The rapist terrorising women in the Gold Coast suburbs takes care not to be seen. Police believe that the same man is responsible for at least eight sexual assaults but because the attacks occur in the dead of night, in darkened bedrooms, his victims have been unable to provide enough details to produce a reliable eyewitness sketch.
To fill in the blanks, investigators have turned to a revolutionary new process: DNA phenotyping. Before, police needed a photograph, a video or a witness with a clear view and a good memory to put a face to a suspect. Now, according to US company Parabon NanoLabs, an accurate facial composite can be created from the genetic data contained in a smear of blood, a flake of skin or a drop of saliva.
"We absolutely stomp our foot that this is a likeness of the individual, and that it should not be treated as a photo ID," says Parabon's Chief Executive, Steven Armentrout, but the possibility of identifying violent offenders using the biological material they leave behind is alluring. Critics say the technology is being introduced before it can reliably fulfil that promise, with potentially dangerous results.
Based on unique DNA found at four of the crime scenes, the company's algorithm predicted the facial shape, eye, hair and skin colour of the Gold Coast rapist.
.jpg)
"We're using it to support and to validate what we've already got … to see whether the science matches up to what we're hearing from some of our victims," says Inspector Scott McLaren of the Queensland Police Service's DNA Management Section.
Investigating detectives aren't ready to release the image to the media. "We don't like to close people off by saying 'this person is a certain age, a certain build' because it does close their minds," says Acting Inspector Matthew Ward. "If we release something too early and close the community's minds to something we're looking for, it can damage us."
Eighty law enforcement agencies worldwide have commissioned a Snapshot from Parabon so far, mostly in cold cases. The technology has provided detectives with new leads and redirected investigations that had taken a wrong turn but the breakthrough case, in which a formerly faceless suspect is brought in looking just like his wanted poster, remains elusive. A triple murder case in Colorado provides a more detailed illustration of the technology's potential, and limitations.
In the early hours of January 16, 1984, a man armed with a knife and a hammer broke in to the Bennett home at 16387 East Centre Drive in Aurora. When her son didn't show up for work at the family-owned furniture store, Constance Bennett drove over to check up on him and was confronted by a vision of unimaginable horror.
.jpg)
Bruce Bennett had fought with the attacker on the stairs. His skull had been smashed in, his throat slit. His wife Debra had been stabbed to death, and their seven-year-old daughter Melissa had been raped and murdered. Their other daughter, three-year-old Vanessa, was still alive but her face and jaw were shattered.
Police connected the murders with another case. Six days previously, Patricia Smith had been sexually assaulted and killed with a blunt object at her house in Lakewood. Tests later revealed that the DNA found in semen at the two crime scenes was from the same man but his genetic profile was not in Colorado's database, nor the FBI's, nor Interpol's.
.jpg)
There had been other non-fatal attacks nearby where the weapon was a hammer but the victims could not describe their attacker. Donna Dixon had no memory of the assault after she woke from a coma. James Haubenschild didn't get a good look before he was knocked out cold, and his wife Kimberly could not provide enough detail for a forensic artist employed to produce an eyewitness sketch. Aurora Police believe one man was responsible for all four attacks. This August, 32 years after the murders, investigators finally released an image of the suspect.
Parabon provided Aurora police with two facial composites of the suspect: one at 25 years old and another at 55. The younger man's image is bland, like a head shot from a college yearbook. It's tempting to see something sinister in his piercing blue eyes and unsmiling countenance. The older looks like a henchman for the Kray twins in a BBC drama. Detective Steve Conner told reporters: "This is the first time we have had some idea of who we're looking for. He is no longer invisible."
Walsh is firm about what the technology can and can't do. Blue versus brown eyes is easy. Simple skin categories like pale, medium and dark can be predicted with 80-90% accuracy.
The Snapshot Prediction Report concluded the suspect was almost certainly of northern European descent. It predicted that his skin colour was fair. The company's statistical model was 92.7 per cent sure that his hair was not black, but whether it was brown or blond was harder to say. His eyes were not black or brown, but where they fell on the spectrum of blue and green could not be determined. He probably did not have freckles.
At US$3600 ($5000) a time plus associated laboratory costs, a Snapshot does not come cheap. Conner admits to the case being "an obsession" for him. "Snapshot is a technology of last resort, because we've used everything we have," he says. "I've ruled out, through this technology, how many millions of people? But I still haven't located the one that I need."
.jpg)
Forensic investigators have been determining ancestry from DNA for more than a decade. In one famous case police hunting a serial killer in Louisiana had narrowed their search to white men using eyewitness reports and psychological profiling, only to discover that their suspect's ancestry was 85 per cent sub-Saharan African. Derrick Todd Lee, a black man, was convicted of the crimes.
Bridget Algee-Hewitt, a forensic anthropologist at Stanford University, says DNA phenotyping has transformed her work identifying people who have died on the long border crossing from Mexico to the USA. Previously, she could only determine age, sex and height from the skeletons. Now she has ancestry, eye, hair and skin colour to work with, making a positive identification much more likely.
.jpg)
"In terms of technology and the value of it, I think it's huge," she says. "The more data you collect, the more DNA you analyse, the better statistics you develop, the higher the probabilities of success are going to be. They've already established that the technology works, that the estimation protocols make sense, so it's just a matter of collecting more material to refine that."
Mark Shriver, a professor of genetics and anthropology at Penn State University, is at the forefront of DNA phenotyping research. In his database he has biological samples and 3D facial scans of more than 10,000 people. He and his colleagues are studying around 10 million gene markers – Single Nucleotide Polymorphisms, or SNPs, known to geneticists as "snips" – to assess which contribute to shaping a face.
.jpg)
"There's at least 40 strong gene effects that we're observing within Europeans that explain a proportion of the facial variation," he says. "The expectation is we're going to be explaining more of the variation and making better predictive faces." Two papers are due early next year.
Assistant Professor Susan Walsh runs the DNA phenotyping laboratory at Indiana University. Her research with colleagues in the Netherlands and Poland has focused on predicting pigmentation from biological samples. So far they've published IrisPlex, which predicts eye colour, and HirisPlex, for hair colour. HirisPlexS, for skin colour, will be next.
Walsh is firm about what the technology can and can't do. Blue versus brown eyes is easy. Simple skin categories like pale, medium and dark can be predicted with 80-90 per cent accuracy. She hopes that predicting "continuous colour," like the shades on a paint chart, will eventually be possible for eyes, hair and skin, but it's not there yet, and as for mug shots …
"Right now there are individuals that promote results to police that they can produce facial images. That is way too far ahead," she says. "Pigmentation is doable. Face is not. We are certainly not anywhere near these outputs that they are giving."
Shriver, who has done consultancy work for Parabon but no longer has a relationship with the company, is more cautious in his criticism. "I do have reservations. Science requires you present your results to a group of peers first, publicly, so that everybody can critically evaluate them … I haven't seen that for any of the commercial presentations of these sorts of predictions."
"Parabon don't tell people what they're doing, which in this field is bad," says Walsh, flatly.
When I asked Armentrout how many subjects are in the company's database, he told me: "We collect samples from a variety of sources, some of which are in the literature and some of which we've collected ourselves. I'll leave it at that." Could he give me a ballpark figure? "I don't wish to."
In a follow-up email, he drew my attention to a blind evaluation of Snapshot carried out by Bruce Budowle at the University of North Texas that was presented at the International Symposium on Human Identification in September. It would soon be submitted to peer-reviewed journals, he added. Budowle did not respond to Fairfax' Medias request for comment.
In the evaluation, Parabon's algorithms got the eye and skin colour "right" for 24 of 25 test subjects, and hair colour right for 23. Some of the composite images presented alongside photographs of the volunteers are impressively accurate, but only five (presumably the best) have been released, and crucially, Parabon was provided with age and body mass data: information that police will not have unless an eyewitness gets a really good look.
In the weeks after Aurora Police released Snapshot's facial composite of their suspect, Detective Conner got "a boatload of calls" saying "this looks like a guy that I know," but some were cranks, some sent photos of black and Hispanic men, others of men who were living thousands of miles from Colorado in 1984. None have turned out to be promising leads yet.
"I don't know what algorithm they use to obtain that physical appearance, but I don't think that the resources are there," Conner says. "Maybe in five years from now, they can refine the features a little bit."
His man could be bald by now, or obese, or wasted by a long-term drug habit. He could be short or tall: geneticists have speculated that as many as 100,000 "snips" may be responsible for variations in height, to say nothing of environmental factors such as diet. Knowing his ancestry, eye, skin and hair colour is useful, but it's hardly the conclusive breakthrough that many people expect when they hear "DNA forensics".
Ward, the detective hunting the Gold Coast rapist, offers a cautious assessment of the technology's value. "The images that we're talking about are an interpretation, in my view, of the certain evidence that we obtained with the DNA ... It's given us an ability to exclude certain features… and that in itself is a productive measure within the investigation."
Forensic criminology is in retreat from an era of misplaced certainty. In 2009, the National Academy of Sciences (NAS) released a withering assessment of supposedly scientific analysis that had been relied on in court for decades. Expert witnesses had been telling juries that bite marks, hair and fibre matches, burn patterns, fingerprints and marks on bullet cases offered proof of guilt. The NAS concluded they proved no such thing: the best that analysts could do was suggest the probability of a match.
DNA phenotyping is unlikely to be relied on in court any time soon, not least because when police catch a suspect using Snapshot, they by definition have a biological sample linking him or her to the crime scene. But it does come with ethical risks, of racial profiling and innocent people brought in for questioning. Belgium and Germany do not allow police to use the technology.
"When people see something so seemingly real; an image that looks almost photographic, and they're told that it's generated from DNA, that has more credence than a normal police sketch," says Stanford University Anthropology Professor Duana Fullwiley.
In an experiment run by psychologists at Harvard University, white and Asian students were asked to read a "crime alert" describing a violent robbery. In half the reports, the suspect was black; in the other half, white. Those who received the black suspect alert were subsequently much more likely to see blacks as dangerous potential criminals. The same blanket prejudice towards whites was not observed among volunteers who received a white suspect alert.
"The Snapshot picture looks realistic, but it's also quite generic: it looks like a lot of people," Fullwiley points out. "If ever we have a success story with DNA phenotyping, we're going to hear all about it. We're going to see the original sketch, we're going to see what the person looks like. We're not hearing about all the negative cases and their stories."
In Queensland, police have been sitting on the image provided by Parabon for weeks.
They'll only release it to the media once the rapist's victims have had a chance to assess it, and then only if detectives are confident it is a good likeness.
"If we put an image of a 25-year-old out and the person turns out to be a 45-year-old, it does damage us. We'll have to exhaust our inquiries, and then explore that option of how the community can assist us further once we take that to the media," says Ward.
McLaren has been contacted by two other Australian police forces enquiring about DNA phenotyping, but can't reveal which ones.
"It's very much a softly-softly approach … The reality is that a lot of the jurisdictions will wait for a success in the utilisation of this technology," he says, a point Ward underscores.
"It's never been used or tested in Australia. I think my confidence will come about when an offender is apprehended and we can look at a face and compare it to the image that's provided to us," he says.
Whether or not it turns out to be instrumental in bringing the rapist to justice, he is confident that DNA phenotyping is here to stay. "Queensland Police, we pride ourselves that we're at the forefront of forensic science, so we are continually looking for this stuff," he says. "This technology's in its infancy. Other cases in Australia will use it, without a doubt."
with Chris Vedelago
DNA phenotyping in Australia
Australian Federal Police: The AFP is examining this new technology however it is not currently being used in active cases.
Victoria: is aware of technology to "provide useful information" on ancestry, hair colour and eye colour from DNA. Researchers at Victoria Police Forensic Services Department are working with other institutes to consider and validate methodologies able to provide this type of information for investigative purposes but it is not currently used for casework.
NSW: The NSW Police Force Forensic Services Group is not considering the use of DNA phenotyping at this stage.
Tasmania: The technique is not currently being used "however, we continue to monitor technology developments in other states".
NT: Not used "at this stage". "It is anticipated if this is required we would use facilities available through the Australian Federal Police".
WA: WA Police have not used the technology and could not confirm if they were considering it.
Chris Vedelago