Bill Questions Proprietary Algorithms Used in Probabilistic Genotyping Software

  • <<
  • >>

 Bill Questions Proprietary Algorithms Used in Probabilistic Genotyping Software

A new bill brought forward by Rep. Mark Takano, D-Calif., is making the legality of using proprietary forensic software to unravel complex DNA mixtures just that—complex.

To either pass or fail the bill, there are essentially two questions the court needs to answer: 1) Are the source codes that power proprietary probabilistic genotyping software considered trade secrets? and 2) Does the public have a right to this source code to ensure equal and fair treatment?

Cybergenetics’ TrueAllele and STRmix Ltd.’s STRmix are the two prevailing, proprietary DNA mixture software platforms used in today’s forensic investigations. The courts have upheld the admissibility of TrueAllele and STRmix—and the idea that their source codes are trade secrets—numerous times in the past decade. Additionally, and most importantly, both companies make their source code, software, validation information and users’ manuals available to defense in confidentiality.

During a 2019 case in Georgia, for example, Judge Cynthia Adams ruled “evidence derived using TrueAllele is admissible under Harper, specifically finding that the TrueAllele probabilistic genotyping system has reached the stage of veritable certainty so that the testing and results ‘rest upon the laws of nature.’” Likewise, earlier this month when the defense counsel in a murder case in Arizona Superior Court challenged the admissibility of STRmix, the court concluded, “the use of probabilistic genotyping, and specifically the software utilized in this case– STRmix–has been generally accepted in the relevant scientific community of forensic DNA…Even the defense’s motion concedes that most state and federal courts have found that STRmix is scientifically valid because of it having been subjected to several peer-reviewed studies that produces favorable results.”

Both TrueAllele and STRmix have been validated and are in compliance with the FBI’s Scientific Working Group on DNA Analysis Methods (SWGDAM), which is an indication of robustness, accuracy and acceptance within the scientific community. Additionally, the algorithms for both software programs have been published in peer-reviewed literature.

“Experience shows that the SWGDAM guidelines are robust and, in our view, it would be challenging to develop and administer standards that improve on them,” STRmix representatives told Forensic.

Still, Takano and supporters of the bill argue the only way to know whether a platform’s findings are accurate enough to be used as evidence in court is to examine its source code for bias and inaccuracy—the same source codes that have been treated as trade secrets but are also available to defense under confidentiality.  

Mark Perlin, Ph.D., founder and Chief Scientific and Executive Officer of Cybergenetics, says the source code isn’t needed to test the accuracy of his TrueAllele platform, or any other forensic algorithm.

“An algorithm describes a procedure. A programmer writes in a computer language, translating the algorithm into source code text. A compiler turns the text into executable software that runs as a smartphone, laptop or other computer app. Algorithms are shared, software is tested. Since software pirates can easily copy text files, trade secret law protects source code confidentiality,” Perlin explains. “You don’t learn how a car works by reading its blueprints; you take it for a test run. Lawyers read, scientists test.”

 

In addition to prohibiting the categorization of probabilistic genotyping software’s source guide as trade secrets, HR 4368 asks NIST to establish the “Computational Forensic Algorithm Standards,” a program that will create and maintain standards for the development and use of computational forensic software. The bill would then require all federal law enforcement agencies and crime laboratories to use only software that has been tested and validated by NIST’s new program.

Traditionally, NIST has been non-regulatory, meaning it cannot enforce any policies on crime labs. This bill would change that slightly by requiring labs to use only software approved by the federal agency if the results are to be admissible in court. When reached by Forensic for a statement, NIST representatives said the federal agency does not comment on pending legislation.

Perlin told Forensic he is worried the bill will stifle forensic innovation by “revoking long-standing trade secret laws that protect visionary innovators.”

“Government protects trade secrets to promote innovation. Probabilistic genotyping companies give source code and testing software to defendants, and publish algorithms; these transparent forensic science advances benefit everyone,” Perlin said. “By eliminating liability for misusing proprietary information, the bill would encourage bad acts that could bankrupt innovators.”

Cybergenetics has entered trade secrets claims in 10 cases where TrueAllele’s findings were used as evidence. Thus far, the company has yet to lose. STRmix, on the other hand, has never filed a trade secrets claim.

To date, there have been 27 TrueAllele admissibility rulings in 14 states and federal court. The states include: California, Florida, Georgia, Indiana, Louisiana, Massachusetts, Nebraska, New York, Ohio, Pennsylvania, South Carolina, Tennessee, Virginia and Washington. Three appellate courts have affirmed (Nebraska, New York, and Pennsylvania). For STRmix, there have been 37 successful admissibility hearings in Colorado, Illinois, Wyoming, New York, New Mexico, Minnesota, Michigan, Connecticut, Florida, California, and Texas. TrueAllele reports have been used in 44 states for thousands of criminal cases—most not involving an admissibility challenge—and STRmix has been used to interpret DNA evidence in more than 160,000 cases around the world.

Related Categories