNA in a national offender database, they must have done SOMETHING. (Although this is where I point out that in the United Kingdom, they take a DNA sample if they arrest you for causing a serious traffic accident.) It must be remembered, that the point behind the justice system is not simply putting bad people, or even people who have done bad things behind bars – it must be to sanction people for an individual bad act that they've been clearly connected to. Many people already have some difficulty in keeping "Has been arrested for a crime" and "Absolutely guilty" straight. We should take a hard look at law enforcement when they start attempting to undermine what should be legitimate questions, and muddy the water even further.

1 comment:

Keifus said...

The database match probability is an important point, and the author's right that it should be reflected accurately at court proceedings (as I strongly doubt that it is). Some spectral analysis tools (mass spec is the most popular and effective one) can compare a measured spectrum against a library of known samples. You can generally get a good idea of how good that match is too, either by some quantitative method, or just looking at the lines. Not sure if DNA is similar, or similarly easy to analyze.

Important facts the article leaves out: how often are forensics scientists presented with degraded DNA evidence? How does the probability compare, even in teh worst cases, to other (subjective) analysis such as ballistics matching, analysis of fibers, tire tracks, etc.?

He also should careful about including in his anecdotes how DNA has been used to rule out a suspect. I think even a poor set of markers can show that a genetic match is extremely unlikely. A better tool for exoneration than for proving guilt maybe.