DNA mixtures can by mind-bogglingly complex. The genetic traces of more than two people can create statistical chaos indicating someone’s DNA is included, or excluded, from a sample – based simply on the judgment of the person doing the testing.
This chaos has been interpreted by mathematical forensic analysis. But it involves subjective estimations by trained experts.
Those estimates – called the Combined Probability of Inclusion – have been used in crime labs nationwide for decades. But that method of statistical analysis has no grounding in science – and needs to be overhauled, a controversial White House report now contends.
The President’s Council of Advisors on Science and Technology findings – known popularly as the PCAST report – instead backs computer programs with complex algorithms to interpret the mathematical riddles. But the report contends that even those tools need further scientific validation.
The report thereby questions potentially thousands of criminal cases, experts told Forensic Magazine.
“Subjective analysis of complex DNA mixtures, including with the widely-used Combined-Probability-of-Inclusion methods, is not foundationally valid,” the PCAST panel contends, “and objective analysis of complex DNA mixtures with probabilistic genotyping software is promising, but has not yet been sufficiently and appropriately validated and their limitations to be considered reliable for all complex mixtures.”
CPI involves a trained expert making estimations to create a simpler way to look at the alphabet-soup of alleles that can show up in the same sample.
But depending on the hand doing the estimations, the process is not standard enough to be presented in courtrooms, the panel contends over six pages at the heart of their lengthy report released last week.
CPI has drawn its critics in recent years – and has even resulted in exonerations. Greg Hampikian, a professor of biology and criminal justice administration at Boise State University, who works with the Idaho Innocence Project, worked with CPI and then probabilistic genotyping computer program TrueAllele to determine that Darryl Pinkins served 24 years in an Indiana prison for a 1989 gang rape he did not commit. Pinkins was released from behind bars in April.
Hampikian told Forensic Magazine that CPI is “voodoo” that has erroneously sent people to prison.
“It’s a well-documented tragedy that people have gone to prison based on faulty CPI methods,” said Hampikian.
Others defend CPI as a method that works, when analysts are trained appropriately, and the work is done objectively. Bruce Budowle, executive director of the Institute of Applied Genetics at the University of North Texas Health Science Center, worked on DNA for decades at the FBI. He told Forensic Magazine earlier this year that CPI works – and works well – when it is properly employed.
PCAST disagreed.
A handful of computer programs with hyper-complex algorithms have cut through the jungle of data to give high-level statistical interpretations to lawyers, judges and juries. They have been used with increasing frequency in courtrooms across the United States.
But PCAST also said these major probabilistic genotyping programs also need more work. The developers of the two major players in the probabilistic genotyping market blasted the report, saying that their tools are already verified – and solving cases.
Mark Perlin, the developer of TrueAllele (the tool that freed Pinkins from prison earlier this year), said TrueAllele has been accepted in courtrooms all over the country – and is solving cases where other tools have failed. His program, he said, is already validated to five contributors in various scientific publications – even though PCAST said the cutoff was three.
“(TrueAllele) fully solved DNA mixtures 10 years ago,” Perlin told Forensic. “NIST pretends there is a problem in order to get funding. They are biased and lack expertise. We don’t need another government boondoggle.
“The PCAST report is well-intentioned, but misguided,” Perlin added. “Forensics needs better science, not more bureaucracy.”
STRmix (pronounced “star-mix”) is the other major DNA-mixture program. John Buckleton, the STRmix developer, also said his program is scientifically-validated – and that the PCAST report was simply overlooking key publications. He provided a list of them in a detailed rebuttal to the group.
“I suggest that insufficient research was undertaken by the committee,” Buckleton wrote to the panel. “The simplest course was to ask us. The conclusions of the committee are incorrect, damaging, and need to be revisited.”
The two programs, however, get different results on the same cases, as the PCAST report noted. Most recently, the Nick Hillary murder trial in Upstate New York drew major headlines. TrueAllele found the various DNA mixtures at the crime scene could not place Hillary there, while the prosecution was relying on a STRmix interpretation of fingernail scrapings that indicated he was there. However, a judge threw out the STRmix evidence just before the trial was set to begin.
Today, the same judge in St. Lawrence County, New York, again threw out the STRmix findings - leaving the prosecution without DNA evidence of any kind. The attorneys for Hillary, accused of strangling 12-year-old Garrett Phillips in 2011, have in the meantime moved to have the judge decide the case - and not a jury.
Still others have blasted the PCAST report itself, including the National District Attorneys Association, which called the report “scientifically irresponsible.”
Rockne Harmon, former longtime prosecutor in the Alameda County District Attorney’s Office, told Forensic Magazine in a phone interview that the report was a way for the committee to “throw defense attorneys another bone,” and to call into question just about every forensic discipline.
The same could be said of DNA mixtures, Harmon added.
“There’s so much out there about TrueAllele and STRmix,” the former DA said. “It’s clear probabilistic genotyping is a solution for all the problems that could arise with complex DNA mixtures.”
