Massively Parallel Sequencing, Forensic Genetic Genealogy and the DeAngelo Paradigm

The Golden State Killer investigation demonstrates the full IGG workflow with forensic precision. The original STR profile from a crime-scene sample had been in the California CODIS system for years without a hit. In 2017, investigator Paul Holes contracted Parabon NanoLabs to generate a phenotype prediction from the crime-scene DNA (the Snapshot facial-composite service). In early 2018, Holes engaged genetic genealogist Barbara Rae-Venter (who had previously identified a Jane Doe cold case using GEDmatch in 2017) to run the IGG analysis.

The workflow Rae-Venter followed has since been formalised:

1. The crime-scene DNA was subjected to lcWGS at Parabon, generating approximately 800,000 SNP calls across the autosomes.
2. The resulting file was uploaded to GEDmatch, which at the time allowed law-enforcement uploads for violent felony cold cases under a policy change Parabon had negotiated with the platform.
3. GEDmatch's comparison engine returned a list of profiles sharing IBD segments with the crime-scene upload, ranked by estimated genetic distance (centimorgans, cM). The highest-ranking matches were third- and fourth-degree cousins (sharing approximately 220 to 55 cM of IBD segments on average).
4. Rae-Venter constructed family trees from each significant match, using public genealogical records (census data, vital records, newspaper obituaries, county court records) and the GEDmatch users' own public family trees to trace the common ancestors.
5. By triangulating the family trees from multiple independent matches, she narrowed the candidate pool to a group of descendants of a particular ancestral couple who would have been the right age to have been born in the late 1940s, with the right California geographic connection.
6. One candidate in that pool was Joseph James DeAngelo, a former Sacramento sheriff's department officer. A discreet DNA sample collected from his vehicle's door handle was STR-typed at 20 loci against the crime-scene profile and returned a match. The suspect was then arrested.

The William Earl Talbott II case (resolved in 2019) followed essentially the same workflow and produced the first IGG-enabled conviction when Talbott was found guilty of murdering a young couple in Washington State in 1987. The Christy Mirack case (December 1992 murder in Lancaster County, Pennsylvania) was resolved in June 2018, weeks after DeAngelo, using GEDmatch and a family tree that traced back through Mirack's killer's maternal line. In the Jane Doe Buckskin Girl case (Ohio, victim killed around 1981, identified in 2018), IGG was used not to identify a perpetrator but to put a name to a long-unidentified victim, demonstrating that the technique works symmetrically for victim identification as well as perpetrator identification.

GEDmatch was founded in 2010 as a free genealogy utility that allowed people to upload DNA data files from any consumer genetic testing company (23andMe, AncestryDNA, MyHeritage, etc.) and compare them against other users' uploads. By 2018 it had accumulated roughly 1 million profiles, all voluntarily uploaded. The site's terms of service in 2018 said nothing explicit about law-enforcement access; Parabon and Rae-Venter used GEDmatch under an informal policy that the platform permitted police use for violent-crime cases.

The privacy controversy following the DeAngelo arrest forced an immediate policy revision. In May 2019, GEDmatch changed its default to opt-out for law-enforcement matching: all existing profiles became invisible to law-enforcement searches unless the user explicitly opted in. This change is generally estimated to have reduced the law-enforcement-accessible database from roughly 1 million profiles to approximately 280,000, materially reducing the reach of the technique for profiles without close relatives who had opted in. GEDmatch was subsequently acquired by Verogen in 2020; Verogen maintains the opt-in architecture and a separate GEDmatch PRO tier for law-enforcement use.

FamilyTreeDNA (FTDNA), a Houston-based consumer testing company, announced in February 2019 that it had been cooperating with the FBI in uploading law-enforcement profiles to its database. This revelation also prompted a policy revision: FTDNA now operates an opt-in system for law-enforcement matching, and the company publishes a transparency report on the number of law-enforcement profiles active in the database.

Neither GEDmatch nor FTDNA allows access to the major commercial consumer databases: AncestryDNA (currently the largest at over 20 million profiles) and 23andMe (over 10 million profiles) do not accept law-enforcement uploads and require a valid court order or legal process before releasing any data about their users. This boundary is maintained by company policy and Terms of Service; in the US there is no federal statute that compels consumer genetics companies to accept law-enforcement profile uploads.

The mathematical logic of third-cousin triangulation is grounded in IBD segment sharing statistics. Two individuals share an IBD segment when they have both inherited a particular chromosomal region from a common ancestor without an intervening recombination event breaking it. The expected amount of shared IBD decreases approximately by half for each additional generation of separation:

• First-degree relatives (parent-child, full siblings): share on average 3,500 to 4,000 cM
• Second-degree (grandparent-grandchild, aunt-nephew): 1,700 to 2,100 cM
• Third-degree (first cousin): 850 to 1,050 cM
• Fourth-degree (first cousin once removed): 425 to 525 cM
• Fifth-degree (second cousin): 212 to 263 cM
• Sixth-degree (third cousin): 106 to 133 cM

At the third-cousin level, an unknown individual (the suspect) will appear in a consumer database match list if one or more of their third cousins has uploaded their own profile. Three-to-five strong third-cousin matches, each traceable to a different common great-great-grandparent pair, are usually sufficient to triangulate the common ancestor pool and narrow the candidate list. Because each American with European ancestry has on average 190 third cousins, the probability of finding at least one in the approximately 1.3-million-profile opt-in GEDmatch database is mathematically high for profiles with Northern European ancestry.

The genealogical reconstruction work is intensive: for the DeAngelo case, Rae-Venter reportedly built family trees covering thousands of descendants to eliminate candidates and isolate DeAngelo. This work is now routinely conducted by a small number of certified genetic genealogists who specialise in IGG, including teams at Parabon NanoLabs, Othram, and independent practitioners trained through the Association of Professional Genealogists.

The Phoenix Canal murders (2015-2016, resolved in 2020 using IGG) involved a suspect of mixed heritage whose match network in GEDmatch was sparser than in the DeAngelo case, requiring more extensive family-tree reconstruction before a candidate emerged. This case demonstrated that the technique functions below the ideal density of third-cousin matches, but at the cost of substantially more genealogical labour.

The US Department of Justice published its Interim Policy on Forensic Genetic Genealogy DNA Analysis on 24 September 2019. The policy applies to all federal law-enforcement agencies and to state and local agencies receiving federal funding. Its key provisions:

• IGG is permitted only for violent crimes (murder, non-negligent manslaughter, sexual assault) and in unidentified-remains cases where the decedent cannot be identified by other means.
• IGG may not be used to identify suspects in non-violent offences, to investigate people exercising First Amendment rights, or to develop information about individuals not directly relevant to the investigation.
• Traditional investigative genealogy techniques (checking genealogy websites and obituaries, tracing family trees through public records) must be exhausted before GEDmatch or FTDNA uploads are made.
• The uploading agency must use only forensic profiles from crime-scene evidence, not from suspects under investigation or from elimination samples from law enforcement.
• FBI approval is required before a federal law-enforcement agency uploads a profile.

The policy is explicitly interim, it is a guidance document, not a statute, and may be amended. Several US states have enacted or proposed state-level legislation that either restricts IGG further (Maryland, Montana) or explicitly authorises it under the DOJ conditions (Virginia).

International divergence is significant. Australia's Attorney-General's Department declined to adopt IGG, citing privacy law incompatibilities with the Privacy Act 1988. The UK Home Office has not authorised IGG use for domestic criminal investigations; the Forensic Science Regulator's office has noted that uploading to GEDmatch from a UK crime scene would likely require a legal basis under the Data Protection Act 2018 and Human Rights Act 1998 that does not currently exist for routine police use. The Netherlands National Police has used Bonaparte for kinship matching in DVI but has not publicly used IGG for criminal suspects. Canada's Privacy Commissioner has raised objections to IGG use that have not been resolved by formal legislation.

India has no specific policy or guidance on IGG. The Bharatiya Sakshya Adhiniyam 2023 (the new Evidence Act, replacing the Indian Evidence Act 1872) governs opinion evidence including DNA evidence, but does not address IGG or genealogy-database searches specifically. Consumer genetic testing uptake in India remains relatively low compared with the US, making large-scale third-cousin matching in existing databases practically limited even if legal questions were resolved.

The Anne Marie Fahey murder (Delaware, 1996): The case involves the murder of Governor Thomas Carper's scheduling secretary by Thomas Capano, a prominent Wilmington attorney. Capano was convicted in 1999 and died in prison in 2011. The case is not an IGG case, but it is routinely cited in the forensic-genealogy literature as an example of a conviction built on circumstantial evidence plus trace DNA that predates the CODIS era, where a genealogy-database search would have been theoretically applicable had the technology existed. Its inclusion in IGG training curricula illustrates the retrospective power of the technique.

The JonBenet Ramsey case (Colorado, 1996): Unidentified DNA from the crime scene has been in the Colorado CODIS system for decades without a hit. In 2024, a Boulder District Attorney's team publicly confirmed that IGG analysis had been applied to the unidentified DNA. Reports indicated that the SNP profile generated from the crime-scene swab had not produced the kind of strong third-cousin match network that the DeAngelo case yielded, likely due to limited genetic representation of the donor population in the opt-in GEDmatch database or due to DNA quality limitations in the 28-year-old sample. The case illustrates the practical ceiling of IGG: even a high-priority, high-resource investigation with DNA quality sufficient for genome-wide SNP profiling may not produce an actionable match if the suspect's relatives are not adequately represented in the accessible opt-in databases.

The JonBenet 2024 reporting also prompted a broader discussion about the reporting threshold for IGG non-results: when should an investigating agency disclose that IGG was attempted but inconclusive, and what obligation does that disclosure create regarding the management of the genealogical data generated from the crime scene?