On an August day in Washington state, fields of crops stretch like seas until they bump up against 100-foot-high pines. Cars are rolling over the High Bridge as locals swim in the coursing stream below. Thirty years ago, this serene scene was likely the site of a grisly crime — part of a double murder that would make legal history.
Under the bridge, on Thanksgiving Day in 1987, pheasant hunters found the body of Jay Cook. The 20-year-old had been beaten around the head and strangled with twine tied to dog collars. A pack of Camel Lights had been stuffed down his throat.
Two days earlier, a passerby, more than an hour’s drive away had also found a body. It was Cook’s girlfriend — 18-year-old Tanya Van Cuylenborg.
Thursday on the PBS NewsHour, William Brangham meets people who have taken DNA tests and as a result been inadvertently involved in murder investigations because of genetic genealogy. Check your local listings for the time, and watch Part 1 here.
“Tanya’s body was found mostly nude on the side of the road,” said James Scharf, who was a Snohomish County patrolman at the time. “She’d been raped and she’d been shot in the head probably right there on the side of the road.”
The couple had come down from Victoria, Canada, to pick up machine parts for Cook’s dad, who owned a furnace company. The couple turned the errand into a road trip through Seattle, camping out in Cook’s family van, a bronze Ford.
The killer probably “parked under a bridge that he knew was a good out-of-the-way place where nobody would see the van,” Scharf told the PBS NewsHour. “That is where he would have killed Jay, and he would have come back and try to set Tanya at ease” before killing her, too.
The van was found in a parking lot nearby Van Cuylenborg’s corpse. There was semen on her body, as well as on the hem of her pants, which were left inside the van.
And that’s where the trail went cold for three decades. No eyewitnesses. No concrete leads.
At the time, there was no DNA database anywhere in the country to compare the semen and search for a potential culprit. But last year Scharf, now a detective in the Major Crimes Unit, heard about a forensic tool called genetic genealogy.
“Without genetic genealogy, this case never would have been solved,” Scharf said. “That’s how powerful genetic genealogy is in solving crimes.”
Genetic or forensic genealogy combines direct-to-consumer DNA tests — like those purchased through 23andMe or Ancestry.com — with the age-old hobby of tracing a family tree with public records, such as birth certificates and land deeds.
The technique relies on the simple principle that, if you go back far enough in history, everyone is related, and therefore has thousands of relatives. Assuming that an average family has 2 to 3 kids, then a typical person would — statistically speaking — have nearly 200 third cousins, 950 fourth cousins and 4,700 fifth cousins.
If a genealogist can find a cousin of an unknown suspect who has left behind DNA at a crime scene, then they can use old school family trees — sometimes literally drawn on paper or white boards — to track down the perp.
In June, a Snohomish County jury convicted 56-year-old William Earl Talbott II for the first-degree murder of Cook and Van Cuylenborg. The episode is historic because Talbott’s trial and conviction were the first made through genetic genealogy.
From a law enforcement perspective, the case for using genetic genealogy is strong. But experts are also flagging concerns about what the method means for people’s legal and DNA privacy.
Big companies like 23andMe, Ancestry.com and MyHeritage don’t allow police to rifle through their records, at least not without a subpoena or search warrant. So genetic genealogy for cases like Talbott’s rely on people who willingly upload their DNA profiles to more public databases — raising the specter of personal health information being exposed.
Genetic genealogy — in truth, any forensics dependent on DNA — can fall prey to the same human biases that plague other aspects of law enforcement. Close relatives or even non-relatives can be accused of the crime if care is not taken with how the genetic genealogy is interpreted.
It happened to Michael Usry.
Back in 2014, Usry, a filmmaker based in New Orleans, was visiting his parents’ house when he received a phone call from the police. Usry said the officers wanted to check his vehicle because it matched the description of a car involved in a hit-and-run. The cops swung by and he agreed to head down to the station.
“When I was walking out the door with them, I turned to the sergeant — who ended up being the sergeant of Idaho Falls Police Department — and I said, ‘So this is about a hit and run?’” Usry said. “And he goes, ‘Well, actually Mr. Usry, we want to talk to you about some other things as well.”
The “other thing” was Angie Dodge, a young woman who was raped and murdered in Idaho in 1996.
But genetic genealogy, which pointed to Usry, is not foolproof. Here’s how it can go wrong.
The types of genetic genealogy
There are three major types of genetic genealogy — mtDNA, Y-DNA and autosomal — depending on what type of DNA is used and what the search entails. MtDNA genealogy relies on mitochondrial DNA inherited from mothers, so it’s useful for tracing maternal heritage. On the flipside, Y-DNA refers to the Y-chromosome acquired by biological males from their fathers — so it can track paternal ancestors.
But when people mention genetic genealogy used in cases like the Golden State Killer or William Earl Talbott II, what they mean is a search that’s dependent on autosomal DNA.
Pretty much every human possesses 23 pairs of chromosomes, and autosomal DNA is the genetic material packed into 22 of them (the last pair consists of the sex chromosomes, which we’ll discuss in a second).
“At the end of 2009, 23andMe introduced the first test where we could use autosomal DNA for genealogy,” said CeCe Moore, a genetic genealogist with Parabon Labs who helped Scharf pinpoint William Earl Talbott II. “That’s the type of DNA that you inherit from all of your ancestral lines and it was immediately obvious to me that the potential for discovery was almost unlimited.”
That’s because those DNA tests read all of a person’s chromosomes like a biography. Everyone’s biography — DNA code or genome — is essentially the same, containing more than 3 billion letters in an order that is 99.6 to 99.8% identical.
But when a person sends a saliva sample to 23andMe or Ancestry.com, they receive back a profile that is mostly focused on 600,000 to 700,000 letters found in those 22 autosomal chromosomes. Those tests reveal subtle differences in those letters — called SNPs — that can explain why people have different hair colors and health conditions. Autosomal SNPs can also show who is related. Everyone carries about 12.5 percent of their great-grandparents’ autosomal DNA.
“We’re looking for people that share long stretches of identical DNA…. The more shared DNA, the closer that ancestor is in time,” Moore said. “If you share about 3 percent of your DNA with someone, then you’re most likely second cousins. If you share about 6 percent, then you’re most likely first cousins once removed. If you share about 12 percent, then you’re most likely first cousins.”
After Moore finds a cousin or cousins who match a suspect’s DNA, she will hunt down all the public records she can find related to their family — birth certificates, death certificates, family connections on social media. That allows her to reverse engineer a family tree — building up to a common ancestor, like a great-grandparent.
She’ll then flip the process — researching so she can build and climb down the family tree until she finds the subsection of the family that contains her unknown suspect. Moore said it’s impossible to predict how long it will take to zero in on a potential suspect. Depending on the size of the family, the process can take months and involve looking up thousands of people.
“The William Earl Talbot II case was the fastest. I zeroed in on him after only two hours,” said Moore.
But none of this can happen without the right DNA database.
Open season for GEDMatch
One common misperception about the rise of forensic genealogy is that 23andMe, Ancestry.com and other big direct-to-consumer companies are somehow involved — that those companies have ratted out their users by allowing law enforcement to dig through their DNA records.
The opposite is true. Most won’t work with law enforcement without a court order. Even if they did, a genealogy search would probably strike out.
That’s because so many of these companies exist. If you subscribe to 23andMe and commit a crime, there’s no guarantee that one of your cousins has done the same. Maybe they used Ancestry.com, MyHeritage or the dozens of other companies on the market.
That’s why GEDMatch was created — to serve as a meeting space for people interested in their ancestry, but who have struggled to find family members through their testing service of choice.
“We never advertised and promoted. This strictly grew by word of mouth and social media,” Curtis Rogers, one of GEDMatch’s founders, told NewsHour. GEDMatch headquarters is in Lake Worth, Florida, mere miles from the ocean and President Donald Trump’s Mar-a-Lago resort. The yellow, beachy bungalow with a giant red rosebush out front doesn’t scream “this is a digital hub of 1.25 million DNA profiles.”
Rogers, an 80-year-old, semi-retired businessman, co-founded GEDMatch because of his fascination with his ancestry that started as a teenager after his grandfather’s brother asked him to fill out a genealogy form.
“That’s how we used to do things in the old days,” Rogers said. “Slow-mail a form to fill out your parents’ names, when they were born, when your siblings were born, that type of thing.”
Fast-forward to the internet era, when Rogers was volunteering for an ancestry website called FamilyTreeDNA. The company asked him to take charge of a surname project that involved tracing the lineages of people named Rogers. “We’re sending emails to people, and it would take hours trying to figure out who the common ancestors were,” he said.
At the time, Rogers had a small personal website and a partner — John Olson — who was good with computers. One day, Rogers asked Olson if the latter could put together an algorithm for matching family trees via computers, rather than doing all the manual back and forth.
GEDMatch launched in 2010, allowing people to download their raw, autosomal DNA profiles from companies like 23andMe and then upload them into a public place. GEDMatch users can use an alias with their accounts, and a search will only yield the account name and contact info.
“People think that we’re showing DNA in our site. We don’t show anyone’s DNA. We show matches,” Rogers said.
That is great for genealogy hobbyists, but also for law enforcement — who have limited means of tracing family ties through DNA.
In 1990, Virginia became the first state to store the DNA records of convicted felons in a database — a move that would be replicated by the other 49 states before the end of the millennia. The Federal Bureau of Investigation eventually united all of those databases, in 1997, into one national network called the Combined DNA Index System, or CODIS.
But CODIS has privacy limitations that inherently prevent its use with genetic genealogy. Whereas a 23andMe-style DNA profile contains info on up to 700,000 spots in a DNA code, a CODIS profile has 20. Those 20 spots are in special zones called short tandem repeats — parts of the genome that don’t define a person’s health status or appearance. Just try writing a person’s biography with only 20 hard-to-decipher letters. Hence the allure of GEDMatch.
But the fact that law enforcement was actually using GEDMatch? “It was a shock to me,” Rogers said. He and Olson didn’t learn that their services were being used for culprits until investigators in the Golden State Killer case announced on April 25, 2018, that they had called on the website to apprehend Joseph James DeAngelo.
“[I] came to the office the next day and there were reporters,” Rogers said. “They drove from Miami…There were satellite trucks out here in this little narrow street; really upset the neighborhood.”
Rogers said he was initially confused, upset and worried that the investigators had violated his users’ privacy.
He said it took him about two weeks to really figure out how his site was being used and that it was not a violation of personal data.
He said their terms of service had always warned that information might be used for purposes outside of genealogy. After the Golden State Killer episode, they updated the terms to permit law enforcement searches in the cases of violent crime, which they defined as rapes and murders. And he disagrees with critics who claim such searches are detrimental.
“If they could see some of the emails from these families that have had some closure, I can’t imagine that anyone would say it’s the wrong thing to do,” Rogers said, adding that, if “you don’t want to get caught, don’t be a criminal.”
Slasher films and the case of mistaken identity
When Michael Usry, the filmmaker from New Orleans, was 19, he had traveled through Idaho Falls on a couple occasions around the same time as Angie Dodge’s murder — once for a Mormon mission trip and another time for a ski vacation in Utah. But in 2014, detectives were interested in Usry because of his dad’s DNA.
Sixteen years ago, Usry’s father had been convinced to submit a cheek swab to the Sorenson Database, as part of a genealogy project in which participation was encouraged by the Mormon church. This DNA database was public, but only contained profiles of Y-chromosomes and mitochondrial DNA — thus, it could only trace paternal or maternal heritage.
But police in Idaho Falls — which has a large Mormon population — saw an opening. When they ran semen from Dodge’s murder scene through the database, they found a close match — 34 of 35 letters — belonged to Usry’s father.
“They immediately eliminated my dad because he was too old,” Usry said.
“But when they started doing research, they saw my short film,” said Usry, who had co-produced a slasher film called “Murderabilia,” which includes the murder of a young girl. Usry said police grilled him for a couple hours before asking him for a cheek swab. “He goes, ‘Well, do you see this warrant right here? It’s signed by this judge and that means that you have to do this for us, right here and right now.’”
Usry was released, but he said he spent a worry-filled month before the police called to say his DNA wasn’t a match.
“There’s a long and documented history of the misuse of forensic evidence and criminal justice,” said Erin Murphy, a law professor at New York University. “It’s an important point to remember because DNA, too, can be misused.”
Here’s why that also applies to genetic genealogy.
Other ways genetic genealogy is fallible
Even though autosomal genealogy — with its 700,000 letters — offers a much more specific portrait of a person, it can still lead to false identifications. Ancestry tests can be misinterpreted, and a direct-to-consumer DNA profile can contain errors — typos in the book. A small study in 2018 found up to 40 percent of the SNPs identified in DNA profile might be false positives, a result mirrored by a second study published this June.
Moreover, autosomal genealogy cannot distinguish between siblings — because their DNA is too similar. If your brother or sister commits a crime, this brand of genetic genealogy can lead detectives to surveil you.
That’s because genetic genealogy is only useful for generating a paper lead — a family tree.
“Because I’ve worked in unknown parentage for so long, I’m acutely aware that sometimes people’s family trees are not what they appear on paper,” said CeCe Moore, referring to confounding factors like adoption. She said she tries to emphasize these caveats — possible siblings that can’t be found through public records — to law enforcement.
That’s why detectives still need a direct piece of a person’s DNA to confirm a match. In the Cook and Van Cuylenborg investigation, Scharf’s team surveilled Talbott for weeks until he discarded a coffee cup with his saliva, and other genetic genealogy cases investigated by their unit have involved tracking siblings.
Benjamin Berkman, deputy director of bioethics at the National Human Genome Research Institute, said sibling surveillance or cases like Usry’s may just be an unfortunate inconvenience of using a powerful tool.
“If we’re weighing convicting someone who’s committed a terrible crime against the idea that an innocent person might have to give a blood sample, that seems like a trade-off that most people would be willing to make,” Berkman said.
He is more concerned about another technique that is rising alongside of genetic genealogy called DNA phenotyping. Investigators claim that it can guess the appearance — eye color, hair color, skin color, face shape — of a suspect or Jane Doe based purely on their DNA.
“The science there is much less well-developed and we need to be very careful about relying on it to target people,” Berkman said. “There is a case in Germany where they thought that the DNA evidence from a murder scene pointed to a certain ethnic minority, and then [law enforcement] went and harassed that population.”
It turned out the sample was contaminated, and regardless, the accuracy of DNA phenotyping varies wildly depending on the trait. For example, it can predict brown and blue eyes with 90 percent accuracy, but it cannot pinpoint other colors like gray, said Amade M’charek, an anthropologist at the University of Amsterdam who specializes in forensics and race.
“For skin color, dark brown is very precise with DNA phenotyping; very pale is very precise. Everything in between is still not that precise,” M’charek added. Likewise, only 40 to 45 percent of facial structure can be precisely judged by DNA phenotyping. M’charek said much of DNA phenotyping still relies on human interpretation, which creates an opportunity for bias.
Given these pitfalls in genetic genealogy and phenotyping, Berkman, M’charek and Moore said everyone involved in these emerging arenas of DNA forensics — from detectives to genealogists — need to follow the highest standards when tracking down possible culprits.
Berkman added that police departments need to be more transparent about when and how they’re using genetic genealogy and other DNA information to prosecute someone.
“I would worry a little bit if genetic information was the only thing leading to prosecutions,” said Berkman. “We still want investigators to have to build a traditional case for prosecution beyond just the genetic evidence.”
William Brangham and Rhana Natour contributed to this report.