Industry News: NIST to Assess the Reliability of Forensic Methods for Analyzing DNA Mixtures

27 Nov 2017


The National Institute of Standards and Technology (NIST) will undertake a study to assess the reliability of forensic methods for analyzing DNA evidence that, if misapplied, could lead to innocent people being wrongly convicted. The study will focus on DNA mixtures involving three or more people, and on very small quantities of DNA also known as touch DNA.

The process of interpreting DNA mixtures and touch DNA can be subjective, and there are currently no clearly defined standards for doing so. The result: different analysts might come to different conclusions given the same evidence.

NIST Fellow and forensic DNA expert John Butler will lead an interdisciplinary team of scientists in conducting the study. The team includes Sheila Willis, the former director general of Ireland’s national forensic laboratory, Forensic Science Ireland (FSI), who NIST recently brought on to work on this project. The researchers will also seek input from outside experts.

Many people consider DNA analysis to be an especially reliable forensic method, and often, it is. If blood, semen or other biological evidence is left at a crime scene, forensic scientists can use it to produce a DNA profile—a sort of genetic fingerprint—that can reliably identify a suspect.

Rigorous scientific studies have shown that when the evidence contains DNA from only one or two people, DNA profiles are extremely reliable. But when the evidence includes a mixture of DNA from three or more people, it can be difficult to tease apart the different profiles, or in some cases, to even determine how many profiles are present.

In addition, DNA methods have gotten so sensitive that investigators no longer need a blood or semen stain to generate a DNA profile. Today, labs will sometimes attempt to generate a DNA profile from, for instance, a few skin cells left behind when someone touched something at a crime scene (thus the name, “touch DNA”). But when analysts dial up the sensitivity of their methods, the data can end up including meaningless information, or “noise,” that makes it difficult to interpret.

Many police agencies now routinely swab doorknobs and other surfaces for touch DNA when investigating property crimes. But if many people have touched those surfaces, the result may be a complex, low-level DNA mixture that is difficult, or impossible, to interpret reliably.

“Some labs won’t do anything with that kind of evidence,” said Butler. “Other labs will go too far in trying to interpret it.”

The goal of the study is to measure the reliability of DNA profiling methods when used with different types of DNA evidence such as two-person mixtures versus four-person mixtures, and with different quantities of touch DNA. Crime laboratories, the courts and other institutions can then use information from this study to decide which methods pass muster.

“The goal is not to undermine these methods, but to determine their bounds of reliability so they can be used appropriately,” Butler said.

While Willis is not a DNA expert, she is a chemist with deep experience in laboratory management. She said the study will be about more than just DNA technology. “We’ll also be looking at what quality control systems and training are needed to support the people who collect and analyze the evidence.”

NIST is calling this study a “scientific foundation review,” and it will be designed in part on recommendations in a landmark 2009 report from the National Academy of Sciences(link is external), which called for studies that demonstrate the validity of forensic methods.

While NIST’s first scientific foundation review will look at DNA methods, the researchers aim to build a framework that will be extendable to other forensic methods. NIST intends to begin a scientific foundation review of bite mark evidence later this year.

This research is part of a larger effort by NIST to strengthen forensic science through research and the promotion of technically sound forensic science standards.