Curtis Rogers’s and John Olson’s database can identify 60 percent of Americans with Northern European ancestry. (Scott Dalton for The New York Times).
Lake Worth, Florida — On Halloween night in 1996, a man in a mask knocked on the door of a house in Martinez, California, handcuffed the woman who greeted him and raped her. Two weeks later, he called the dental office where she worked. Investigators tried to track him down through phone records, but got nowhere. They obtained traces of his semen, but there was no match for his DNA in any criminal database.
In September — two decades after the crime — the Sacramento district attorney’s office tried something new to finally crack the case of this serial rapist, who had attacked at least 10 women in their homes. Investigators converted the assailant’s DNA to the kind of profile that family history websites such as 23andMe are built on, and uploaded it to GEDmatch.com, a free site open to all and beloved by genealogists seeking to find biological relatives or to construct family trees.
Within five minutes of reviewing the results, the investigators had found a close relative among the million or so profiles in the database. Within two hours, they had a suspect, who was soon arrested: Roy Charles Waller, a safety specialist at the University of California, Berkeley.
The arrest marked the 15th time that GEDmatch had provided clues leading to a suspect in a murder or sexual assault case, starting with the arrest in April of Joseph James DeAngelo, a former police officer, for the rapes and murders committed across California in the 1970s and 1980s by the notorious Golden State Killer.
And no one has been more surprised than the two creators of GEDmatch — Curtis Rogers, 80, a retired businessman in Florida, and John Olson, 67, a transportation engineer from Texas. Their tiny outfit, which began as a side project, has unintentionally upended how investigators are trying to solve cold cases.
Within three years, the DNA of nearly every American of Northern European descent — the primary users of the site — will be identifiable through cousins in GEDmatch’s database, according to a study published on October 11 in the journal Science.
Since Joseph James DeAngelo’s arrest, law enforcement agencies have turned to GEDmatch to solve crimes. (Jason Henry for The New York Times)
«It’s kind of been a shock to all of us how these things developed», said Mr. Rogers, who was drawn to the research when he decided to investigate whether he was related to a man who came to America on the Mayflower in 1620.
Since Mr. DeAngelo’s arrest, law enforcement agencies across the United States have turned to the site to solve decades-old cold cases. Increasingly, it is being used in recent cases as well.
Initially, Mr. Rogers was outraged at how law enforcement was using his website, but he now feels proud. «Within a year I think it will be accepted», he said. Some genealogists find that notion problematic, given the many ethical and privacy issues that have emerged as investigators have come to rely on a privately owned family history site to solve crimes.
Mr. Rogers works out of GEDmatch’s headquarters, a small house in Lake Worth, Florida, and Mr. Olson, his business partner, works out of his home in Texas. Three retired computer scientists sometimes help remotely.
Roy Charles Waller, right, was charged in at least 10 rapes across California. (Renee C. Byer/The Sacramento Bee, via Associated Press)
The GEDmatch database can now be used to identify at least 60 percent of all Americans of European ancestry through their cousins, according to analyses by genetics researchers.
But GEDmatch has no lab. Rather, the site serves as a place where people, who have had their DNA analyzed elsewhere, can find more relatives and dive deeper into their ancestry. Some users are interested in finding long-lost cousins; others are retirees digging into old family mysteries or adoptees tracking down biological parents. Over 10,000 people likely have used the site over the past eight years, according to two genealogists who teach people how to conduct searches.
What it offers to researchers and criminal investigators is tremendous flexibility. There are now more than 17 million DNA profiles in genealogical databases, but most of the bigger sites restrict what can be uploaded, banning not only crime scene evidence but anything processed by an external lab. GEDmatch will take it all — blood processed by an obscure lab, spit processed by 23andMe — for free, so long as it’s in the right format.
The site is also useful for people building an extensive family history. The average person can find any number of cousins on existing genealogy sites. But the key, for a genetic sleuth, is figuring out precisely how those cousins are related to a person of interest, and to each other. The tools that Mr. Olson created enable users to see the precise genetic segments where cousins overlap. From the site’s one million or so profiles, a skilled genetic detective can often puzzle out an individual’s identity from a single third cousin match.
«There’s nothing else like it», said Barbara Rae-Venter, a genetic genealogist who used the site to help crack the Golden State Killer case.
Using GEDmatch in this way is not easy. Most investigators require the help of a skilled genetic genealogist.
On April 25, the Sacramento County district attorney’s office announced that there had been a breakthrough in the Golden State Killer case.
Mr. Rogers saw the news while watching TV in bed. «I’d never even heard of the Golden State Killer», he said.
But when one of the newscasters mentioned «a new form of DNA technology», he asked his wife, «Do you suppose I was involved?»
It seemed possible; around six months earlier, two companies involved in criminal investigations had asked for his blessing to use the site.
American law enforcement agencies have their own database for criminal investigations: Codis, which contains more than 16 million DNA profiles. But forensic profiles contain only a tiny fraction of the hundreds of thousands of genetic markers that genealogy sites rely on. If investigators are unable to find an exact match there, a site such as GEDmatch is better for tracking down suspects through their relatives.
The site’s privacy agreement had always been vague, essentially stating that its owners had no control over how any individual’s genetic or family tree data would be used. But explicitly sanctioning a law enforcement presence felt different.
«There’s probably no way I could stop you», Mr. Rogers said he told Parabon, a forensic consulting firm, and the DNA Doe Project, an organization focused on identifying bodies. «But we can’t give you permission. I have to protect the site».
Mr. Rogers was furious when he confirmed that a third set of investigators, without first telling him, had involved GEDmatch in the Golden State Killer case. It seemed inevitable that the news would drive thousands of people off the site.
While giving a tour of his home, Mr. Rogers handed over a stack of emails from that first week. On top, an expletive-filled note accused him of violating users’ privacy. But underneath it was email after email of congratulations.
Mr. Rogers and Mr. Olson had not expected such an outpouring of support. Neither did they anticipate 5,000 new uploads to the site shortly after Mr. DeAngelo’s arrest — a daily record, Mr. Olson said.
Two weeks later, Parabon announced that it was joining forces with CeCe Moore, a genetic genealogist, to use GEDmatch to solve crimes.
«This simply would not be possible without the courage of John and Curtis to allow law enforcement to use this database», said Ms. Moore, who has helped identify more than 12 suspects in murder and sexual assault cases using the site in the past five months.
As praise has flowed in, both men began to relax. By May, they had tweaked the privacy agreement to explicitly mention that users’ profiles might be used in a homicide or sexual assault investigation. By September, any lingering doubts they had were gone.
«I have absolutely no concerns that a person’s privacy is violated, because there are so many people whose DNA helped get to a capture», Mr. Rogers said.
But many observers disagree. When any one person’s DNA can lead investigators to hundreds of a suspect’s relatives, the standard model of individual consent does not hold up, said Rori Rohlfs, a professor at San Francisco State University. She finds it ironic that police in California must get approval from a judge to search criminal databases for a murder suspect’s brother, but can upload DNA to GEDmatch to identify cousins without any restrictions.
The excitement around cases that have not yet gone to trial also risks reinforcing the notion that a DNA match is proof of guilt, some researchers warn.
As more law enforcement agencies have begun experimenting with genetic genealogy, the GEDmatch database has grown by about 1,800 profiles every day, Mr. Olson said.
«I don’t like it, I don’t like it one bit», said Mr. Rogers’ wife, Janet Siegel Rogers, an artist.
«What don’t you like?» Mr. Rogers asked.
As it turned out, she didn’t like her husband’s email habits. «He’s at it 24-7», she said.
As she spoke, Mr. Rogers took a sip of wine.
From the trace of saliva he left on the glass, a skilled genetic genealogist could use the website he created to reveal his name and a dozen close relatives. It would also be possible to determine that while he is not related to any pilgrims, he is in fact the descendant of Vikings.