Defense Attorneys Seek Access to DNA-Matching Software’s Source Code
TrueAllele DNA testing software has been employed in hundreds of criminal cases around the country since 2009. The software is used to analyze evidence containing mixtures of genetic material and determine whether it contains a match to a suspect or DNA archived in a database. A similar DNA testing program, STRmix, has also been used in some criminal cases, including by the California Department of Justice.
While prosecutors, and even some defense attorneys, have hailed TrueAllele, some contend that in order for defendants to have a fair trial, the core “engine” of the software – its source code – must be open to scrutiny and validation. Both prosecutors and TrueAllele’s creators have consistently fought to prevent such scrutiny.
The software is a product of Pittsburg, Pennsylvania-based Cybergenetics. The company was founded by its chief scientist and executive officer, Dr. Mark Perlin, when he was a faculty member at Carnegie Mellon University.
Perlin, a computer scientist, has refused defense attorneys’ requests for access to the program’s source code, claiming it’s a patented trade secret. Defense lawyers maintain that without access to the source code, they cannot check the program for errors to ensure its results are accurate. Dr. Perlin has countered by noting that seven scientific articles have been published verifying the accuracy of the program; however, he was the author or co-author of all those articles.
Law enforcement uses TrueAllele to test mixtures of DNA usually found at crime scenes. Conventional crime labs often refuse to calculate the probability that a suspect contributed DNA in weak and contaminated crime scene samples, rejecting test results below a certain threshold of reliability. TrueAllele uses all test results, regardless of how weak they are. This means there are large differences between the conclusions reached by TrueAllele compared to conventional DNA testing.
For example, after crime scene DNA was matched to a sample in a national database, convicted sex offender Martell Nathaniel Chubbs was charged with a 1977 murder in Long Beach, California. After testing vaginal swabs from the 17-year-old victim, private company Sorenson Forensics concluded that the DNA profile occurred in about one out of 10,000 African Americans. TrueAllele examined the same sample and said the likelihood it belonged to Chubbs was “1.62 quintillion times more probable than a coincidental match to an unrelated Black person.” Of course, that probability is greater than the planet’s entire population. Chubbs, 56, pleaded no contest to second-degree murder and second-degree burglary on March 18, 2016, and received a prison term of seven years and eight months.
In an Allegheny County, Pennsylvania double homicide case, the county crime lab found a mixture of several peoples’ DNA discovered on a bandanna left at the crime scene was too complex to analyze. However, TrueAllele had no problem finding a match to suspect Michael Robinson, which the company stated was “5.7 billion times more probable than a coincidental match to an unrelated black person.” Robinson, who had no prior criminal record, was charged with murder in 2013. Prosecutors initially sought the death penalty.
Attorney Noah Geary, who represented Robinson, filed a discovery motion for access to TrueAllele’s source code. Geary asserted that, under the Sixth Amendment, his client had the right to confront his accuser by “[challenging] the validity and reliability of the methodology” of the software, including its source code.
“They’re coming up with these astronomical, insane results,” said attorney Angelyn Gates, who represented Chubbs and had also challenged the use of TrueAllele’s test results. “When you put data into a computer and it spits out something, you’d like to know how it did it.”
An appellate court reversed the trial court’s exclusion of the TrueAllele evidence in the Chubbs case, and Gates appealed to the Supreme Court of California.
“TrueAllele is being used on the most dangerous, least information-rich samples you encounter, and typically in the most important cases,” observed Dan E. Krane, a Wright State University biology professor and defense expert who has sought to examine TrueAllele’s source code. “And I don’t know how it arrives at its answers.”
In February 2016, Pennsylvania Court of Common Pleas Judge Jill Rangos denied a discovery motion in the Robinson case, stating, “[t]he source code is not material to the defendant’s ability to pursue a defense.” See: Commonwealth of Pennsylvania v. Michael Robinson, Court of Common Pleas of Allegheny County (PA), CC 201307777.
Prior to that decision, both the state and Cybergenetics had argued against granting the motion, warning that disclosure of the program’s source code may cause financial harm to the company or compromise its ability to act as a witness for the prosecution.
Following Judge Rangos’ ruling, Robinson’s counsel requested that the Pennsylvania Superior Court review and overturn it. In their appeal – underscoring the potential human loss in such cases, as opposed to the potential pecuniary loss to Cybergenetics – Robinson’s defense team wrote, “without production and defense review of the computer instructions, not only will the petitioner be denied his constitutional right to a fair trial – he risks being wrongly executed.”
Recent forensic debacles include the discredited field of forensic bite mark analysis, doubts about forensic hair comparison, errors in the FBI’s national DNA database and crime lab scandals in numerous states. [See, e.g., PLN, Dec. 2016, p.22; April 2015, p.1; Oct. 2010, p.1]. These scandals have led courts to be more open to challenges to the once-unassailable field of forensic analysis. A similar controversy exists where DNA testing of mixed genetic material is concerned.
For decades, more traditional forensic laboratories have relied on Combined Probability of Inclusion (CPI) analysis of mixed-DNA evidence. Dr. Perlin and others have criticized CPI, stating that probabilistic genotyping tools that rely on the processing of complex data sets through use of advanced computer algorithms, such as those used by TrueAllele, are far superior because they eliminate human error. CPI, unlike probabilistic genotyping software, largely involves the work of trained human technicians.
Then again, anyone who has dealt with computer programs knows that they are seldom error-free. Yet with programs like TrueAllele and similar software, the results can literally be a matter of life or death in criminal prosecutions.
In September 2016, the White House issued a report from the President’s Council of Advisors on Science and Technology (PCAST). The report cast grave doubt over the efficacy of CPI analysis. While critical of CPI, the PCAST report also stated, “objective analysis of complex DNA mixtures with probabilistic genotyping software is promising, but has not yet been sufficiently and appropriately validated....”
Though Cybergenetics and prosecutors have consistently fought to keep the TrueAllele source code from third-party scrutiny – and, therefore, from independent validation – they also loudly criticized the PCAST report.
Following the release of the report, the National District Attorney’s Association called the PCAST findings “scientifically irresponsible.”
In an interview with Forensic Magazine, former Alameda County (California) District Attorney’s Office prosecutor Rockne Harmon took on an almost conspiratorial tone, saying that PCAST’s mild criticism of TrueAllele was intended to “throw defense attorneys another bone” and further undermine forensic science.
For his part, Dr. Perlin, in a September 2016 interview with Forensic Magazine, also seemed eager to err on the side of paranoia.
“[TrueAllele] fully solved DNA mixtures 10 years ago,” he said. “[The National Institute of Standards and Technology] pretends there is a problem in order to get funding. They are biased and lack expertise. We don’t need another government boondoggle.”
Yet regardless of the accuracy or inaccuracy of TrueAllele, in criminal cases the jury has the final word.
And on February 7, 2017, jurors acquitted Michael Robinson of murder, finding him not guilty of all charges – despite the TrueAllele DNA test results that implicated him by a factor of “5.7 billion.” Just prior to trial, the prosecution had withdrawn its notice that it would seek the death penalty. The jury rejected statements by a witness who claimed she saw Robinson commit the crime, whose testimony was impeached by Robinson’s defense team. His attorneys also attacked the TrueAllele evidence, saying it had a questionable basis and Dr. Perlin was basically a salesman who promoted the DNA testing program.
According to the Post-Gazette, a “juror who asked not to be named said after the trial that the jurors had a lot of unanswered questions about the case and that the prosecution did not present enough evidence to convince them that Mr. Robinson was the perpetrator. She noted, too, that the DNA on the bandanna was difficult for them, because the county crime lab said it could not analyze it, but Dr. Perlin said he could. Further, she said that the bandanna was not recovered until the day after the shooting, and that it was possible it had been there even before the shooting.”
While TrueAllele’s results failed to sway the jury in Robinson’s case, also consider the program can be a double-edged sword, in that it can be used to exonerate defendants if it excludes their DNA. In fact, the program was used to clear Darryl Pinkins, convicted of rape, who served almost 25 years in Indiana’s prison system before being exonerated and released based on DNA evidence in April 2016. It was the first time TrueAllele was used to prove a prisoner’s innocence. The test results also excluded Pinkins’ co-defendant, Roosevelt Glenn, who had previously been released on parole.
“This technology holds the key not just to answering complex DNA problems, but the literal key to freedom for men like Darryl Pinkins and Roosevelt Glenn,” said Greg Hampikian, director of the Idaho Innocence Project.
Assuming that the software program is accurate and reliable, that is – which is hard to determine when defense attorneys aren’t allowed access to its source code.
Sources: Wall Street Journal, www.arstechnica.com, www.post-gazette.com, www,cbsnews.com, www.forensicmag.com, www.huffingtonpost.com, www.cybgen.com, www.buzzfeed.com
Related legal case
Commonwealth of Pennsylvania v. Michael Robinson
|Cite||Court of Common Pleas of Allegheny County (PA), CC 201307777|