10 ways DNA technology is solving cold cases decades later

The evidence in many unsolved crimes was never the problem. A semen stain, a few hairs or a bloodstained shirt could sit in a property room for decades, holding information no laboratory of its era could read. What changed was the science. Laboratories can now recover a usable genetic profile from a trace sample. Databases can compare that profile against stored records, and the combination has turned old evidence into new leads.

That shift explains why DNA cold cases are now being solved decades after the original investigations stalled. Two cases from April 2018 show the method at work. Investigators uploaded a crime-scene profile to the genealogy database GEDmatch and used distant-relative matches to build family trees. That work led to the arrest of Joseph James DeAngelo, the Golden State Killer, in Sacramento County, California. He pleaded guilty in June 2020 to 13 murders. A second April 2018 case involved a young woman found dead in Ohio in 1981. Known for decades only as Buckskin Girl, she was identified through genetic genealogy as Marcia King, 21, of Arkansas.

This list covers 10 methods and cases. It is a selection, not a complete or ranked account, and the entries fall into three groups.

The first group is databases. The FBI's Combined DNA Index System, known as CODIS, links local, state and national collections of DNA profiles. Its national level has operated since 1998.

The second is laboratory technique. Short tandem repeat profiling, single nucleotide polymorphism testing, touch DNA recovery and mitochondrial DNA analysis each read genetic information in a different way. Some of these methods work on samples too small or degraded for older tests.

The third is investigative strategy. Genetic genealogy, familial searching, the testing of long-stored sexual assault kits and the naming of unidentified remains each apply DNA to a distinct problem. They range from finding an unknown suspect to identifying a long-unnamed victim.

The same science can also work in reverse. Re-testing old evidence can show that the wrong person was convicted. The Innocence Project, founded in 1992 by Barry Scheck and Peter Neufeld, uses post-conviction DNA testing to challenge wrongful convictions.

Some methods have limits. Trace samples can hold DNA from several people, and mitochondrial testing points to a maternal line rather than a single individual. Each slide explains one method or case, including what it can and cannot establish.

Illustration: Quartz

Forensic genetic genealogy identified the Golden State Killer by matching crime-scene DNA to distant relatives in a public database. Investigators uploaded a crime-scene profile to GEDmatch in 2018. That search led to the arrest of Joseph James DeAngelo.

The method does not need the offender's own DNA to be on file anywhere. GEDmatch is a public genealogy database. When investigators entered the crime-scene profile, the matches that came back belonged to distant relatives of the person who left it.

Those relatives were the starting point, not the answer. A distant match shows that two people share some DNA. That shared DNA means they also share ancestors somewhere in their family lines. It does not say who the unknown person is, or which branch of a family tree leads to that person.

Investigators used the matches to build family trees. By tracing how the relatives connected to one another, they could narrow the field of people who might have left the DNA at the crime scene. Each tree turned a broad set of genetic links into a shorter list of possible candidates.

A family tree produces a lead, not proof. It can point investigators toward a person, but it does not on its own establish that the person left the evidence.

DeAngelo was arrested in Sacramento County, California, in April 2018. In June 2020, he pleaded guilty to 13 murders. The plea resolved the case without a trial verdict. He was sentenced in August 2020 to life without parole.

The approach differs from a standard criminal database search. A search of offender records can only find a person whose profile is already stored. A genealogy database can reach relatives who have never had any contact with the justice system. Through those relatives, it can lead to an offender who was never on file.

Illustration: Quartz

The FBI's Combined DNA Index System, known as CODIS, links DNA databases kept at the local, state and national levels. A profile developed from crime-scene evidence in one jurisdiction can be compared against profiles held across that wider network.

The national tier is the National DNA Index System, or NDIS. It began operating in 1998. Evidence collected before then could not be searched nationally when it was gathered. That tier did not yet exist. Local and state databases sit below it, and the three levels together make up the network that CODIS connects.

The system runs two broad kinds of comparison. The first checks a profile from crime-scene evidence against profiles taken from known people. Those reference profiles come from convicted offenders and from arrestees. A match of this kind connects a piece of evidence to a person whose DNA is already on file.

The second comparison sets crime-scene profiles against one another. Two profiles from separate scenes can match even when the person who left them has never been entered as an offender or arrestee. A hit of that kind links the cases. The two scenes do not need to sit in the same jurisdiction, because the network reaches across all three tiers. The match indicates that the same unknown individual left DNA at both scenes, even though no name is attached.

An unsolved case with no named suspect can gain new leads this way. If its evidence profile matches evidence from a separate case, investigators can read the records of both cases together.

CODIS can only compare profiles that have been entered into it. Evidence from an old case has to be tested, and a usable profile developed, before it can be searched against the network at all.

CODIS is a government system built on offender, arrestee and crime-scene profiles. It is separate from the consumer genealogy databases that investigators have used to trace a suspect through distant relatives.

Illustration: Quartz

Short tandem repeat profiling, or STR analysis, identifies people by counting how many times short DNA sequences repeat at specific locations in the genome. The CODIS core set of those locations grew from 13 to 20 loci, effective Jan. 1, 2017. That change gives labs more points of comparison for every sample.

Each STR locus holds a short sequence of DNA bases repeated back to back. The number of repeats varies from person to person. A person inherits one version of each locus, called an allele, from each parent. A profile therefore records two repeat counts per locus. It records a single count when both parents passed on the same version.

Labs first copy the target regions using the polymerase chain reaction, or PCR. That copying step lets analysts work from small amounts of starting material. The copied fragments are then sorted by length using capillary electrophoresis. A longer fragment means more repeats. The result is a string of numbers that can be stored in a database and compared with other profiles.

No single locus identifies anyone. Many unrelated people share the same repeat count at any one location. The power to tell people apart comes from combining many loci. Each added locus makes it less likely that two unrelated people would share a full profile by chance.

The 2017 expansion added seven loci to the original 13: D1S1656, D2S441, D2S1338, D10S1248, D12S391, D19S433 and D22S1045. Several of the new markers are also used in European DNA databases. That overlap makes profiles easier to compare when an investigation crosses borders.

For cold cases, the larger core set changes what old evidence can show. A sample typed years ago at fewer loci can be retested if material remains. The new test produces a profile with more data points. A match backed by stronger statistics can carry more weight when the evidence is presented in court.

Retesting depends on what survives. Evidence consumed by earlier testing, or stored in poor conditions, may not yield a full 20-locus profile. A partial profile can still be searched, but it carries less power to distinguish one person from another.

Illustration: Quartz

SNP testing reads crime-scene DNA in the same format that consumer genealogy databases use, which standard forensic profiles cannot do. Sequencing-based methods can also recover usable data from small or degraded samples that conventional STR testing could not handle.

A single nucleotide polymorphism, or SNP, is a variation at one position in the genetic code. At that spot, one person might carry one chemical base while another person carries a different one. Reading many of these positions across the genome produces a detailed genetic profile.

Forensic labs rely on a different marker. Short tandem repeat analysis, the standard behind CODIS, measures repeat lengths at a fixed set of locations. That profile is built to match a sample against known people in a forensic database. It is not the kind of data genealogy databases hold.

The gap has practical consequences for cold cases. Consumer genealogy databases store SNP data, not the STR markers used in CODIS. A crime-scene STR profile cannot be compared directly with the family records in those databases. Investigators who want to search for relatives of an unknown offender first need the evidence converted into SNP form.

Private labs perform that conversion. Parabon NanoLabs and Othram both turn crime-scene DNA into SNP data that genetic genealogists can work with. That profile can then be compared with genealogy records to look for relatives of the person who left the sample.

Sequencing also widens the range of evidence worth testing. Some decades-old samples are too small or too degraded for standard STR testing to produce a usable result. Sequencing can work on material in that condition.

Agencies holding decades-old evidence can revisit items that earlier tests left without a profile. A swab or stain once set aside as unusable may now yield enough genetic information to produce an investigative lead. Any lead from a SNP profile still points investigators toward a family, not straight to a named suspect.

Illustration: Quartz

Familial DNA searching scans an offender database for profiles that come close to a crime-scene sample without matching it exactly. A near match of that kind can point investigators toward a relative of the unknown offender. California used the method to identify Lonnie Franklin Jr., the man known as the Grim Sleeper.

The method rests on inheritance. Each person receives half of their DNA from each parent. As a result, close family members carry many of the same genetic markers. A standard database search asks whether a crime-scene profile is identical to a stored one. A familial search asks something else: whether a stored profile shares enough markers with the evidence to suggest a parent, child or sibling.

A partial match does not name a suspect. It names a family. Investigators still have to work out which relative, if any, fits the crime, using conventional checks such as age, location and timeline. The lead holds up only when a DNA sample from the suspected person is compared directly against the crime-scene evidence. An exact match is what confirms the identification.

Franklin was arrested in July 2010. He was convicted in May 2016 and sentenced to death.

Familial searching is a different tool from the forensic genetic genealogy used in other cold cases. It runs inside government databases of convicted offenders. Those databases hold STR profiles built from a limited number of markers. That is enough to flag a parent, child or sibling, but not a distant cousin. Genetic genealogy relies instead on consumer databases built from SNP data, which can reveal much more distant family connections.

Familial searching also reaches people a routine search does not. A familial hit can bring police attention to someone who has never been charged with a crime, solely because a relative's profile is on file. The search works only if a close relative of the offender already has a profile in the database.

Illustration: Quartz

Touch DNA is genetic material recovered from skin cells a person leaves on objects such as clothing or tools. Lab testing is now sensitive enough to produce a profile from very small samples of it.

That change bears directly on evidence collected long ago. An item that once yielded nothing usable can now return a profile. That gives investigators a reason to pull stored evidence and test it again.

The term describes where the DNA comes from. Blood, semen and saliva are not the source. The source is skin cells left behind through contact. A shirt grabbed during a struggle or the handle of a tool used at a scene can carry this kind of trace.

That widens the range of evidence worth examining. In a case with no visible biological stain, the objects a suspect touched may still hold genetic material.

The same sensitivity makes results harder to interpret. A small sample from a handled object can hold DNA from several people at once. Forensic analysts call this a mixture. Separating one person's profile from a mixture is harder than reading a sample that came from a single source.

Innocent transfer is the second complication. DNA can end up on an object without its owner having any part in the crime. A profile on a piece of evidence shows that a person's genetic material was present. It does not show how or when that material got there.

Those two limits shape what a touch DNA result can establish. A match can point investigators toward a person. It cannot by itself show that the person committed the crime. Investigators still have to account for how the DNA reached the object. With evidence stored for years, that question covers everyone who handled the item after it was collected, not only people at the scene.

Illustration: Quartz

Mitochondrial DNA testing can recover genetic information from a hair shaft that has no root attached. That makes it an option for old evidence when the nuclear DNA used in standard profiling is absent.

Evidence boxes in unsolved cases can hold hairs that were collected without their roots. Without a root, the nuclear DNA that conventional profiles depend on may not be there to test. Mitochondrial DNA is present in the shaft itself. A rootless hair therefore does not have to be written off as unusable evidence.

Mitochondrial DNA is inherited through the maternal line. A mother passes the same sequence to each of her children. Her daughters pass it on to the next generation. Her sons carry it too, but they do not pass it to their own children. That inheritance pattern is what makes the test useful. It also sets the method's main limit.

The limit is that mitochondrial DNA cannot single out one person. A mother, her children, her siblings and other relatives on the maternal side share the same sequence. A match between a hair and a suspect places that suspect within a family line rather than identifying the suspect alone.

For that reason, a mitochondrial match does not close a case on its own. Investigators still need other evidence to tell apart the people who share the sequence, such as two brothers or a mother and her son.

The test works in the other direction too. If the sequence from a hair differs from a person's sequence, that person is not the source of the hair. In a cold case, a mismatch can rule someone out and narrow the list of people under consideration.

The shared sequence also means a maternal relative can provide a reference sample when the person in question cannot be sampled directly. That relative carries the same mitochondrial sequence as the individual being sought.

Illustration: Quartz

Sexual assault kits that sat untested in storage have produced DNA matches on old cases in several U.S. cities. Labs develop profiles from the stored evidence and search them through CODIS, the FBI's Combined DNA Index System.

Detroit shows how large a backlog can grow. In 2009, the city found more than 11,000 sexual assault kits that had never been tested.

Each kit holds biological evidence collected after an assault. That evidence cannot be compared with anything until a lab develops a DNA profile from it.

Once a profile exists, it can be searched against the offender profiles stored in CODIS. A hit can give investigators a name in a case that had gone cold.

A profile from one kit can also match evidence from a different crime scene. That match ties separate cases together before any suspect is known.

For cold cases, the backlog itself is a store of evidence. The material was collected years earlier, when the original assaults were investigated. The missing step was testing, which turns stored material into a searchable profile.

Illustration: Quartz

DNA can give a name back to a victim, not only point to a suspect. In April 2018, genetic genealogy identified the young woman long known as Buckskin Girl as Marcia King, 21, of Arkansas.

Her body was found in Ohio in 1981. No one knew who she was. For decades, the case carried a nickname instead of her name.

From 1981 until 2018, the open question in the case was not who had committed a crime. It was who the dead woman had been.

This makes unidentified remains a distinct kind of cold case. Their first task is to establish who the person was.

The identification also connected two states. King was found in Ohio, but the DNA result traced her to Arkansas.

The Buckskin Girl case shows a second use for genetic genealogy. The approach that can lead investigators to an unknown suspect produced a name for a victim here.