New science is making things a little more complicated when it comes to DNA. One state is now looking at thousands of cases to make sure their analysis of identification probabilities was accurate for the last 15-plus years.

The Texas Science Forensic Commission sent a notice to its labs statewide in August that DNA mixture interpretation – involving genetic material from more than one person – is currently being reevaluated, based on improvements in DNA lab technology at the FBI.

“While in many cases the changed protocols may have no effect, it is also possible changes to results may be considered material by the criminal-justice system, either in terms of revisions to the population statistics associated with the case or to the determination of inclusion, exclusion or an inconclusive result,” the Commission wrote.

Some 24,000 cases involving DNA mixtures now have to be combed through, said Lynn Robitaille Garcia, the general counsel for the Commission, in an interview with Forensic Magazine last week.

“What we’re trying to do is understand as a state and as a commission where we think the CPI fell, in terms of statistical probabilities,” Garcia said. “There is not a cookbook of what should be done.”

The reevaluation began with a notification from the FBI early this year that they had corrected and updated their allele frequency data to all CODIS laboratories in the nation. However, those data brought an unexpected result: the Combined Probability of Inclusion and Combined Probability of Exclusion (CPI and CPE, respectively) could have changed – and changed significantly, in some cases.

This means that a 1-in-674,000 chance of identifying a suspect’s genetics could have fallen to less than 1-in-100, according to some reports.

Stephen Fienberg, a noted professor of statistics and social science at Carnegie Mellon University, told Forensic that the scientific approaches were still evolving – and each lab and case is inherently unique.

“My sense is that there really are multiple models for dealing with mixtures and they do yield potentially different results,” Fienberg wrote in an email. “So the mixture problem is not just an issue of what tests to do in a lab, and there is not a real consensus on the ‘best’ statistical approach.”

The inclusion/exclusion analysis models were based on a series of black-and-white decisions which could potentially have errors, said Bruce Weir, chairman of the Department of Biostatistics at the University of Washington. However, the latest computer analyses take into account more than one person – and instead have a series of probabilities – essentially a cumulative accounting of the shades of gray because of genetic pieces of information from more than one person, Weir told Forensic.

“Any calculations you do should take into account there’s more than one person there,” Weir said.

“It’s a matter of detail, rather than a fundamental shift,” he added.

In Texas, the state authorities are embracing the new science – and proactively trying to update all their protocols. The new technology allows a “better, crisper, clearer understanding,” Garcia explained.

The group of 24,000 cases could be significantly whittled down, since it includes every case involving mixtures of more than one person – and will also include situations where no charges were filed, Garcia said. But it also accounts for only nine of the 16 labs in the state accepting samples since 1999, she added.

“Isn’t the primary concern getting these probabilities right?” Garcia said. “We have some work to do.”