Robert Xie’s small wooden garage in Sydney’s north-west was unremarkable but for a tiny clue that hid a terrible secret.
There, on the concrete floor, underneath the bottom of an unused wooden tallboy, was a faded bloodstain, smaller in size than a cigarette lighter.
“Stain 91”, as the dried patch of blood would come to be known in court, would forensically connect Xie to the murders of five of his wife’s family. Good old-fashioned police work had led to the discovery of Stain 91, but only a futuristic forensic tool could unlock the critical DNA that would prove so damning for Xie. Had Xie’s murders happened several years earlier, Stain 91 would have been too difficult for forensic scientists to decipher.
Until recently, labs struggled to untangle and deliver reliable results on crime scene samples that were mixed up with DNA from two or more people.
Initial tests on Stain 91 revealed it was an incredibly complicated sample. In it was blood from at least four of Xie’s five victims.
Mark Tedeschi, QC, one of the lead crown prosecutors on the Xie case, pinned his hopes on a company in Pittsburgh, in the United States, called Cybergenetics.
Cybergenetics’ founder Dr Mark Perlin had developed very powerful software capable of unravelling tremendously complicated, mixed DNA samples. Perlin’s lab ran Stain 91 data through its TrueAllele software, churning it through untold complex mathematical algorithms, and emailed the report to Tedeschi.
“The results were astounding,” Tedeschi says.
Working outside the limits of traditional DNA test methods, TrueAllele calculated devastating match statistics in Stain 91 of at least four victims beaten to death by Xie. The software also managed to show in-depth and undeniable comparisons between Stain 91 and a blood spot found on a mattress at the murder scene.
Xie’s lawyer tried, and failed, to fight the admissibility of the TrueAllele evidence. In January, Xie was found guilty of five counts of murder.
Tedeschi tells The Saturday Paper that Stain 91 was probably the most complex DNA sample ever introduced to a criminal trial in Australia.
“They were very significant statistics, which we were able to use,” he says.
But not everybody is comfortable with the new forensic software, which has turned previously indecipherable DNA into often-profound courtroom evidence.
Since 2012, police forces across Australia have increasingly used a DNA computer program called STRmix, which was jointly developed by scientists in South Australia and New Zealand. STRmix evidence has been used to prosecute some of the country’s most vicious rapists, child molesters and murderers. Its emergence as a key forensic tool in Australia has gone largely unreported.
Probabilistic genotyping software, such as STRmix and TrueAllele, run DNA data through statistical computer algorithms to calculate a likelihood ratio that a particular person’s DNA is present in a mixture.
In December 2014, the forensic laboratory that analyses crime scene DNA for Queensland Police became aware of “a minor miscode” in the STRmix software.
Queensland Health Forensic and Scientific Services executive director Paul Csoban tells The Saturday Paper that the miscode was limited to just 60 cases. He says “an internal review” had confirmed that 4500 other samples analysed by STRmix had not been affected.
When contacted by The Saturday Paper, a spokesperson for STRmix said in a statement that the issue “was largely restricted to Queensland and reflected the way that they used the software”. Both STRmix and Queensland Health stated the miscode had resulted in a small distortion to some DNA match statistics. Csoban says 23 of the 60 cases required retesting to produce new data.
A right-to-information request submitted by The Saturday Paper to Queensland Health and Queensland Police revealed the 60 miscoded cases included alleged crimes of murder, rape, incest, sexual assaults against minors and armed robbery.
STRmix and Queensland Health were in conflict over which organisation had first noticed the bug in the software. There were reports Queensland Health had not purchased the most up-to-date instruction manual to properly operate STRmix.
“STRmix developers released information that a minor miscode was contained within the software that the [Queensland Health] laboratory was using,” Csoban says.
However, STRmix developer Dr John Buckleton says the bug only became apparent when Queensland Health approached their office in South Australia. He adds that Queensland Health had been unwilling to share potentially important and useful details about the miscode.
Michael Bosscher, a Brisbane-based defence lawyer, says he received an “arse-covering letter” from Queensland Health, generically advising of a forensic miscode. At the time, Bosscher was representing Andrew Burke, a teenager charged with the rape and murder of pregnant Queensland mother Joan Ryther.
Only when Bosscher followed up the issue did he discover how his client Burke had been affected, and that his DNA samples had required retesting.
Bosscher says he has “serious concerns” about the rise of STRmix and probabilistic genotyping software in Australia.
“Until disclosure that there was a miscode in the program, nobody, including Queensland Health, was aware of it. What concerns me is, what else has been miscoded?”
Mark Tedeschi rejects suggestions the software has tilted the playing field in favour of prosecutors. He says STRmix was extensively verified and robust.
“I think what defence lawyers are grumbling about is that STRmix evidence is very powerful and very convincing, and it influences juries,” he says.
STRmix’s Dr Buckleton points to multiple scientific validation studies as evidence of the software’s reliability in solving critical DNA riddles. However, he refuses to explicitly confirm that STRmix is 100 per cent reliable.
“We do everything we can to make it as good as possible, but we always hold open the need for checking,” Buckleton says.
“If we ever said it was 100 per cent reliable we might be inviting complacency and we don’t want to do that.”
In 2015, the Federal Bureau of Investigation in the US began using STRmix. It has also been rolled out across various jurisdictions in the US, Canada and Europe.
The biased view is DNA stands for Do Not Acquit, according to Clinton Hughes, a New York attorney who works for The Legal Aid Society. Most of Hughes’s casework involves helping clients implicated by DNA, or wrongly imprisoned men who could potentially be exonerated by compelling forensic evidence.
In 2016 Hughes was involved in a high-profile New York murder trial where STRmix and TrueAllele, the world’s two leading probabilistic genotyping products, generated dramatically different results on the same data.
Underneath the fingernail of a 12-year-old boy who had been strangled to death was a tiny unknown amount of DNA from someone else. Prosecutors believed the DNA had come from the child’s mother’s boyfriend, Oral Hillary.
TrueAllele ran the data and produced an inconclusive result. But STRmix, consulting for the prosecution, showed Hillary’s DNA was strongly connected to the boy’s fingernail.
The DNA sample was controversially microscopic, infinitely smaller than Stain 91. The judge ruled that the STRmix result was unreliable and therefore inadmissible. One month later, Hillary was acquitted of murder.
Hughes describes the STRmix and TrueAllele evidence as “Alien versus Predator”. Had STRmix evidence been allowed into evidence Hillary would have “definitely been in trouble”, Hughes says.
“Every lawyer who practises defence has huge concerns when we can’t really confront the evidence that implicates our clients – particularly DNA evidence.”
Both STRmix and TrueAllele say they work with defendants to help make sense of results and the technology. Yet there are only so many scientists who understand the software to go around.
Tedeschi, however, is unmoved. DNA has always been complex for juries to comprehend, he says.
“Part of a lawyer’s job is to reduce DNA to a basic element so the jury can understand it. It’s no different to any other type of forensic evidence that needs convincing.”
TrueAllele, which specialises in the most intricate of cases, such as Stain 91, has been used in more than 400 criminal trials across 32 American states.
But there is another face to the technology. Last year Cybergenetics helped exonerate two men wrongfully convicted of gang-raping a woman in 1989.
“We don’t just work for one side,” Mark Perlin said. “We work for the truth.”
This article was first published in the print edition of The Saturday Paper on June 17, 2017 as "Crime screen".
For almost a decade, The Saturday Paper has published Australia’s leading writers and thinkers. We have pursued stories that are ignored elsewhere, covering them with sensitivity and depth. We have done this on refugee policy, on government integrity, on robo-debt, on aged care, on climate change, on the pandemic.
All our journalism is fiercely independent. It relies on the support of readers. By subscribing to The Saturday Paper, you are ensuring that we can continue to produce essential, issue-defining coverage, to dig out stories that take time, to doggedly hold to account politicians and the political class.
There are very few titles that have the freedom and the space to produce journalism like this. In a country with a concentration of media ownership unlike anything else in the world, it is vitally important. Your subscription helps make it possible.
Select your digital subscription