No witnesses. Leads that never panned out. Three decades searching for a suspect whose identity and appearance were completely unknown.
Some 32 years after Margit Schuller, 34, was found shot to death in the laundry room of the Palmetto Apartments in Beaufort, investigators were ready to try anything. Their one clue, a DNA sample left behind, seemed a dead end, with no matches found.
But investigators at the Beaufort County Sheriff’s Office believed that DNA sample could still tell a story. To use a technique called forensic DNA phenotyping, they contracted with a Virginia lab to give that unknown suspect a face.
The technique uses DNA to predict a person’s physical appearance, including eye color, hair color and skin color. It’s new and still developing, and not all of the scientific and legal communities are on board.
Skeptics say it’s unlikely that a full, accurate face can be predicted solely based on DNA, and critics worry the faces produced by this technology may prejudice investigations toward anyone in a minority community who vaguely resembles the rendered likeness.
For Maj. Bob Bromage, a cold case investigator in the Schuller case, what matters is the results. The department released the DNA-based composite sketch for the unknown suspect in mid-September, and it’s already generating new leads, he said.
The sketch shows what the suspect may have looked like in 1987 at an estimated 25 years of age, and how that person might have aged to today.
“It’s encouraging,” Bromage said. After 32 years, it gets less and less likely they’ll get an ID for their suspect, so any new information helps.
Keeping a cold case alive
The Columbia Police Department was the first department in the country to release a sketch based on DNA to the public. The department released a sketch in 2015, just as the technology was getting off the ground, in hopes of solving a 2011 cold case.
They arrested a suspect in 2017 after matching his DNA to a handprint found at the scene. The department declined to answer questions on how, or if, the DNA phenotyping assisted the investigation.
For Beaufort investigators, it’s a bit of an experiment — they may catch a suspect who looks nothing like the sketch, Bromage said. The true test will lie in being able to test a potential suspect’s DNA against the sample left at the 1987 crime scene.
Bromage wasn’t swayed by critics who worry the technology isn’t accurate enough. The science is valid, he said, and while it’s not good for all cases, it can open up new avenues of investigation that didn’t exist before. Most importantly, it can keep the case alive for family members hanging onto hope, he said.
The Beaufort County Sheriff’s Office worked with Parabon NanoLabs, based in Reston, Va., to create the composite sketch of their suspect. The company was founded in 1999 by Paula and Steve Armentrout as a software company, and they began working with law enforcement organizations on forensic DNA phenotyping in 2015. The company received funding from the Department of Defense to create the technique, Vice President Paula Armentrout said.
Their lab uses a predictive model for each feature of physical appearance. According to the company, they have a database with thousands of samples of phenotypes and the genotypes that create them, and they’ve looked at how genes interact to create certain phenotypes such as brown hair or blue eyes. They create a composite based on which traits are most likely, but acknowledge that it’s not 100 percent certain.
There’s been an uptick in interest for the technology, Paula Armentrout said, especially with the growing popularity of genetic genealogy and services like 23andMe that let people send in DNA tests to learn about their ancestry and other genetic information.
Parabon NanoLabs focuses on how that technology can be harnessed for police investigations, particularly cases running out of other options. They can take DNA samples and link a suspect to possible family members in existing databases, or create renderings of what that suspect may look like.
Genetic genealogy helped crack the “Golden State Killer” case in California, where a former cop was arrested in 2018 for a spree of killings and rapes between 1976 and 1986. Police used online DNA database GEDmatch to link the suspect’s DNA with that of distant relatives, eventually leading investigators to an arrest.
Like forensic DNA phenotyping, genetic genealogy has sparked concerns that the technology is moving ahead of people’s understanding on how such techniques should ethically be used. With genetic genealogy, those who submit their DNA to databases in hopes of tracing ancestry or finding relatives could unwittingly be identifying those relatives as potential suspects in police investigations.
And the success rate is growing. Since May of last year, Parabon NanoLabs has helped solve more than 69 cases, dozens of which involved only the forensic DNA phenotyping technique, said Paula Armentrout.
According to the company, an initial assessment with basic phenotyping for hair and eye color and genealogy information costs $1,500, while further forensic DNA phenotyping costs around $1,680.
The potential for misuse
Critics say that many of the composite sketches produced via DNA could cast suspicion on entire communities because they often resemble a generic black or Hispanic face. In Paula Armentrout’s perspective, her company does the opposite.
“We are giving law enforcement very objective info,” she said. Their technique is more reliable than eyewitness testimony, she added.
Legal experts have doubts on the ethics of using this technology in police investigations. Susan Dunn, legal director and interim executive director of the American Civil Liberties Union of South Carolina, said the ACLU believes it should only be used as a screening tool and nothing more.
“We really don’t think that this should be sent out to the public, because it’s not reliable enough for that,” Dunn said.
While the science has potential, it needs to become more accurate, Dunn said.
“It’s not the same as a DNA match. It’s a guess,” she said.
According to Dunn, the sketches she’s seen have been too generic and could easily make all members of a certain race a target because they “sort of look” like a sketch.
She acknowledged the usefulness of the technique for police to use internally to narrow their search, but she cited concerns that the technology could easily add to racial biases.
Genetic privacy may also be endangered by these new DNA advancements in forensics, she said. The ACLU worries that the technology will increase pressure on people to submit their DNA to databases.
Gray areas in DNA research
According to Steve Armentrout, president of the company and principal investigator for the lab’s phenotyping, the technique is intended to help police narrow down their fields of investigation.
“If you have a DNA sample but no match, you essentially have the entire world’s population as a suspect pool,” Steve Armentrout said.
Four years after they started marketing the technology, the lab is working on collecting additional data, increasing their database and enhancing their algorithms. They’re also trying to develop new phenotypes for features like hair curl and height.
Experts agree that accurately predicting sex, eye color and hair color is scientifically valid, but faces aren’t so easy. Steve Armentrout acknowledged that faces are harder to predict because they’re not on a single continuum like hair or eye color.
Researchers aren’t close to knowing all the genes that can shape a face, according to the latest studies published. There’s an additional complication in that genes alone don’t determine what a person looks like. A person can dye their hair, wear colored contacts, gain weight, or have distinct facial hair, among many other changes.
Many in the scientific community have questioned just how accurate forensic DNA phenotyping can be when so much remains unknown about how DNA shapes the face.
Researchers have called for Parabon NanoLabs to publish their methods in a peer-reviewed scientific journal, which they have so far declined to do. Steve Armentrout cited their proprietary algorithms as an obstacle in the way of publishing their work.
“No one really knows what their method is,” said Chris Korey, who teaches genetics at College of Charleston. He said without peer review of what data the company is using and how they get their results, there’s no way to tell if the technique is reliable or mere guesswork.
He doubts that scientists will be able to predict faces in the short or long term. Predicting faces for people not of Western European descent can be particularly difficult, as the current data available doesn’t include DNA sequencing for many non-Europeans.
Korey also worries that sketches from DNA phenotyping can give the impression that DNA is more powerful than it actually is. Without knowing how a person’s environment and random chance affected their physical appearance, there’s too much uncertainty to accurately predict what a person looks like, Korey said.
“DNA is not everything in determining who we are,” Korey said.
