Advertisement

DNA mixtures are a growing concern in the world of forensic science. The detection of even miniscule touch DNA in a pool of blood, or the skin cells left behind on the handle of a kitchen knife that later becomes a murder weapon, is now more possible than ever before with hypersensitive instruments. Forensic experts contend detectives and scientists need to be able to understand how the invisible ATCG alphabet soup at a crime scene may implicate a murderer, an innocent person—or both.

Forensic science has been grappling with this, and software solutions involving complex statistical modeling and big-data computing have emerged to replace previously subjective expert analysis. DNA mixtures comparison was a major topic of presentations yesterday at the annual International Symposium for Human Identification, with the industry leader STRmix the focus of several talks. The forensic analysis of complex mixtures will be a major research project, the National Institute for Standards and Technology also announced Tuesday. But the stiff competition between the two major technologies—STRmix and TrueAllele—has led to accusations that “conflicts of interest” at the government agency favor one of the tools, and are therefore driving the science.

STRmix, Validated by the FBI

Several ISHI presentations focused on STRmix, a technology developed by scientists in New Zealand, where it has been in use for routine casework since 2012.

An FBI analyst presented at the Seattle meeting Wednesday morning, showing how the Bureau had completed internal validation of the tool. That validation involved assessing more than 2,800 mixtures of 3 to 5 different people, using the tool at 31 different laboratories nationwide, according to Tamyra Moretti of the FBI.

Another analyst showed how STRmix statistics have cracked criminal cases. Cristina Rentas of the Florida-based DNA Labs International presented several cases, one of which showed how STRmix comparisons actually changed the direction of detectives’ suspicions. A three-person mixture was from a weapon, and the genetic material showed the victim was almost certainly in there. But the suspect was a different story. First, the STRmix analysis showed the suspect was almost assuredly part of the mixture (12 million times more likely). But the second run was distinctly less certain (only 17 times more likely). What had changed were the assumptions programmed into the software at the beginning of the tests, Rentas said. Once they started doing more genetic testing, adding more data from people surrounding the case, they hit on the shocking result: the main suspect’s brother was actually the culprit.

“It was head-spinning for us—it didn’t make sense to us, but then we decided to break down the likelihood ratios,” said Rentas. “As more and more labs are starting to bring online STRmix, and more and more labs are starting to use probabilistic genotyping, it’s important to share that this is something that can make a difference in the case—and bring a case in a new direction.”

STRmix is currently in 29 laboratories across the U.S., as well as in places like Australia, according to John Buckleton of the New Zealand-based Institute for Environmental Science and Research, the creator of STRmix. In the U.S., the software is gaining traction, coming online in dozens more laboratories across the nation, the company recently announced. The Bureau of Alcohol, Tobacco, Firearms and Explosives (ATF) announced last month it would use it, as would a new handful of sheriff’s offices in California, Florida and Illinois.

Buckleton has pointed to recent cases like one in Michigan where STRmix solidified a prosecutor’s case against a rapist based off a DNA mixture on the victim’s toes.

“I am excited at the growth STRmix has experienced in the U.S. The U.S. would now be considered a leader in the field of probabilistic genotyping,” Buckleton told Forensic Magazine this week.

The three developers of the software from Australia and New Zealand “do not receive any benefit direct or indirect from sales of STRmix” because they are civil servants paid by their governments, according to Buckleton’s website.

NIST to Assess

The DNA mixture question has been the subject of critical reports such as the watershed 2009 National Academy of Sciences report “Strengthening Forensic Science,” and last fall’s President’s Council of Advisors on Science and Technology, or PCAST, report on forensic science. That PCAST report recommended a complete reevaluation of DNA mixture analysis—including STRmix and TrueAllele.

NIST will now assess the reliability of picking apart DNA mixtures from multiple contributors. The “scientific foundation review” will be led by NIST Fellow John Butler, a noted DNA expert, and Sheila Willis, the former director of Ireland’s national forensic laboratory, who is a chemist.

“The goal is not to undermine these methods, but to determine their bounds of reliability so they can be used appropriately,” said Butler, in a NIST statement.

NIST did not specify which DNA mixture tools will be analyzed in their announcement.

Buckleton had served as a guest lecturer and researcher at NIST for roughly two years, until earlier this year—after his role there was questioned by the competitor, TrueAllele.

TrueAllele—and Conflicts of Interest Alleged

The major competitor to STRmix is TrueAllele, a software made by the Pittsburgh-based Cybergenetics. The software was created by Mark Perlin, who heads the company, and is a former Carnegie Mellon scientist.

TrueAllele is currently in at least seven labs across the country, and another seven have access to the system, Perlin told Forensic Magazine. (TrueAllele was used by the New York State Police crime lab for several years. But the lab was later enmeshed in allegations and counter-allegations involving accreditation, and a lawsuit alleging civil rights violations.)

TrueAllele has been used in more than 40 U.S. states, by both prosecutors and defense teams. Two men who served years in the Indiana prison system for a 1989 gang rape were exonerated early this year after TrueAllele proved they could not have been in the DNA mixtures collected.

Perlin says his software is superior with more complex mixtures. But he alleges his technology has suffered from “conflict of interest” at NIST—where Buckleton, the STRmix creator, was a guest lecturer up until last year.

Last year, Perlin sent a letter to the U.S. Department of Commerce asking for a look into the “special status” of Buckleton, ESR and STRmix.

Buckleton’s place within NIST had given STRmix an unfair advantage, despite the scientific considerations, Perlin alleged in the July 8, 2016 letter.

“The closeness between NIST and ESR is apparent to the forensic DNA community,” he wrote. “This special relationship between your agency and a large foreign company may have unleveled the playing field for a small American innovator.

“The impact goes far beyond commerce, since widespread usage of weak crime-fighting DNA technology affects justice for all Americans,” he adds.

Within months of that letter, Buckleton was no longer at NIST.

Earlier this year, NIST referred Forensic Magazine to the Department of Commerce for comment on the conflict allegations, and Buckleton's  departure. The Department of Commerce at first said a response was forthcoming, and did not respond to months of further emails from Forensic Magazine.

Buckleton, this week, said he did not want to delve into the specifics of his departure from NIST.

“Suffice to say that I enjoyed the two years I spent at NIST and was sorry to leave,” he wrote to Forensic Magazine. “My departure, though, led me to work with Professor Bruce Weir with the Department of Biostatistics and the Director of the Institute for Public Health Genetics at the University of Washington Seattle. Professor Weir is a leader in the field of evidence interpretation, so working with him has been tremendously exciting.”

The acting NIST director sent Perlin and TrueAllele an April 2017 letter that indicated the agency wouldn’t necessarily favor STRmix.

“Commercials products, materials, and instruments are identified in NIST’s communications and documents for the sole purpose of adequately describing experimental or test procedures,” wrote Kent Rochford, the acting NIST director. “In no event does such identification imply recommendation or endorsement by NIST of a particular product; nor does it imply that a named material or instrument is necessarily the best available for the purpose it serves.

“NIST strives to avoid even the appearance of endorsing commercial products, while working with a breadth of measurement techniques,” Rochford added.

Perlin told Forensic Magazine this week that NIST shouldn’t be conducting the tests on DNA mixture software at all.

“A biased referee cannot conduct a fair study,” he said. 

Tests at a New York Murder Trial

The two programs were put to the test in a high-profile murder trial in Upstate New York last year. Oral Nicholas Hillary stood accused of strangling the 12-year-old son of his ex-girlfriend, in a case with racial undertones within a small community near the Canadian border.

TrueAllele was first consulted on the case. Every sample distanced Hillary from the scene—including the fingernail scrapings, the company said. That evidence was not used in court, however.

STRmix was enlisted by prosecutors next. Buckleton told the court he had assessed the complex genetic mixture under the boy’s fingernails—and determined Hillary was in there.

However, the judge disagreed with the stated methodology in his decision, ruling he would not allow Buckleton to “pick and choose data from different ‘reliable sources,’” for the prosecution’s case, since the science had not been proven within the state’s crime labs.

“Here, the lack of internal validation by the New York State Police crime lab, as candidly admitted by Dr. Buckleton, precludes use of the STRmix results,” the judge ruled. “Neither the STRmix nor the (Random Match Probability) results may be used in this case.”

Hillary walked free, with many believing him innocent, and other onlookers convinced he was guilty. 

Advertisement
Advertisement