Forensic genealogy helped nab the Golden State Killer in 2018. Now investigators across the country are using it to revisit hundreds of unsolved crimes.
In October 2016, the remains of three murder victims, dead for three decades, were laid to rest in Newton County, a rural corner of Indiana.
Two were young men, likely teenagers, the victims of a serial killer in 1983. The third was a woman found dead in 1988 on the bank of a creek. She had been shot in the head, covered with car tires and lit on fire.
Their bones, stored in tattered cardboard boxes and black trash bags, had been passed down from one county coroner to the next. When Scott McCord took the job in 2009, he gave the remains names: Adam, Brad and Charlene. He ordered anthropological and dental analyses, facial sketches and DNA tests in an effort to find their true identities.
Nothing panned out. So McCord gave Newton County, a community of about 14,000, a chance to mourn their "kids," as he called them. He paid for three small coffins, and a local florist donated flowers. Nineteen high school students volunteered to be pallbearers. After a ceremony at a county building, the teens piled into a yellow school bus and McCord, a part-time bus driver, followed three hearses, each donated from a different funeral home, to the cemetery.
"I didn't think we'd ever see resolution to any of the cases," McCord said.
As it turned out, he was wrong.
In late 2019, McCord, his deputy and a county prosecutor decided to try a complicated forensic technique that had nabbed the infamous Golden State Killer a year earlier. The effort turned into a yearlong crash course in a niche area of science: using genetic markers to build multigenerational family trees. It would require a million-dollar DNA sequencing machine, a custom-built computer in Texas and the utmost patience of volunteer genealogy buffs.
McCord and his team are among a growing number of investigators that have joined the scientific vanguard to revive cold cases. Likely hundreds of cases, of both victims and perpetrators, have been solved. Some have involved extracting DNA from decades-old bones, hair or minute traces of skin cells. Others have benefited from the most comprehensive and expensive type of DNA testing, known as whole-genome sequencing. In turn, a cottage industry has emerged to help.
Critics worry that the widening use of this investigational method could lead to what is essentially a national DNA database for law enforcement, giving police access to highly personal information from a wide swath of the public without their explicit consent. The only significant limit is the cost — typically several thousand dollars per case — and that is dropping rapidly, as demand surges.
"My county was more than happy to pay," McCord said.
Everyone and their grandmother
Genetic genealogy debuted more than 20 years ago as a pastime for ancestry enthusiasts. The customer sent a saliva sample to a company like FamilyTreeDNA and could then log in to a website showing how closely their genetic markers matched with those of other people — long-lost relatives — in the company's databases.
Margaret Press, a software developer and mystery writer, had used the method for years to help adoptees find their biological parents. In 2017, while reading a novel based on an unsolved murder, she realized that her skills might be equally useful to law enforcement. "It just hit me," she said. "The same technique that we were using for adoptee searches and finding unknown parents was adaptable to bones and unidentified remains."
She co-founded a nonprofit, called the DNA Doe Project, to try to match unidentified remains with genetic profiles that had been uploaded to an open-source genealogy database called GEDMatch. A set of unknown remains might match to a distant, known cousin, for example. An investigator could then build out a large family tree, first identifying the ancestors the two cousins have in common, such as great- or great-great-grandparents, and then investigating individual branches from those ancestors. Only some of those people would have lived in the right place and time as the unknown victim. More research, and the process of elimination, could ultimately reveal the Doe's identity.
On April 10, 2018, the DNA Doe Project announced that it had used this method to positively identify a 21-year-old woman, previously known only as the "Buckskin Girl," who had been found strangled in Ohio in 1981; it was the first time the public learned about a cold case solved with genetic genealogy. Two weeks later, police officers in California arrested Joseph James DeAngelo, the so-called Golden State Killer who had murdered 13 people and raped dozens in the 1970s and '80s. The technique employed was again the same, except this time the genetic sample came from the unknown culprit — a trace of semen left at a crime scene — not from a victim's remains.
That case made headlines for weeks, and "the floodgates opened," Press said. Investigators across the country were eager to try the technique on their own cold cases, and an industry sprang up to help. Parabon NanoLabs, a forensics company based in Reston, Virginia, was an early pioneer, and has since worked on more than 550 cases for law enforcement.
The Golden State Killer and many other cases relied upon a DNA test called a microarray, which generates a subset of key markers from a person's DNA code, like an abridged version of a book. But increasingly, investigators are turning to private companies for whole-genome sequencing, which reconstructs a person's entire DNA code. This more sensitive test is often best for old and degraded DNA, such as from skeletal remains heavily contaminated with bacteria. HudsonAlpha Discovery, a lab in Alabama, has worked on about 1,100 forensic cases. Astrea Forensics, in California, began as part of an academic paleogenetics lab, and Othram, in Texas, has worked on hundreds of cases and raised more than US$10 million in venture capital.
"It really has exploded," Press said. "Everyone and their grandmother is now setting up shop."
As a result, many cold cases have become quicker and cheaper to resolve. In the Golden State Killer case, six investigators worked full time for four months to narrow in on the culprit. Now cases are often solved in weeks or days.
The DNA Doe Project, run by Press and dozens of volunteers, has taken on about 120 cases since 2017, and has fielded inquiries from another 200 or so. Many were from small sheriff's or coroner's offices with few resources, Press said. They are often just as skeptical of the technique, she said, as they are of the psychics who frequently call with tips on open cases: "For many of them it's in the same category — do I call the psychic back or the genealogist back?"
Testing Charlene
McCord and his team in Newton County had not considered genetic genealogy until Rebecca Goddard, the county's chief deputy prosecutor, heard about the DNA Doe Project on a podcast called "Crime Junkie." "I understood very little," Goddard said. "I just sort of understood the concept of using ancestry to create a family tree."
McCord reached out to the DNA Doe Project for help, and a collaboration began. Initially, his team divided up the three open cases. As a prosecutor, Goddard took the only one that was still an open homicide: Charlene Doe, whose charred remains were found by a fisherman in 1988. McCord took Adam Doe, and his deputy (and girlfriend), Heidi Cobleigh, took Brad Doe. A serial killer, Larry Eyler, had confessed to both boys' murders years earlier but had not known their names.
The plan was for all three investigators to take an online course in genetic genealogy, which would walk them through a mock case. But McCord and Cobleigh quickly realized that they were not cut out for the work. It involved hour upon hour of computer searches for obituaries, family trees and other public records in order to deduce the precise familial relationships of people with shared genetic markers.
"We sat down and tried to do it and it was just overwhelming," McCord said. "It made zero sense to me."
But Goddard immediately took to it, and dove into Charlene's case.
Back in 2010, McCord had commissioned a Texas lab to run a DNA test from one of Charlene's teeth. That test resulted in just a couple of dozen genetic markers — too sparse to be compatible with genealogy databases. Luckily the lab, which runs the National Missing and Unidentified Persons System, had kept the remaining DNA sample in storage.
So now, a decade later, the DNA could be retested with the far more sophisticated whole-genome technology. There was just one problem: The sample contained only 0.3 nanograms, or 300 trillionths of a gram, of genetic material.
"When we heard how much it was, we were definitely like, well, not sure if this will work but we'll give it a shot," said Kevin Lord, a bioinformatician and private investigator in Belton, Texas, and the lab liaison for the DNA Doe Project.
Charlene's sample, in a tiny plastic tube, was wrapped in foam and cold packs, and shipped overnight to HudsonAlpha's lab in Huntsville, Alabama. There, the DNA went into a NovaSeq 6000, a million-dollar machine resembling a photocopier.
Human DNA is made of 3 billion "base pairs," a unique set of molecular letters that creates a book of instructions for our cells. At HudsonAlpha, Charlene's DNA was split into tiny fragments, then placed in the sequencing machine to decode the precise sequence of letters in each. The machine then created a file containing all of these tiny sequences and uploaded it to a server.
Lord downloaded that file — 71 gigabytes of raw data — onto a custom-built computer and began piecing all of those fragments back together in the right order, producing Charlene's full genetic code. From there, he focused on a subset of roughly 1.5 million key markers, known as SNPs, that make up the genetic "profiles" used by genealogy companies. He then sent Charlene's profile to the two genealogy databases, GEDMatch and FamilyTreeDNA, that make themselves available to law enforcement for such cases.
Then came the hard part.
Database hunting
Goddard was determined to hone her new sleuthing skills on Charlene's case. Last May, she began training with L. Elias Chan, a volunteer at the DNA Doe Project who runs a genealogy research business in Seattle.
Each database produced a list of people who were genetically related to Charlene, sorted by the percentage of their shared DNA. If a match shared 50 per cent of her DNA, for example, it could be a parent, child or full sibling. At lower percentages, the familial possibilities grow; a 3 per cent match could be a second cousin, a first cousin twice removed, a half first cousin once removed or a befuddling array of other relationships.
Goddard and Chan met via video chat, sometimes several times a week, to work through Charlene's matches. Chan would assign Goddard an intriguing match, and then suggest ways to track it down. "If she got stuck, she'd send me a message, like, What do I do here?" Chan recalled.
Success in such cases often depends on the size of the database: The larger it is, the more likely it will produce a match. The GEDMatch and FamilyTreeDNA databases are enormous, including millions of people who have voluntarily uploaded their own genetic information. (Ancestry.com and 23andMe also have huge databases but have opted not to make them accessible to law enforcement.)
GEDMatch and FamilyTreeDNA have not always been transparent with users that the DNA profiles might be scrutinised by the police. In 2019, under public pressure, GEDMatch abruptly changed its access policy, requiring users to explicitly "opt in" to law enforcement searches. Overnight, investigators lost a raft of promising leads.
That policy was in effect when Lord uploaded Charlene's DNA profile to the site, in May 2020. The only matches were to distant cousins, which Goddard and Chan dutifully worked through for more than seven months. (To protect the privacy of the victim's family, investigators have not disclosed the precise familial relationships discovered during the search.)
"We were all well in the throes of the Covid pandemic, and so I was spending a lot of time at home," Goddard said. "And I remember feeling like I had spent countless hours, but wasn't a whole lot closer, after all that time, than we had been after the first week or two."
Goddard caught a break in January, when GEDMatch changed its terms of service again, allowing its entire database to be used for forensic searches related to unidentified remains.
Chan called Goddard with some good news: Charlene now had multiple close matches at the top of her list.
True names
It didn't take long to deduce Charlene's real name: Jenifer Noreen Denton. In 1988, when Denton was 24, she suddenly went missing from her Illinois home, leaving behind her belongings and a 1-year-old daughter.
In February, Goddard talked to Denton's daughter, now grown, for the first time. "I just felt extremely emotional about it," Goddard recalled. Over her months of research, she had often imagined what it must have been like for Charlene's family. "There were people out there who surely loved her and were missing her and were probably desperate for answers." (Denton's family declined, through an intermediary, to be interviewed.)
Denton's daughter agreed to a DNA test for confirmation. It came back as a strong match with Charlene Doe. "That's when I felt comfortable breathing again," McCord said.
He is now bullish about the technology's potential to solve thousands of other cases, especially his own.
In February, Brad Doe's DNA was sequenced by HudsonAlpha, and on April 2, Lord uploaded it to GEDMatch. At the top of the list of matches were several close relatives. Three weeks later, and nearly 38 years after the boy's murder, the team announced Brad's real identity: John Ingram Brandenburg Jr., a 19-year-old from Chicago who went to a friend's house one day and never returned.
In 1983, after being tipped off by mushroom hunters, police found his remains with those of Adam Doe and two other boys near an abandoned barn off US Route 41 in Newton County. All were victims of Eyler, also known as the Highway Killer, who murdered at least 21 young men and died in prison in 1994.
McCord said that the teenager's mother was distraught to learn how he had died. "She woke up every morning for the past 37 years wondering, 'Is Johnny going to come home today?'" he said. "We hit her kind of hard." (The family declined, through an intermediary, to be interviewed.)
The Newton County team is now focused on solving their last case, of Adam Doe. His long-stored genetic sample was sequenced at HudsonAlpha and uploaded to GEDMatch, but was not robust enough to return any good matches. The team is planning to enlist a lab specialising in ancient organisms to extract a new sample of DNA from the bones.
And although Charlene now has a real name, her 1988 murder case remains open. Her remains had been discovered with those of a man who was identified a year later by police. "There were some leads back in the day," McCord said. "Now they can go back and hopefully confirm some of those leads, and we can get this resolved fairly quick."
The coroner, who organised Denton and Brandenburg's first funerals, now plans to attend their final memorial services. Brandenburg's will be held this month.
"That was a specific request from his mother that I be there because she said I've known Johnny almost as long as she has," McCord said. "When she said that, it literally brought me to tears."
Written by: Virginia Hughes
Photographs by: Kaiti Sullivan, Jason Henry, Wes Frazer and Jovelle Tamayo
© 2021 THE NEW YORK TIMES
