The 1998 murder of Renee Sweeney has haunted the small northern Ontario city of Sudbury for nearly two decades. On Monday, the Sudbury police finally released a photo of their suspect. With short-cropped dirty blonde hair and a boring, rectangular pair of glasses, he looks like any number of graduate students in the world.
But there's no guarantee that the murderer looks like this photo in reality. This is because it was prepared using only DNA samples found under Sweeney's fingernails and on a discarded jacket and pair of gloves by an emerging—and controversial, in a policing context—technique called DNA phenotyping.
The Sudbury police want people to keep an eye out for anyone who looks like the 3D visage in the digital image. Investigators are also using the snapshot internally to narrow down lists of suspects and prioritize tips.
"I definitely believe that it's accurate, and the science has proven that," said Det. Sgt. David Toffoli of the Sudbury police in a phone interview.
At Pennsylvania State University, Mark Shriver leads a lab that attempts to identify these superficial genes and model their effects on the face. According to Shriver, the technology presents a compelling area of academic work, but police relying on it to generate leads from the public and to narrow down suspects is fraught.
A phenotype is the expression of genes into characteristics like hair, eye, and skin colour. Since all humans are far more genetically similar than not, Shriver said, accurately predicting phenotypes comes down to a small number of genes—not an easy task. Moreover, appearance can be dramatically altered by a person's environment.
"People are so used to looking at faces—I don't want them to imagine that what they're seeing is the face of the perpetrator, or whoever was thought to leave that DNA sample," Shriver said over the phone. "It's really critical to be up-front with investigators about how you got that information and what it means."
There's also the risk that the emerging technology, in the hands of police, can misrepresent visible minorities due to potential shortfalls in the computer model making the phenotype predictions.
"We need more participants, and to sample more people from different populations," Shriver said. "That's going to be critical for prediction, because unless you over-sample for minority groups, you're going to over-generalize a minority individual."
Aside from the general state of the science, the Virginia-based company that did the phenotype prediction for the Sudbury police—Parabon Nanolabs—isn't very forthcoming about its methods.
This is despite numerous police forces in the US using the company's technology to generate suspect photos. In 2016, Parabon confirmed to the Toronto Star that police in Canada had used the technology but would not divulge where or when.
The company's webpage compares its 3D renderings with photos of real people to showcase how accurate its approach is, but Parabon hasn't published any research on their approach—a red flag for any company trading in emerging science.
When I contacted Parabon over email, I was linked to a scant document that Ellen Greytak, director of bioinformatics, described as a "scientific poster."
"Snapshot composites cannot be expected to represent a subject's exact appearance," Gerytak wrote in an emailed response. "Such environmental influences on an individual's appearance include, but are not limited to, age, weight, scars, exposure to smoking, exposure to sun, tattoos, hairstyles, and presence of facial hair."
In addition, all of Parabon's 3D composites are automatically given an age of 25 and an average BMI of 22. The algorithms are, of course, proprietary, Gerytak wrote, but a scientific paper is forthcoming.
Shriver helped Parabon secure a government grant for their work several years ago, but hasn't had contact with the company since, he said. At the time, he said that he and his colleagues were concerned that the company was moving too quickly to market with unproven technology.
As for the company's guidelines for police, what individual departments do with Parabon's product is up to them.
Perhaps the upcoming paper will assuage some critics' doubts, but it all raises the question: What exactly do police learn about a suspect from a DNA composite, and is it worth it—or hell, ethical—to ask the public to watch for people who resemble it? And can it actually help solve a nearly 20-year-old murder?
These questions notwithstanding, the image has been released to the public, so we might be about to find out.
Get six of our favorite Motherboard stories every day by signing up for our newsletter.