https://amp.cnn.com/cnn/2018/06/02/us/golden-state-killer-unsealed-warrants/index.html
When the suspected Golden State Killer drove into a Hobby Lobby parking lot in April, investigators were waiting nearby. As he walked into the craft store, it gave them a perfect chance to collect a secret DNA sample.
Police swabbed the driver's side handle of [the suspect's] car, according to arrest and search warrants released Friday.
Authorities sent it for testing and matched it to semen recovered at some of the Golden State Killer's crime scenes, the arrest warrant said.
[...] The stop at the Hobby Lobby was just one of several ways investigators used to zero in on a suspect. Earlier this year, police tracked him down by comparing genetic profiles from genealogy websites to crime scene DNA, according to investigators.
On April 23, a day before his arrest, police say they collected multiple samples from a trash can outside DeAngelo's home in Citrus Heights, a town 16 miles northeast of Sacramento. They had watched the home for three days, the warrant said.
Previously: DNA From Genealogy Site Led to Capture of Golden State Killer Suspect
GEDmatch: "What If It Was Called Police Genealogy?"
(Score: 3, Interesting) by AthanasiusKircher on Tuesday June 05 2018, @02:28AM
First, I should be clear that what I meant was serious statistical misrepresentations in the way evidence was presented. Sometimes it's due to actual errors, but more often it's just playing off a jury's ignorance and when they hear "DNA match," they just assume it's ironclad.
There have been loads of coverage of the potential problems with this sort of thing in recent years. See articles about it here [sciencemag.org], here [theatlantic.com], here [innocenceproject.org], here [pbs.org], here [latimes.com]. And the FBI even admitted [usatoday.com] to a pattern of statistical errors in DNA calculations over 15 years. Granted, the magnitude of the FBI errors mentioned in the last link may have been small enough not to cause wrongful convictions, but if such errors could go overlooked for a long time, it's certainly likely that other estimates are sometimes off.
That's likely true, but we probably are significantly less likely to find out about wrongful convictions due to statistical error or misrepresentation, since usually it's significantly harder to challenge a ruling where a match was found. Evidence of contamination is a clear reason to challenge a ruling, but proving that a statistical misrepresentation (or even a blunder) had an effect on a jury is a lot harder.
If you're referred to ERRONEOUS cases, yes DNA is probably much more likely to be used to exonerate a bad conviction than to be used to erroneously convict. But again, part of the issue is that DNA evidence is seen by many juries (and judges) without nuance -- a "match" is a match, and so we just may not have gotten to a big wave of challenges yet because not enough judges are convinced to examine an appeal on this basis yet.
I'm not saying it's hugely widespread in leading to false convictions. But exaggeration of DNA matches has happened and is definitely a problem.