Description
Read the following articles and link. Based on what you have learnt this week about forensic genealogy, in which cases/samples would be useful to apply this technique? What do you think about the ethical implications? How about its implementation on a regular basics in crime labs? Main post and 2 replies.What Your Second Cousin’s DNA May Say About You – The ISHI Report April 2019 (foleon.com)
Unformatted Attachment Preview
correspondence
Forensic genealogy and the power of defaults
To the Editor — In March 2019,
FamilyTreeDNA, the fourth largest directto-consumer genetics firm, announced
that it was updating its terms of service
and privacy policy to enable users to
decline to participate in a particular
type of DNA data sharing: the sharing of
their genetic data with law enforcement
for purposes of investigating crimes1. In
May 2019, GEDmatch, a free, searchable
online genetics database, announced a
similar update2. Since the April 2018 arrest
of the alleged ‘Golden State Killer’, law
enforcement has eagerly embraced the use
of consumer genetics platforms to identify
suspects for investigation and, in some
instances, arrest. Already, nearly three dozen
investigations using these platforms have
delivered suspects to law enforcement.
As many as 26 million people have used
a consumer genetics service to date3, though
most of their genetic profiles have remained
inaccessible for routine law enforcement
use. Initially, only GEDmatch permitted
law enforcement free rein to search its
data to investigate crimes. In late January
2019, however, the world learned that
FamilyTreeDNA had secretly been working
with the US Federal Bureau of Investigation
(FBI) for nearly a year to analyze crimescene DNA samples and compare them
to the genetic profiles in its database—
without explicitly informing its users
about this practice4. Indeed, all the while,
FamilyTreeDNA held itself out as committed
to user privacy, summarizing its privacy
policy as a commitment that “we won’t
share your DNA5.”
Bringing wrongdoers to justice and
resolving cold cases is undoubtedly
laudable, yet news of FamilyTreeDNA’s
secret law enforcement cooperation
was met with considerable outrage and
dismay. FamilyTreeDNA hastily issued
a letter to users apologizing for its poor
communication but defending its decision
to permit law enforcement access to its
database of user genetic profiles. In that
letter, FamilyTreeDNA president Bennett
Greenspan explained that users should
have expected law enforcement to make
use of the FamilyTreeDNA database.
Greenspan claimed that permitting law
enforcement use of the database was
consistent with FamilyTreeDNA’s terms of
service. Those terms, in turn, stated that
FamilyTreeDNA services must not be used
“for any law enforcement purposes, forensic
examinations, criminal investigations, and/
or similar purposes without the required
legal documentation and written permission
from FamilyTreeDNA6.”
And then, in March of this year,
FamilyTreeDNA backed away from its
prior position that law enforcement access
to its DNA database is much the same
as ordinary users’ access for purposes
of genealogy research. Recognizing that
law enforcement use involves different
expectations, interests and implications
than traditional uses of consumer genetics
platforms, FamilyTreeDNA announced that
users may opt in or out of “law enforcement
matching,” while retaining the ability to
match with non-law-enforcement-related
genetic relatives1.
Similarly, at GEDmatch, the creation of
a law enforcement option soon followed
public disclosure of law enforcement use
of the platform in a way that users did not
anticipate. The operator of the GEDmatch
platform had authorized a one-time law
enforcement use of the GEDmatch database
to investigate an aggravated assault—a crime
not within the site’s then-existing terms of
service2. Shortly after an arrest in that case
was announced, GEDmatch made at least
two significant updates to its user policies.
First, like FamilyTreeDNA, GEDmatch
enabled users to decline to participate in
law enforcement use of their genetic data.
Second, GEDmatch expanded its definition
of ‘violent crime’ to include not only
homicide and sexual assault crimes, but
also certain manslaughter, robbery and
assault crimes2.
Interestingly, FamilyTreeDNA has
adopted different defaults for users with
accounts located in the European Union
(EU) and those with non-EU accounts. As
FamilyTreeDNA explains, in general, “[i]
f you do not wish to be matched with these
designated law enforcement registered
users, you have the ability to opt out by
adjusting your Matching Preferences,
which now includes an option to opt-out
of Law Enforcement Matching.” Users with
EU-based accounts, however, have already
“been opted out of Law Enforcement
Matching, but may choose to opt in1.”
By contrast, GEDmatch has adopted a
uniform default of non-participation in law
enforcement matching, while permitting
users to opt in to such use2.
Presumably, FamilyTreeDNA—and
perhaps also GEDmatch—opted out
EU-based user accounts by default to
comply with the EU’s General Data
NATURE BIOTECHNOLOGY | VOL 37 | JULY 2019 | 707–713 | www.nature.com/naturebiotechnology
Protection Regulation (GDPR). The GDPR
establishes extensive data protection and
privacy rules to enable individuals to better
exercise control over their personal data and
how it is used.
The distinction between opting users
in or opting them out by default may seem
at first blush to be largely superficial. After
all, users can avoid sharing their genetic
information with law enforcement under
either system. However, this default could
have far-reaching—and perhaps intended—
consequences. Making law-enforcement
matching the default in the United States,
while adopting the opposite default in the
EU, could lead to dramatically different
outcomes. FamilyTreeDNA’s US customers
will be far more likely to share their genetic
data with law enforcement than their
European counterparts. The nature of
defaults offers an explanation.
Defaults are notoriously sticky. That is,
people often do not change them. When a
default option is in place, users tend to stick
with the default. Defaults are therefore very
powerful and can have serious implications.
Certain tendencies in human behavior
explain why. Defaults are what happen when
people do nothing. Research shows that
individuals have a tendency toward inertia,
particularly when decisions are complex7.
They will simply put off a complicated
decision and, as a result, end up staying with
the default option. Importantly, defaults are
so sticky that people may stay with a default
even when it is against their best interests.
One study in the United Kingdom found
that lower-income employees kept a default
savings plan that had a high contribution
rate, despite the plan’s being suboptimal,
arguably to the point of being harmful8.
Defaults, in other words, have the power to
lock in undesired and undesirable results.
Policymakers can harness the
stickiness of defaults to encourage certain
behaviors. For instance, some scholars
have advocated for adopting an opt-out
system to increase organ donation, relying
on the power of defaults in prescribing
policy. Thus, selecting one default rule
over another is necessarily strategic. When
an entity like FamilyTreeDNA chooses a
particular default, it nudges people—often
intentionally—toward a particular outcome.
And that outcome appears to be precisely
what FamilyTreeDNA wanted to do.
Using different default settings for
EU and non-EU user accounts was not a
minor decision. After all, rolling out two
707
correspondence
different defaults required FamilyTreeDNA
to determine which user accounts were
EU-related and which were not, while a
single default rule could have been rolled
out seamlessly to all users at once. Similarly,
using two default rules also required
FamilyTreeDNA to write different code
to instruct its systems to select a different
initial default option based on user location.
FamilyTreeDNA’s differential default
rules thus raise questions about why the
service would go to the extra effort to adopt
different rules for different users. The
answer, it is likely, turns on FamilyTreeDNA’s
internal determination to facilitate as
widespread a law enforcement matching
database as possible.
But as FamilyTreeDNA’s own adoption
of a law enforcement matching preference
option demonstrates, the availability of
consumer genetic data for law enforcement
use is not an unalloyed good. Law
enforcement access to consumer genetic
data raises significant privacy, ethical
and legal concerns9. These concerns,
moreover, go beyond the users of a service
like FamilyTreeDNA themselves to affect
the broad swath of genetic relatives who
may also now be identifiable to police
through the DNA of their kin. And these
concerns go to a determination about the
appropriateness of a near-national DNA
database by happenstance, rather than by
democratic deliberation and legislation10.
In opting non-EU users in by default,
FamilyTreeDNA is effectively arrogating to
itself to a decision about the appropriateness
of forensic genealogy, familial forensic
identification, and a de facto near-national
DNA database. That is not FamilyTreeDNA’s
decision to make.
❐
Natalie Ram1 and Jessica L. Roberts2*
University of Baltimore School of Law, Baltimore,
MD, USA. 2University of Houston Law Center,
Houston, TX, USA.
*e-mail: [email protected]
1
Published online: 12 June 2019
https://doi.org/10.1038/s41587-019-0172-5
References
1. FamilyTreeDNA. Updates. https://mailchi.mp/familytreedna/
updates-to-our-terms-of-service-and-privacy-policymarch19?e=dfef197239 (12 March 2019).
2. Aldhous, P. Tis genealogy database helped solve dozens of
crimes. But its new privacy rules will restrict access by cops.
BuzzFeed News https://www.buzzfeednews.com/article/
peteraldhous/this-genealogy-database-helped-solve-dozens-ofcrimes-but (19 May 2019).
3. Regalado, A. More than 26 million people have taken an
at-home ancestry test. MIT Technology Review https://www.
technologyreview.com/s/612880/more-than-26-million-peoplehave-taken-an-at-home-ancestry-test/ (11 February 2019).
4. Hernandez, S. One of the biggest at-home DNA testing
companies is working with the FBI. BuzzFeed News https://www.
buzzfeednews.com/article/salvadorhernandez/family-tree-dnafi-investigative-genealogy-privacy (31 January 2019).
5. FamilyTreeDNA. https://perma.cc/A6TW-SW5P (1 April 2019).
6. FamilyTreeDNA. A letter to our customers. https://mailchi.mp/
familytreedna/letter-to-customers?e (3 February 2019).
7. Fleming, S. M., Tomas, C. L. & Dolan, R. J. Proc. Natl Acad. Sci.
USA 107, 6005–6009 (2010).
8. Beshears, J. et al. Te limitations of defaults. National Bureau of
Economic Research, Retirement Research Center http://www.nber.
org/programs/ag/rrc/NB10-02%20Beshears,%20Choi%20et%20
al%20SUMMARY.pdf (September 2010).
9. Ram, N. Genetic privacy afer Carpenter. Va. Law Rev. (in
the press); https://papers.ssrn.com/sol3/papers.cfm?abstract_
id=3265827
10. Erlich, Y., Shor, T., Pe’er, I. & Carmi, S. Science 362,
690–694 (2018).
Acknowledgements
Funding was provided by a Greenwall Foundation Faculty
Scholars Grant.
Competing interests
The authors declare no competing interests.
Pre-market development times for biologic
versus small-molecule drugs
To the Editor — Before approval by the
US Food and Drug Administration (FDA),
prescription drugs must be adequately tested
in preclinical studies and clinical trials for
safety and efficacy. To allow manufacturers
sufficient time to earn a profit on resources
invested in conducting these studies, brandname drugs are protected by patents, which
last 20 years from the date of application1.
The key patents protecting a drug—such
as those associated with the drug’s active
ingredient—tend to be filed shortly after the
new drug is discovered or synthesized. Thus,
longer periods of drug development leading
up to FDA approval result in reduced time
remaining on this fundamental patent
during the post-approval period before
market entry by competitors2.
Biologic drugs—complex drugs derived
from living cells—are thought to be
particularly time- and resource-intensive
to develop3. The lengthy development
process attributed to biologic drugs was
cited by legislators when the US Congress
passed the Biologics Price Competition
and Innovation (BPCIA) in 2009, which
granted new biologics 12 years of guaranteed
exclusivity. Similarly, the recently proposed
708
renegotiation of the North American Free
Trade Agreement (NAFTA)—known as the
United States–Mexico–Canada Agreement
(USMCA)—would require Canada and
Mexico to provide 10 years of exclusivity
protections for biologic products for all new
drugs4 (Canada currently provides 8 years);
expanded biologic exclusivity protections
have been proposed in other trade
negotiations, such as with Japan5. However,
sales of biologic drugs now account for
approximately one-third of prescription
drug spending in the US, and high prices for
these products have led some policymakers
and consumer advocates to consider
reforms to the BPCIA3,6–8.
Despite their policy importance, the
development times for biologic drugs are
poorly understood, primarily because
patent data for biologics have not been
easily accessible. Previous studies have
measured the amount of time required for
human testing (phase 1 to phase 2 or
3 trials and regulatory review), but these
have not also accounted for development
time before human testing9,10. To expand on
this previous work, we used the key patent
filing dates associated with small-molecule
and biologic drugs to determine whether
there is a difference in the amount of time
these drugs spend in development before
FDA approval (Box 1).
We found that, between 2007 and
2016, the FDA approved 275 new drugs,
of which 212 (77%) were small-molecule
drugs and 63 (23%) were biologic drugs
(Supplementary Table 1 and Supplementary
Data)11. Key patents could be identified for
92% (252) of these products using data from
US Patent and Trademark Office (USPTO)
patent term restoration data, and for 89%
(245) of products using the Merck Index.
Across the study cohort, median total
development times—from first patent filing
to FDA approval—were similar between
the USPTO data (12.4 years; interquartile
range (IQR), 9.7–15.3 years) and the Merck
Index (12.1 years; IQR, 9.2–17.7 years). Total
development times appeared to be stable
or decreasing slightly over the past decade
(Supplementary Fig. 1).
Median total pre-market development
times were not different between biologic
and small-molecule drugs using USPTO
data (12.4 versus 12.4 years; P = 0.68) and
were shorter for biologic drugs using the
NATURE BIOTECHNOLOGY | VOL 37 | JULY 2019 | 707–713 | www.nature.com/naturebiotechnology
Forensic Science International: Genetics 36 (2018) 186–188
Contents lists available at ScienceDirect
Forensic Science International: Genetics
journal homepage: www.elsevier.com/locate/fsigen
Commentary
The Golden State Killer investigation and the nascent field of forensic
genealogy
T
Chris Phillips
Forensic Genetics Unit, Institute of Forensic Sciences, University of Santiago de Compostela, Spain
A R T I C LE I N FO
A B S T R A C T
Keywords:
Genetic genealogy
DTC genetic testing
GEDmatch
Criminal investigative practice
SNPs
The likely genetic analysis steps taken to identify suspect Joseph DeAngelo in the recently resolved Golden State
Killer investigation are discussed. The consequences for the forensic genetics community of introducing much
more detailed SNP analysis regimes, as used by the Golden State Killer investigators, are reviewed along with
some of the limitations in accuracy and sensitivity that may be involved in such approaches.
As a long-standing advocate of the application of genomics to forensic DNA analysis, I followed news reports of the successful detection
of the Golden State Killer with increasing interest. Much of the press
coverage centered on the ethical issues raised by certain actions taken
by the investigators to identify Joseph DeAngelo, namely: surreptitious
sample collection; accessing the community-sourced GEDmatch database to detect a criminal suspect – a “citizen science” website originally
established to help reunite kinship members or adoptees with their
biological parents; and passing off the variant data obtained using DNA
from a preserved rape kit from the 80’s, as a personal genome report.
The companion commentary by Denise Syndercombe Court discusses
these and other ethical aspects to the case. Here I will review the genetics involved (as far as this is possible to discern from often scant or
inaccurate reportage), and the consequences of the analyses made by
the Golden State Killer investigation, herein GSKI, for forensic DNA
analysis and investigative practice in the future.
While the description of the genomic analysis steps taken to find
DeAngelo appear to be the latest groundbreaking technology, investigators have been considering the use of freely accessible personal
genetic testing data for some time. This is partly a response to their
access being barred to customer data held by the four biggest US directto-consumer (DTC) genetic testing vendors: AncestryDNA; 23andMe;
Family Tree DNA; MyHeritage (in order of user base). There has been a
degree of success from extending the concept of familial searching beyond CODIS data and its confinement to close relative matches. For
example, in 2015 Bryan Miller was identified as prime suspect in the
Phoenix Canal Killer case by genealogist Colleen Fitzpatrick who uploaded a Y-Filer profile to a number of public Y-STR databases (probably Ysearch, the Sorenson Molecular Genealogy Foundation and
AncestryDNA, although the latter has ceased offering Y-STR tests) [1].
Interestingly, Fitzpatrick had volunteered her genealogical services to
the forensics community after giving a talk at the 25th ISHI conference
https://doi.org/10.1016/j.fsigen.2018.07.010
Received 27 June 2018; Accepted 12 July 2018
Available online 17 July 2018
1872-4973/ © 2018 Elsevier B.V. All rights reserved.
in Phoenix the year before [2]. Y-STR analysis started much of the initial developments in genetic genealogy and the tools hosted by GEDmatch evolved from a ‘Rogers’ surname-matching group run by its cofounder Curtis Rogers [3]. GSKI first tried limited Y-STR matching in
2017 (one of the genotypes was described as atypical in European Y
lineages), but this misidentified a 73-year old man in Oregon. A recent
report indicates that GSKI used the open-access Ysearch database (now
closed) and issued a subpoena to Family Tree DNA to reveal the details
of a match of 12 Y-STRs [4] – instigating this failed investigative lead.
As early as 2014, investigators were using a Y-STR profile to attempt
the identification of DNA from the murder of Angie Dodge, but analyses
were similarly restricted by a lack of genetic data and an innocent
person was misidentified by the adventitious hit obtained in this case
[5]. It might be the case that a study in 2013 by Gymrek et al. [6]
indicating “triangulation” methods could identify otherwise anonymous DNA data in open genealogical databases (combining Y-STR
genotypes with other metadata) led to over-confidence in adopting the
same approach for criminal investigations. However, the major advances in genetic genealogy in unknown parentage and missing persons
cases have come from use of large-scale autosomal marker sets, not Y
data [7].
Genealogists have identified missing persons with increasing precision, as an additional form of analysis to the established help given to
those wishing to discover their distant relatives using the proprietary
company databases and GEDmatch. This is the inevitable outcome of
both the large number of individuals accumulated by vendor- or community-databases, and the density of the genetic marker data all four
vendors provide. At the time of writing, AncestryDNA has 10 million
DNA customers and 23andMe 5 million, while the other vendors and
GEDmatch have about 1 million each. All vendors use customized
versions of the Illumina OmniExpress Array or Illumina Global Screening
Array which both comprise over 650,000 loci. A key study in 2012 by
Forensic Science International: Genetics 36 (2018) 186–188
C. Phillips
Henn, et al. [8] showed that 2nd to 9th cousinships were detectable
using identity-by-descent (IBD) measurements of a 550 K SNP intersect
from the 650 K SNP data of the HGDP-CEPH panel and the 580 K SNPs
genotyped by 23andMe at that time. Therefore, the tools to detect close
and distant relatives using high-density SNP datasets have been in place
for some time. These consist of: i. IBD analysis that can detect very
distant cousin pairs and thus provides a depth of information that can
range from manageable numbers of related individuals sharing greatgrandparents, up to much more extensive but distant kinships; ii. high
density SNP data from the DTC vendors to allow customers to query the
vendor’s proprietary databases or share their data in GEDmatch to
broaden a search out to the genealogy community (GEDmatch has a
Genesis database allowing comparisons between data from different
SNP arrays); and iii. growing availability of extensive online genealogy
databases (e.g. censuses; birth, marriage and death indices; digitized
newspapers; etc.). As a whole, these factors allow detailed family tree
reconstruction, which not only routinely handles centimorgan values,
but compiles a wide range of other details, including birthdates, places
of residence and kinship member’s personal characteristics (see [9] for
the blog details of three authoritative practitioners). It is notable that
GEDmatch have changed their terms and conditions to directly highlight the possibility of investigative use of community data held on the
website and now allows users to withdraw their data; but at the same
time, they have made it easier for investigators to upload probative
DNA data without the need to use an alias. These steps are to be
commended, as transparency is a crucial factor in preserving user’s trust
and it is important that those who are not concerned about distant
relatives being detected by police are free to continue exploring their
personal data and benefiting from the custom tools that have made
GEDmatch so effective for this purpose.
Given the developments described above, it might seem surprising
that police have only just started to obtain major investigative breakthroughs (and when they do, three separate cases identifying murder
victim Marcia King [10], and suspects Joseph DeAngelo and William
Earl Talbott [11] are reported within weeks of each other). Yet the
obvious missing piece in the jigsaw of genomic analyses made in such
cases is how they were able to generate a sufficient amount of SNP data
from limited evidential material. In the absence of information from
GSKI, a large degree of supposition is needed here to reconstruct the
processes likely to have been used to obtain the SNP data eventually
uploaded to GEDmatch. Therefore, more detail may eventually come to
light that changes this perspective and the purported steps taken might
be inaccurate in specific details. Initially, DNA was taken from a duplicate rape kit [12] and treated as single source (no details of the
applied extraction processes are available), then analyzed with a SNP
array compatible with the DTC vendor databases [4]. If the SNP genotypes generated centimorgan values from “chromosome painting” that
revealed a specific kinship, connected in part to the geographic distribution of criminal activity, then investigators could begin to compile
supporting details about places of residence, lifestyle and physical
characteristics of age- and sex-matched kinship members. This process
can provide an effective approach even with very extensive family
trees, of which only a handful of members may have had their DNA
tested by DTC vendors. The information gathered does not always need
to be particularly specific: Bryan Miller was one of tens of thousands of
Millers living in Phoenix, but the only one on a list of suspects at the
time [1]. Although the principal vendor- and community-databases
hold much more data from US customers than from other countries, the
application of the genealogical analyses described here have inevitably
brought interest from police elsewhere, such as Sweden, where an estimated 40,000 people have taken a genetic ancestry test [13].
Descriptions of large-scale genomic analyses may seem disconcerting when, until recently, forensic DNA analysis has dealt with
just a small number of well-established STRs and has not used sequencing beyond the analysis of mtDNA. The jump from 24 A-STRs and
27 Y-STRs to more than 650,000 SNPs is a quantum leap indeed, and
most forensic laboratories will not have the bioinformatics infrastructure or analysis skills that such data scales demand. SNPs have
been perceived to be markers reserved for small-scale bespoke tests still
at an early stage of forensic use, such as ancestry analysis and phenotype prediction. Equally, the utmost care is taken in forensic testing to
control contamination; to ensure genetic markers are fully validated;
and to confirm the genotypes are specific to one donor, not a mixture of
signals from multiple DNA sources. It can also be suggested that investigators have now found a way to circumvent the carefully prepared
checks and balances put in place to regulate familial searches of national DNA databases. The important context here is the difference
between standardized forensic identification practices and obtaining
new genetic information by novel means that might provide vital investigative leads. In routine forensic practice, we all rightly concern
ourselves with precision and sensitivity; in other words, it is critical to
produce the best possible DNA profile from minimal contact traces and
which uniquely identifies an individual. An investigation does not need
to deal in such certainties and the police will willingly add the extra
detail DNA can provide to the body of information already gathered in a
complex criminal enquiry. Investigators can process information on a
scale of reliability, usefulness and relevance to the direction of enquiries – akin to the Admiralty System for evaluating compiled items of
intelligence [14]. I recently dismissed the suggestion that DNA from a
licked stamp on a letter sent by the Zodiac Killer [15] would be sufficient to obtain a whole genome sequence or SNP array data with over
half a million genotypes. Yet modern sequencing technologies potentially open the door to detailed genomic analysis of minimal DNA and
are likely to show such perceptions from an orthodox forensic viewpoint are too pessimistic. We could be witnessing the emergence of
genomic genealogy as a completely new field in forensic genetics – one
that uses the extended haplotype block, not STRs, as the forensic
marker of choice.
Acknowledgements
I particularly wish to thank Debbie Kennett, Honorary Research
Associate, UCL, London, for her invaluable help in piecing together the
probable DNA analysis steps employed in the GSKI process. I am also
indebted to Professors: Mark Jobling, Leicester University; and Denise
Syndercombe Court, Kings College, London, for helpful guidance and
informative discussions. Lastly, I would like to thank Dennis McNevin,
University of Technology Sydney, and Runa Daniel, Forensic Services
Department, Victoria Police, Australia, for ongoing discussions about
the future of police investigative practice and its use of emerging forensic DNA technologies.
References
[1] Phoenix Canal Killer genealogical analysis: https://eu.azcentral.com/story/news/
local/phoenix/2016/11/30/how-forensic-genealogy-led-arrest-phoenix-canalkiller-case-bryan-patrick-miller-dna/94565410/.
[2] 25th ISHI discussions on Y-STR lineage analysis: https://www.ishinews.com/
solving-the-phoenix-canal-murders/.
[3] Growth and development of GEDmatch: https://www.theatlantic.com/science/
archive/2018/06/gedmatch-police-genealogy-database/561695/.
[4] GSKI misdirection in 2017 using Y-STRs to an unrelated man in Oregon and subsequent success with a high density SNP set: https://www.buzzfeed.com/
peteraldhous/family-tree-dna-subpoena-golden-state-killer.
[5] The misidentification of Michael Usry Jr. in the Angie Dodge murder investigation:
http://www.postregister.com/articles/chris-tapp-coverage-featured-news-dailyemail/2017/07/29/contradictory-dna-results-put.
[6] M. Gymrek, A.L. McGuire, D. Golan, E. Halperin, Y. Erlich, Identifying personal
genomes by surname inference, Science 339 (2013) 321–324.
[7] Initial development of genetic genealogy: http://jogg.info/pages/vol8/editorial/
moore/moore-history.html.
[8] B.M. Henn, L. Hon, J.M. Macpherson, N. Eriksson, S. Saxonov, I. Pe’er,
J.L. Mountain, Cryptic distant relatives are common in both isolated and cosmopolitan genetic samples, PLoS One 7 (2012) e34267.
[9] ‘Cruwys news’, the blog of genetic genealogist Debbie Kennett: https://cruwys.
blogspot.com ‘Your Genetic Genealogist’ the blog of genetic genealogist CeCe
Moore: http://www.yourgeneticgenealogist.com/ ‘DNAeXplained’, the blog of
187
Forensic Science International: Genetics 36 (2018) 186–188
C. Phillips
genetic genealogist Roberta Estes: https://dna-explained.com.
[10] Identification of murder victim “Buckskin Girl”, Marcia King: http://cases.
dnadoeproject.org/mesmerize/buckskin-girl/.
[11] Identification of murder suspect William Earl Talbott: https://www.
washingtonpost.com/news/morning-mix/wp/2018/05/21/a-genealogy-websiteused-to-crack-another-cold-case-police-say-this-one-a-1987-double-homicide/?
utm_term=.2ae49615aa.
[12] Storage and use of a duplicate rape kit in the GSKI: https://abcnews.go.com/US/
inside-terrifying-golden-state-killer-crime-spree-key/story?id=54849196.
[13] Swedish police consider accessing Swedish DTC data: https://www.thelocal.se/
20180507/swedish-police-mull-using-dna-family-tree-websites.
[14] E. Bruenisholz, A. Ross, S. Prakash, C.P. Roux, M. Morelato, T. Raymond, S. Walsh,
T. O’Maley, The Intelligent Use of Forensic Data, Researchgate PDF at: https://
www.researchgate.net/publication/286930641_The_Intelligent_Use_of_Forensic_
Data.
[15] Genealogical analysis proposed for Zodiac Killer DNA: https://www.independent.
co.uk/news/world/americas/zodiac-killer-golden-state-investigation-dna-testscalifornia-vallejo-sacramento-a8343086.html.
188
Forensic Science International: Genetics 36 (2018) 203–204
Contents lists available at ScienceDirect
Forensic Science International: Genetics
journal homepage: www.elsevier.com/locate/fsigen
Forensic genealogy: Some serious concerns
T
Denise Syndercombe Court
King’s Forensics, Faculty of Life Sciences and Medicine, King’s College, London, UK
A R T I C L E I N F O
A B S T R A C T
Keywords:
Golden state killer
Genealogy
Ethics
DNA
Genealogical databases have provided links to possible perpetrators of crimes in several cold cases in the US.
This commentary discusses some of the ethical issues associated with this approach while recognising the underlying value of the identifications.
The use of genealogy to identify murderers is not new, as Chris
Phillips has discussed in his accompanying piece: but the risks of false
identification and intrusion into privacy is obvious to all. A false Y
chromosome STR match in the case of Chen Long-Qi, for example, led
to him being wrongly imprisoned in Taiwan for four years [1]. As the
direct-to-consumer (DTC) genomic industry has expanded significantly
as a consequence of the decreasing costs of large scale SNP typing, so
the industries’ ability to identify close and more distant relatives within
its databases has shifted the emphasis to the use of autosomal loci.
Indeed, one DTC supplier, Ancestry, stopped their Y-chromosome and
mitochondrial DNA typing service in 2014. Nevertheless, using this
approach to uncover relatives may not be that simple if the relationship
is more distant: the number of familial relationships increase exponentially so that we have almost 1000 fourth cousins and close to
5000 fifth cousins, only a proportion of whom will be on any given
database. To validate the links, genealogists use powerful triangulation
techniques to identify shared segments, but searching for an identity
where DNA is the sole source of the information is likely to require the
significant use of more traditional genealogical methods.
As forensic geneticists, we did not consider the methodology used to
identify Joseph DeAngelo as being of particular concern, despite
warnings from those with an interest in genomic privacy [2]. We are
familiar with the inherent difficulty of being able to obtain sufficient
cellular material at a crime scene for analysis, though semen is perhaps
a notable exception, and sexual homicides provide fertile territory for
the approach used in the case of DeAngelo. The detailed linking information available in a genealogy database is, of course, private to
those directly concerned, and DTC companies increasingly provide information about how they protect the p
