Back to Presentations

On the threshold of injustice: manipulating DNA evidence

M.W. Perlin, W.P. Allan and C. Hughes, "On the threshold of injustice: manipulating DNA evidence", American Academy of Forensic Sciences 69th Annual Meeting, New Orleans, LA, 17-Feb-2017.

PowerPoint presentation with studio recording of Dr. Perlin's talk.

PDFDownload Handout
PPTDownload PowerPoint


After attending this presentation, attendees will understand how DNA evidence can be manipulated by adjusting thresholds and other subjective data interventions.

This presentation will impact the forensic community and criminal justice by showing how DNA data can be subtly adjusted to achieve a biased result. In the provided case example, exclusionary evidence was used to produce an inclusionary match statistic.

Fingernail DNA evidence was obtained from a homicide Victim. Short tandem report (STR) laboratory analysis showed a possible DNA mixture containing the Victim, and perhaps a second person. Working for the prosecution, an Operator used his probabilistic genotyping Software program to compare this possible mixture with a Defendant.

Reliability Rule 702 of the Federal Rules of Evidence (FRE) for the admissibility of expert testimony requires sufficient data, a reliable method, and that the method be reliably applied to the data.

Sufficient data. The fingernail data contained potentially exculpatory DNA peaks between 30 and 50 rfu that did not come from the Defendant. According to the Software, the mixture ratio was around 250:1.

Reliable method. The Software's validation studies have been published using a data threshold of 30 rfu. None of these studies used a 50 rfu threshold that discards more DNA peaks.

The Software's validation studies have gone down to a mixture ratio of 25:1. No published studies have examined ratios as low as 250:1, the level of a few cells.

At a ratio of 250:1, the minor contributor may be too low to detect. Studies by the developer show that the Software is unreliable at this level. The program cannot accurately distinguish between true and false matches. It can give a positive association, whether or not a person's DNA is actually present in the mixture.

Reliably applying method to data. The expert Operator found a 250:1 mixture ratio. Such a tiny purported minor component may not even be from a real person. The Software was only validated for 25:1 mixtures, not for a 250:1 ratio. Applying an unreliable method to insufficient data is not reliable.

The Operator chose a threshold of 50 rfu. But the fingernail evidence contained potentially exculpatory evidence between 30 and 50 rfu. And the Software was validated for using more peaks at 30 rfu, not fewer at the higher 50 rfu level. Applying an unreliable method to insufficient data is not reliable.

In fact, running the Software at a validated 30 rfu threshold would exclude the Defendant. The fingernail evidence is exculpatory. The Software proves that the Defendant's DNA is not present.

A Software operator can subjectively choose their data to get the answer they want. Running the validated Software program on all the relevant data excludes the Defendant, and helps prove his innocence. The Operator's invalid data choices give a scientifically baseless answer that could lead prosecutors to a wrongful conviction.

Relevance Rule 403 balances the probative value of evidence against its prejudicial nature.

Probative. Ignoring crucial exculpatory evidence (i.e., data peaks between 30 and 50 rfu) does not prove anything. Running an unreliable computer program (i.e., unvalidated at 50 rfu, or for 250:1 mixture ratios) does not prove anything. The prosecution's expert ran an unreliable method on insufficient data to report a nonscientific result. When there is no science behind an expert’s opinion, it has no probative value.

Prejudicial. DNA is highly prejudicial evidence. Telling a juror that DNA connects important evidence to a defendant strongly conveys guilt. If that DNA connection is wrong, no amount of good science to the contrary can unring the bell. The jury should not hear flawed DNA evidence that can unfairly harm a defendant and irreversibly taint a trial.

Prosecutors can present DNA to jurors as infallible evidence. But expertly manipulating the data can alter its interpretation, and give subjective or inaccurate results. Most statistical DNA software lets the user choose their data. This case study points out the danger of such practices.