This Professor Found a Faster Way to Crack Cold Cases

Barbara Rae-Venter had unmasked the Golden State Killer, but now she was stumped.

An investigative genetic genealogist, she helped pioneer a technique that combines DNA analysis with family-tree research to identify criminal suspects. In 2018, after she received the infamous serial killer’s genetic profile, it took her just 63 days to identify Joseph James DeAngelo, who later admitted to committing dozens of rapes and murdering 13 people.

Investigative genetic genealogy is a powerful tool for solving crimes, but it can be painstakingly complex. In December 2024, Rae-Venter and her team were stuck on a cold case involving a brutal sexual assault in Cleveland nearly three decades earlier.

The suspect’s DNA did not correspond to any profiles in a national law enforcement database. Using genetic genealogy websites, Rae-Venter and her team built a family tree of the suspect’s potential distant relatives, but none were in Ohio, where she believed he lived. Not many people in the databases shared DNA with the suspect, and those who might were unwilling to provide genetic samples. After almost four years of research, she recalls, “We couldn’t figure out what to do next.”

Then Rae-Venter got an email from Lawrence Wein, professor emeritus of operations, information, and technology at Stanford Graduate School of Business. He told her he’d developed an interactive algorithm that could speed up her research. The first case she sent to Wein was the Cleveland assault. His algorithm told her to look further back in the family tree for a relative who might lead her to the suspect’s identity.

“I don’t know what the magic fairy dust is that tells him where we need to go,” Rae-Venter says. “Looking at the data, it was not at all obvious to do what he suggested.”

Eventually, the cues from Wein’s algorithm led Rae-Venter to a couple whose only son had lived in the Cleveland area until his death. Law enforcement confirmed he was the perpetrator through a DNA sample from a close relative. “When someone is deceased, that’s the end of the trail. But for survivors, naming him means he is no longer a boogeyman,” Rae-Venter says.

With the help from Wein’s algorithm, her team had cracked the case in 10 days.

A High-Stakes Math Problem

Police departments and other law enforcement agencies began working with investigative genetic genealogists to identify suspects in violent crimes about eight years ago, following the Golden State Killer breakthrough.

The process typically begins when a law enforcement agency runs an unknown suspect’s DNA through the Combined DNA Index System, a national database that includes DNA profiles of convicted felons and missing people. If no identical matches are found, agencies may choose to work with an investigative genetic genealogist. A detailed genetic profile generated from crime-scene DNA is uploaded to GEDmatch or FamilyTreeDNA, commercial genealogy sites where users share their DNA profiles to find relatives. (Users must consent to share their data with law enforcement.)

Once there’s a list of people who share DNA with the suspect, the law enforcement agency enlists a genealogist to build out a family tree the old-fashioned way, hunting down records for births, marriages, deaths, and other personal details. Their goal is to identify the most recent common ancestor of genetic matches from either the mother’s or father’s side. They then try to identify this ancestor’s descendants until they land on a couple that connects both sides of the tree.

From there, they follow the branch to the suspect’s parents. At that point, law enforcement tries to confirm the suspect’s identity, often by collecting DNA from a discarded item or requesting a sample from a close relative.

Using these techniques, Rae-Venter has identified more than 80 suspects who were ultimately arrested. She’s zeroed in on a suspect in as little as eight hours. “However, if the people sharing DNA are distantly related, that can be very time-consuming,” she says.

Wein didn’t know forensic genetic genealogy existed until he read about Rae-Venter solving the Golden State Killer case. Yet Wein, an operations expert who had worked to make fingerprinting, ballistic imaging, and sexual-assault investigations more effective, believed he could help. “I understood that this was just a math problem,” he says. “Until now, this has been done quite inefficiently, and unfortunately, the stakes are very, very high.”

Before Wein worked with Rae-Venter, he and Mine Su Ertürk, PhD ’22, tested a predecessor of his new algorithm with the DNA Doe Project, a nonprofit that investigates unidentified remains. Applying their approach to real data, Wein and Ertürk found that it solved cases around 10 times faster than existing methods.

Wein’s interactive algorithm, FIGGsolver (FIGG stands for forensic investigative genetic genealogy), is designed to speed up investigations by identifying the most efficient path for a researcher building a suspect’s family tree. “Our problem is one of finding a needle in a haystack,” Wein writes in a paper describing how the algorithm works, cowritten with Malo Sommers, a Stanford statistics PhD student, and published in Forensic Science International: Synergy. The “needle” is the suspect’s parents; the “hay” is every ancestral couple in the family tree who might turn out to be them. Step by step, FIGGsolver tells forensic genealogists how to shrink the haystack with the least possible effort.

Sleuths like Rae-Venter still must put in the legwork to uncover the identities of people in the tree. “The algorithm is meant to aid, not replace, genealogists’ decisions,” Wein says.

“A Fabulous Tool”

When Wein contacted Rae-Venter about deploying his algorithm to solve crimes, he assumed they’d start with something simple. Instead, she handed him a batch of sexual assault cold cases from Cuyahoga County, Ohio. “We threw him in the deep end,” Rae-Venter says. “These are ugly cases that we were completely stuck on.”

Since solving the Cleveland case in January 2025, Wein and Rae-Venter have closed out another 14. Given the gaps in the available data, “the fact that we’ve been able to solve anything is absolutely amazing,” Rae-Venter says. “The algorithm has been a fabulous tool. It’s able to do something we’re not by visually inspecting the matches.”

It does this by using mathematical models to weigh the costs and benefits of thousands of possible decisions against a backdrop of uncertainty, a process Wein compares to the mechanism a self-driving car uses to respond to its environment. The algorithm can’t analyze everything in a suspect’s family history: It doesn’t consider geography, surname, or ethnicity. Nor does it account for shared DNA resulting from intermarriage within insular communities. Forensic genealogists can still hit roadblocks when public records are missing — for instance, birth records for adoptees or documentation of Black ancestors who were enslaved.

But Wein says that speeding up the research process is invaluable. “Police pay genealogists a lump sum for a certain number of hours,” he says. “Solving cases faster means solving more cases.” In October, police in Rhode Island arrested a suspect in a 1985 sexual assault and murder case after contacting Wein for assistance.

Wein is currently working to make the FIGGsolver algorithm widely available to forensic investigative genetic genealogists. “Even if we can get 10% to use it, we can help solve hundreds of cases,” Wein says. “These are heinous acts — murder and rape. You’re taking dangerous people off the street so they won’t keep doing this. And you’re bringing closure to families.”