Forensic Science International: Genetics 43 (2019) 102129

Contents lists available at ScienceDirect

Forensic Science International: Genetics

journal homepage: www.elsevier.com/locate/fsigen

Research paper

The effect of FBI CODIS Core STR Loci expansion on familial DNA database T
searching
⁎
Efthymia Karantzalia, , Phaedra Rosmarakia, Anastasios Kotsakisa,
Marie-Gaëlle Le Roux-Le Pajolecb, Georgios Fitsialosa
a
The European Center for Genetics and DNA Identification, DNAlogy. 98 Vouliagmenis Ave. Glyfada, 16674, Athens, Greece
b
Institut Génétique Nantes Atlantique IGNA, 1 Avenue des Lions, CS 40193, 44802 Saint-Herblain Cedex, France

A R T I C LE I N FO A B S T R A C T

Keywords: 1000 profiles chosen randomly from an in-house database of 6000 profiles were searched against the database
CODIS expansion for matches with at least one shared allele per locus. The database contains profiles that have been analyzed with
Familial searching Identifiler Plus (15 markers) for biological relationship and DNA identification purposes and both true and false
20 STRs matches are expected to be obtained. 100 pairs of at least one true paternity match and one false match were
SE33
selected and were initially supplemented with 5 additional STRs representing the new Core CODIS set. Study of
Kinship analysis
Likelihood ratio
the LR value showed that when false matches were treated as paternity matches, the expansion of the marker set
Identity-by-state severely diminished the LR values obtained compared to true matches and the false positive ratio of familial
database searching. When false matches were treated as full-sibling matches, the expansion to 20 STRs also
diminished the number of false matches and the corresponding LR values compared to true full-sibling cases, but
the effect was less dramatic. Addition of the SE33 marker, further promoted distinction between true and false
matches both in paternity and full-sibling cases. Counting the number of shared alleles presented improved
distinction efficiency between true and false matches after STR expansion to20 and 21 STRs but remains a less
valuable method of familial DNA database searching compared to LR.

1. Introduction The initial scope of database searching was to match unidentified

DNA profiles identified in crime scenes are nowadays routinely
searched against databases which include reference and evidence pro-
files from previously committed crimes, in order to identify the donors.
In 1997 the European Union invited all member states to establish their
national databases [1] and in 2001 determined the European Standard
Set (ESS) of 7 genetic loci (D8S1179, D21S11, D3S1358, TH01,
D18S51, vWA, FGA) [2] which was later expanded with 5 additional
loci (D1S656, D2S441, D10S1248, D22S1045, D12S391) [3]. The 2008
EU legislation based on the Prϋm Treaty required that all member-states
establish a national DNA database for the investigation of criminal of-
fences and share information for a cross-border combat against crime
[4]. On the other side of the Atlantic, the Act of 1994, authorized the
U.S states to maintain DNA databases which since 1998, included of-
ficially the 13 - CODIS Core Loci (D8S1179, D21S11, D5S818, CSF1PO,
D3S1358, TH01, D13S317, D16S539, TPOX, D18S51, vWA, D7S820,
FGA) [5] which were recently expanded to 20 with the addition of 7
new STRs (D1S1656, D2S441, D2S1338, D10S1248, D12S391,
D19S433 and D22S1045) [6].
DNA profiles found in crime scenes with their donors who may have
previously committed other crimes and for this reason their profiles are
already in the database. However, since this is not true for every of-
fender and there are cases where no match is retrieved, lab scientists
incorporated new strategies of indirect matching through database fa-
milial searching, a method that has already been used for missing
person identifications and humanitarian mass disasters [7–9]. In such
cases, “orphan” DNA profiles are deliberately searched against the da-
tabase with low stringency criteria in pursue of first-degree biological
relationships (father-child, full-siblings) [10]. This approach is based on
findings from several studies that have linked crime and family
[11–16].
Familial database searching has been used successfully in several
cases [17–22]. In addition, many studies have been conducted in this
field, based mainly on simulations, in order to evaluate the usefulness
of this approach (prior to CODIS expansion) and to determine the most
appropriate strategy for optimal results, as there are a lot of parameters
that should be taken into account during searching, including the
strategy that should be followed (rare alleles, number of shared alleles

⁎
Corresponding author.
E-mail address: karantzali@dnalogy.eu (E. Karantzali).

https://doi.org/10.1016/j.fsigen.2019.07.008
Received 5 November 2018; Received in revised form 10 July 2019; Accepted 12 July 2019
Available online 27 July 2019
1872-4973/ © 2019 Elsevier B.V. All rights reserved.
E. Karantzali, et al.
or likelihood ratio (LR) value), the number of STRs used and the
thresholds that should be set [23–30]. All studies support familial
searching as an invaluable tool. Recently, the SWGDAM Ad Hoc
Working Group issued recommendations on familial searching in order
to minimize the false positive results during search, an operational goal
of familial searching [10].
However, several ethical and other considerations have arisen from
the use of familial searching, concerning mainly the privacy of those
already in databases, their close relatives and unrelated people that
may be put under the authorities’ microscope due to false positive
matching, targeting of ethnic minorities that are over-represented in
databases and waste of crucial investigation time [27,31,32]. In an
effort to minimize the number of adventitious matches during database
search and avoid these phenomena but also to increase international
compatibility and the power of discrimination for criminal and missing
person cases, the Federal Bureau of Investigation (FBI) CODIS Core Loci
Working Group announced in 2015 the expansion of CODIS Core Loci
from 13 to 20 [6].
The aim of the current work is to study the effect of the CODIS Core
Loci expansion on the number of adventitious matches obtained in a
familial DNA database search. 100 pairs of true and false paternity
matches obtained after familial DNA database searching, were ex-
panded from 15 to 20 STRs representing the new Core CODIS set. False
matches were also treated as full-sibling matches and were compared to
true full-sibling cases handled in the laboratory. Changes in the LR
value and the number of shared alleles were studied. Further con-
tribution of SE33 was also evaluated for potential future inclusion.
2. Materials and methods
2.1. DNA database searching
A database of 6000 profiles which have been generated with the use
of AmpFlSTR® Identifiler® Plus PCR Amplification Kit (Applied
Biosystems) was used. The profiles come from unidentified samples (no
personal data of the donors are registered in the database) and belong
to paternity, biological relationship and DNA identification cases han-
dled in the laboratory. 1000 random profiles from the database (1000
successive entries, random starting point) were searched against the
database to find matches with at least one allele shared at each locus.
100 pairs of one true and one adventitious match originating from the
same searched profile were selected for further study. True matches
were identified by the profile’s unique code which includes the case
number and type (paternity, full-siblings etc.). All matches selected
were between males which allows for a validation of true and ad-
ventitious matches through Y-STR analysis without adding any bias to
the study.
2.2. PCR amplification and capillary electrophoresis
15-STR database profiles were generated using the AmpFlSTR®
Identifiler® Plus PCR Amplification Kit (TFS, Walham, MA, USA).
Supplementary loci (D1S1656, D2S441, D10S1248, D12S391,
D22S1045, SE33) were provided after analysis with the AmpFlSTR™
NGM SElect™ PCR Amplification Kit (TFS). Penta_E and Penta_D were
provided from PowerPlex® 16 HS System (Promega, Madison, WI,
USA). Y-STR analysis was performed with the Yfiler kit (TFS). PCR
amplification was performed on a GeneAmp 9700 thermal cycler (TFS)
according to manufacturer’s protocols. CE analysis was conducted on an
ABI PRISM 3130xl Genetic Analyzer (TFS).
2.3. Statistical analysis
Database search, LR calculations and counting of shared alleles were
performed with the GeneMarker® HID v2.4.0 software. Caucasian allele
frequencies were used for LR calculations [33–35]. Likelihood ratio
2
Forensic Science International: Genetics 43 (2019) 102129
(LR) is the probability of data given that the first comparing hypothesis
is true, divided by the probability of data given that the second com-
paring hypothesis is true: LR = P(data|Hyp 1)/P(data|Hyp 2) for two
comparing hypotheses Hyp1 vs Hyp2. The hypotheses considered for LR
calculation regarding paternity matches were father-child vs. unrelated
and regarding full-siblings vs. unrelated for full-sibling matches. No
mutations were considered during searching but mutation rates from
AABB were used for LR calculations after STR expansion [36]. A
θ = 0.01 value was used for calculations to allow for population sub-
structuring. 95% Confidence Intervals calculation was based on the
exact binomial test [37,38]. The Estimated Kinship Ratio (EKR) is the
Likelihood Ratio (LR) divided by the sample size (N) of the searched
database, EKR = LR/N [39]
3. Results
3.1. Paternity testing
3.1.1. 15-STR database search
1000 genetic profiles of 15 STRs (13 CODIS loci plus D19S433 and
D18S1338), were randomly selected from a database of 6000 profiles
and were used in a familial DNA database search against the database,
for matches sharing at least one allele per locus. 348 profiles returned at
least 1 and up to 12 adventitious matches (Suppl. Excel 1) while the rest
652 profiles did not return any adventitious matches. In total, 555
adventitious matches were observed, 100 of which were selected for
further study. The selection was made on the basis that the adventitious
match should be between males and that the searched profile should
also return a true paternity match with a male person, serving as a
control. All true matches presented the same Y-STR haplotype (single
locus mismatches reported in 6 cases were attributed to one-step mu-
tational events) while all adventitious matches had different Y-STR
haplotypes.
The LR values for true and adventitious matches were calculated for
the Caucasian population, as this population is the reference population
for nearly all of the samples analyzed. As expected, true matches show
in general greater LR values than adventitious matches (Fig. 1). By
comparing each adventitious match with the corresponding true match,
4% present a greater LR value than their paired true match, 17% pre-
sent a LR value of the same order of magnitude and 79% present lower
values.
True and adventitious matches present a great overlap in their LR
values (Fig. 1); 86% of the adventitious matches overlap with 83.8% of
the true matches, making it nearly impossible to separate true from
false matches based on the LR value at the level of 15 STRs. By applying
a statistical threshold of LR ≥ 102, which is the lowest LR value ob-
tained for true matches, all true matches would be included but 86% of
false matches would be included as well (Figs. 1, 2A). A higher
threshold, LR ≥ 103, would still include almost all of the true matches
(96.5%) while the number of false matches would lower down to 48%,
which however, is still a very high percentage. A threshold of LR ≥ 104
would include only 14% of the false matches but a lot of true matches
would now be excluded, lowering the percentage above threshold to
77.6%. True positive rates (TPR) and false positive rates for all LR
thresholds are presented in Fig. 2A.
The SWGDAM Ad Hoc Committee on Partial Matches [39] sug-
gested calculation of the Estimated Kinship Ratio (EKR) as a different
approach in order to focus on matches that are more likely to represent
related pairs of people. This is a modification of the standard kinship
index which adjusts for database size searched. In our analysis EKR
calculation estimates that all true matches and 86% of false matches are
more likely to represent true relatives (Suppl. Excel 4). These findings
show that the use of the specific set of 15 STR markers cannot sa-
tisfactorily distinguish between true and adventitious matches based on
the LR value.
E. Karantzali, et al.
Fig. 1. LR values of true and adventitious paternity matches with the use of 15,
20 and 21 STRs. Vertical lines represent LR thresholds. The curves are estimated
density functions based on kernel smoothing. 58 and 100 datapoints were used
respectively.
3.1.2. Expansion to 20 CODIS Core STR Loci
True and adventitious matches described above were supplemented
with 5 additional loci (D1S1656, D2S441, D10S1248, D12S391,
D22S1045), to represent the newly proposed expanded Core CODIS set,
in an attempt to study the effect on the list of candidates acquired with
the use of 15 STRs. At the level of 20 STRs, all true paternities con-
tinued to share at least one allele in all loci, raising their LR values and
shifting their LR distribution to the right (Fig. 1). On the other hand,
93% of the adventitious matches showed at least 1 and up to 4 locus
mismatches while 7% of the adventitious matches continued to share at
least one allele in all 20 loci. Pairs with 3 or 4 locus mismatches (41%)
were excluded as candidates (LR = 0). As shown in Fig. 1, the degree of
overlap between true and adventitious matches is substantially dimin-
ished after expansion to 20 STRs; 10% of the adventitious matches
overlap with 34% of the true matches.
By applying a threshold of LR ≥ 103, which is the lowest LR value
obtained for true matches, only 10% of the adventitious matches are
included as candidates while all true matches are above threshold
(Fig. 1; Fig. 2A). Higher thresholds (LR ≥ 104; LR ≥ 105) retain most of
the true matches (96.5% and 93.3% respectively) while lowering the
number of false matches to 6% and 1%. The EKR estimates that all true
matches but only 14% of false matches are more likely to represent true
relatives (Suppl. Excel 4), supporting a better distinction between true
and false matches, reflected in our findings as well.
3.1.3. Further expansion with the addition of SE33 marker
Expansion from 15 to 20 STRs had a dramatic effect in the number
of candidates that would be selected for further investigation after a
3
Forensic Science International: Genetics 43 (2019) 102129
familial DNA database search. We expanded the set of loci even more,
by adding the highly polymorphic SE33 [40–44] and repeated the
statistical analysis. 75% of the adventitious matches showed a mis-
match in the SE33 locus and this raised the number of false matches
that were excluded, from 41% (20 STRs) to 71% (3, 4, 5 mismatches,
LR = 0) (Fig. 1; Suppl. Excel 1). A threshold of LR ≥ 104, which is the
lowest LR value obtained for true matches, included all true matches
and only 4% of the adventitious matches (Fig. 1; Fig. 2A). Addition of
SE33 in true matches did not result in any locus mismatches and in-
creased the LR value in 75% of the cases (Suppl. Excel 2) presenting a
direct opposite effect compared to adventitious matches. The overlap
between true and adventitious matches was slightly lowered, with 4%
of the adventitious matches overlapping with 34% of the true matches.
The EKR estimates that all true matches are more likely to represent
relatives while the respective number of false matches is only 7%, re-
presenting an even greater distinction between true and false matches
(Suppl. Excel 4).
Most surprisingly, 2 adventitious matches continued to share at
least one allele in all 21 loci and had a high LR value in the order of 105.
Since Y-STR analysis had excluded these matches as true paternities,
full-siblings or any other patrilineally connected relatives, increased
compatibility is most likely coincidental as has been also reported in
other studies [22]. More distant biological relationships could also be
an explanation [45] even though distant pairwise relationships share a
relatively low number of alleles and present low LR values [46]. Further
analysis with the use of Penta_D and Penta_E finally revealed one
mismatch in both cases.
3.2. Full-sibling testing
3.2.1. 15-STRs
The adventitious matches obtained after database search described
above could also represent candidate pairs of full-siblings, since it is not
unusual for siblings to share at least one allele in each of the 15 STR
loci. The LR values for the 100 adventitious matches were re-calculated
for full-siblings (SI). As these matches share at least one allele at each
locus, the analysis includes pairs with elevated numbers of shared al-
leles, (at least 15 alleles in common while true full-siblings may share
less alleles), which in turn may elevate the LR values as well. This kind
of matches is expected to demand stricter thresholds in order to be
distinguished from true matches.
30 pairs of true full-siblings that have been analyzed in the la-
boratory were used as positive controls (Suppl. Excel 3). Similarly to
paternity matches, true and adventitious full-sibling matches present a
great overlap in their LR values (Suppl. Fig. 1). As full-sibling testing
can present low LR values depending on the degree of similarity be-
tween siblings, the lowest statistical threshold in favor of a relationship,
LR = 10, was used to generate an initial list of candidates. By applying
this threshold, 96.7% of true siblings show LR values above threshold
while 58% of the adventitious matches are also included (Suppl. Fig. 1,
Fig. 2A). If a stricter threshold is chosen, LR ≥ 102, then the number of
false matches would be lowered down to 24% but in parallel 10% of
true matches would also be ignored while for LR ≥ 103 the true and
false positive rates would be 76.7% and 7% respectively. True positive
rates (TPR) and false positive rates for all LR thresholds are presented in
Fig. 2A. EKR estimates that 93.3% of true and 24% of false matches are
more likely to represent true full-siblings (Suppl. Excel 4).
3.2.2. Loci expansion to 20 CODIS STRs
LR values for true and adventitious full-sibling matches were re-
calculated after addition of 5 STRs to represent the new Core CODIS set
(Suppl. Excel 3). 61.3% of true full-siblings increased their LR value
while 72% of the adventitious matches presented a decrease, shifting
their LR distribution to the right and left respectively (Suppl. Fig. 1).
After application of the same statistical thresholds LR ≥ 10, LR ≥ 102
and LR ≥ 103, the number of adventitious matches decreased from 33%
E. Karantzali, et al. Forensic Science International: Genetics 43 (2019) 102129

Fig. 2. A. TPR was plotted against FPR for paternity (PT) and full-sibling (SI) matches based on LR values (PT-LR, SI-LR) and number of shared alleles (PT-IBS, SIIBS)
for 15, 20 and 21 STRs. (values are available in Suppl. Excel 4). B. TPR was plotted against FPR for paternity (PT) and full-sibling (SI) matches based on combinations
of LR values and number of shared alleles (LR-IBS), for 15, 20 and 21 STRs. Uncombined LR and IBS values have been added for comparison. (values are available in
Suppl. Excel 4).

to 14% and 1% respectively while 96.7%, 93.3% and 90% of true
matches were included (Suppl. Fig. 1; Fig. 2A), providing a better se-
paration compared to 15 STRs. The elevated discriminative power
provided by the use of 20 STRs is supported by the EKR as well which
estimates that 96.7% of true matches and 14% of adventitious matches
are more likely to represent true full-sibling pairs (Suppl. Excel 4).
3.2.3. Further expansion with the addition of SE33 marker
Addition of SE33 decreased the LR value of 50.5% of the ad-
ventitious matches while increased the LR value of 55% of the true
matches. All true matches presented LR values > 10, compared to 24%
of adventitious matches (Suppl. Fig. 1; Fig. 2A). A higher threshold
(LR ≥ 102) included only 8% of the adventitious matches and almost all
true matches (96.7%). Further reduction of the percentage of false
matches to 5.1% was achieved with a threshold of LR ≥ 103 but with a
concomitant reduction of true matches to 90%. The EKR supported a
further distinction between true and false matches including 100% of
true matches vs. only 8% of false matches (Suppl. Excel 4).
3.3. Number of shared alleles (IBS)
A second comparison regarding the number of common alleles (IBS;
Identity by State) in each match was also performed. This approach is
independent of the type of biological relationship or the relevant po-
pulation but it has a lower discrimination power than the LR approach
[27,45]. At the level of 15 STRs, a nearly complete overlap in the
number of shared alleles of true and adventitious matches was ob-
served, both for paternity and full-sibling matches (Fig. 3), which was
expected, as the database search required at least one shared allele in
each locus.
After expansion to 20 STRs, true matches show in average greater
values of shared alleles than adventitious matches. The degree of
overlap is substantially diminished for paternity matches while a
slighter effect was observed for full-sibling matches. For 19 and 20

4
E. Karantzali, et al.
Fig. 3. Number of shared alleles of true and adventitious paternity (A) and full-sibling (B) matches with the use of 15, 20 and 21 STRs. 58, 30 and 100 datapoints
were used for true paternities, true full-siblings and false matches respectively.
alleles IBS, almost all true paternity and full-sibling matches are in-
cluded along with more than half of false matches (Figs. 2A, 3). For
higher thresholds (IBS ≥ 21;≥22;≥23), the percentage of false mat-
ches was 45%, 24% and 13% respectively, while the percentage of true
paternities was 96.5%, 93.1% and 86.2% and the percentage of true
full-siblings has dropped < 90% at all cases. TPR and FPR for all IBS
thresholds are presented in Fig. 2A. Addition of SE33 did not provide
any better distinction, showing that even at the level of 21 STRs,
counting the number of shared alleles remains less discriminative
compared to the LR method. However, it is shown that CODIS expan-
sion is able to induce a remarkable degree of separation between true
and adventitious matches. As the number of STRs increases (15, 20, 21
STRs), the degree of separation increases as well, which means that the
method could possibly be more effective if more than 21 markers are
analyzed.
3.4. Combination of LR and IBS thresholds
Ge et al. [28] suggested that the LR values and the number of shared
alleles can be combined to improve the accuracy of searching by ex-
cluding a substantial proportion of the unrelated profiles with only a
relatively small proportion of relatives. This suggestion received severe
criticism by Balding et al [29] who supported that the LR method is the
most powerful statistic for distinguishing between two hypotheses and
that its efficiency cannot be enhanced by the IBS method. Moreover,
they questioned the gain from the decreased false positives when this is
accompanied by an increased number of false negatives. In our study,
we combined the two methods and studied TPR and FPR for paternity
and full-sibling matches at the level of 15, 20 and 21 STRs (Fig. 2B).
Joint distributions are presented in Suppl. Table 1 and Suppl. Table 2.
For paternity matches including 15 STRs, a threshold of LR ≥ 102
combined with IBS ≥ 15, resulted in the same rates as using the LR
threshold alone (TPR:100%; FPR:86%). By raising the IBS threshold
to ≥ 16, both rates are only slightly decreased (TPR:96.5%; FPR:79%)
5
Forensic Science International: Genetics 43 (2019) 102129
while an even higher IBS threshold (≥17) reduces TPR to 84.5% and
FPR to 51%. The same observations were made when a threshold of
LR ≥ 103 was used. An IBS ≥ 15 threshold, resulted in the same rates as
using the LR threshold alone (TPR:96.5%; FPR:48%), an IBS ≥ 16
threshold slightly decreased both rates (93.1%; 44%) while an IBS ≥ 17
threshold reduced TPR to 81% and FPR to 28%. In accordance with the
findings of Ge et al. [28], we observed that a remarkable reduction of
FPR is feasible when the two methods are combined but only if a
proportion of true matches is also excluded. When a higher LR
threshold (LR ≥ 104) at the level of 15 STRs or LR thresholds at the
level of 20 and 21 STRs were used - where in both cases most of the
false matches have been already excluded - further addition of an IBS
threshold would exclude a large proportion of true matches in order to
achieve only slight reduction in the number of false matches. The re-
sults were similar for full-sibling matches. No reduction was observed in
FPR without a concomitant, slight or large, reduction in TPR.
4. Discussion
In the present study, familial DNA database searching was per-
formed using 1000 15-STR profiles registered in the laboratory’s data-
base of 6000 profiles, which were searched against the database for
matches sharing at least one allele per locus. 100 pairs of one true and
one false match including the same searched profile were chosen for
further study. LR calculation for parent-child vs unrelated led to the
conclusion that the use of the specific set of markers is not a highly
discriminative method, as the LR values presented a great overlap and
21% of the adventitious matches presented the same or greater LR value
than the paired true match. It has been shown that the percentage of
true relatives ranked as #1 decreases as the size of the database in-
creases [6]. The above findings show that even with a small database, a
true relative is not always ranked as #1 (though ranking is out of the
scope of the present study). Even after the application of statistical
thresholds, in an effort to control the number of false positives, either a
E. Karantzali, et al.
large proportion of false positives were included or a proportion of true
positives were excluded. Addition of 5 genetic loci to represent the new
Core CODIS set of 20 STRs, transformed a large number of false matches
to paternity exclusions and severely diminished the degree of LR value
overlap between true and adventitious matches. Application of several
statistical thresholds produced high TPR/FPR ratios with ≥ 90% of true
matches and ≤ 10% false matches included in the lists generated. These
ratios could be tolerable in a familial DNA database search, in combi-
nation with the fact that adventitious matches present the lowest LR
values, as investigation would go sequentially through a ranked list of
candidates, sorted by the LR values. Expansion to 21 markers with the
addition of the highly polymorphic SE33 further excluded a large
number of false matches by adding an additional locus mismatch and
increased the TPR/FPR ratio with no loss of true matches, making SE33
a very valuable marker and an exceptional choice for complementary
use or future expansion of CODIS. Surprisingly, 2 adventitious matches
continued to share at least one allele in all 21 markers with high LR
values, supporting the fact that adventitious matches with high LR
values is a possibility even when a great number of STRs has been
analyzed, though a more distant biological relationship cannot be ex-
cluded. The Y-STR analysis used for validation of true and adventitious
matches in the present study was exclusionary for all adventitious
matches supporting the use of Y-STR analysis for further refinement of
the candidate list, where applicable.
The LR values of the 100 adventitious matches were recalculated for
full-sibling testing and were compared to 30 full-sibling pairs that had
been tested positive in the laboratory. Similarly to paternity testing, at
the level of 15 STRs the degree of overlap between true and false
matches was very high and the application of any statistical threshold
either included a large number of false matches or excluded true mat-
ches. This is in accordance to previous studies [47] where the use of the
specific set of 15 STR markers gave low discriminative power between
the true and false full-sibling matches. Expansion to 20 and 21 STRs
reduced the LR value overlap between true and false matches and in-
creased the TPR/FPR ratio, however, the effect was slighter compared
to the paternity matches.
Population substructure was incorporated to LR calculations with a
use of a θ = 0.01. A comparison between 0 and 0.01 theta values has
been shown to induce a 10-fold decrease in the LR values when the
more conservative 0.01 value is used [28]. No mutations were taken
into account during database searching while mutation rates were used
after expansion in the number of STRs. The incorporation of mutation is
expected to have a slight effect on LR values (< 5%) [44].
The study on the number of common alleles led to the conclusion
that this approach is characterized by a lower discrimination power
compared to the LR method. However, the overlap between true and
false matches decreased as the number of STRs increased, both in pa-
ternity and full-sibling cases, which means that if a larger number of
STRs is analyzed, there is a possibility that this method can also be used
more effectively. A combination of LR and IBS thresholds was better
than the IBS method alone but could not improve the use of the LR
method alone as any decrease in the number of false matches was al-
ways accompanied by a decrease in the number of true matches as well.
Overall, based on a true database, the findings of the present study
showed that expansion of the number of STRs from 15 to 20 or 21, can
greatly improve distinction between true and false matches during a
DNA database search and are in support of the idea that the CODIS Core
Loci expansion from 13 to 20 STRs, will severely diminish the number
of adventitious matches obtained in familial DNA database searching.
This effect will definitely save a lot of investigation time and will aid in
the protection of privacy of suspects’ family members and innocent
people that would be investigated based on database search findings, a
major ethical consideration against the method.
The strategy used in the present study applied a fixed LR threshold.
However, other LR-based strategies for familial database searching
have also been studied (top-k, profile-centered, conditional) [30]. Even
6
Forensic Science International: Genetics 43 (2019) 102129
though the fixed threshold has been characterized as the most efficient,
the choice of the suitable strategy depends on each lab’s requirements,
e.g. length of the candidate list, probability to detect a true relative etc.
No method can be used alone to include all true matches and exclude all
false matches and as a result, finding the best-suited strategy in pursuit
of this goal requires combination of several data (database search hits,
additional case data, Y-STR analysis, gender, geographical details, age
etc). Refinement of the candidate list with the use of additional STRs is
a promising future possibility for improved performance.
Declaration of Competing Interest
The authors declare no conflict of interest
Acknowledgements
This research did not receive any specific grant from funding
agencies in the public, commercial, or not-for-profit sectors.
Appendix A. Supplementary data
Supplementary material related to this article can be found, in the
online version, at doi:https://doi.org/10.1016/j.fsigen.2019.07.008.
References
[1] EU - Council Resolution 97/C 139/02 https://eur-lex.europa.eu/legal-content/EN/
TXT/PDF/?uri=CELEX:31997Y0624(02)&from=EN.
[2] EU - Council Resolution 2001/C 187/01 https://eur-lex.europa.eu/legal-content/
EN/TXT/PDF/?uri=CELEX:32001G0703(01)&from=EN.
[3] EU - Council Resolution 2009/C 296/01 https://eur-lex.europa.eu/legal-content/
EN/TXT/PDF/?uri=CELEX:32009G1205(01)&qid=1550671331259&from=EN.
[4] EU - Council Decision 2008/615/JHA https://eur-lex.europa.eu/legal-content/EN/
TXT/PDF/?uri=CELEX:32008D0615&from=EN.
[5] Pub. L. No. 103-322 (1994). https://www.congress.gov/103/bills/hr3355/BILLS-
103hr3355enr.pdf.
[6] D.R. Hares, Selection and implementation of expanded CODIS core loci in the
United States, Forensic Sci Int Genet. 17 (July) (2015) 33–34.
[7] C.H. Brenner, B.S. Weir, Issues and strategies in the DNA identification of World
Trade Center victims, Theor Popul Biol. 63 (3) (2003) 173–178 May.
[8] C.H. Brenner, Some mathematical problems in the DNA identification of victims in
the 2004 tsunami and similar mass fatalities, Forensic Sci Int. 157 (2–3) (2006)
172–180 March 10.
[9] B. Budowle, F.R. Bieber, A.J. Eisenberg, Forensic aspects of mass disasters: strategic
considerations for DNA-based human identification, Leg Med (Tokyo). 7 (July 4)
(2005) 230–243.
[10] Recommendations from the SWGDAM Ad Hoc Working Group on Familial
Searching, (2014).
[11] D.C. Rowe, D.P. Farrington, The familial transmission of criminal convictions, 35
Criminology (1997) 177–201.
[12] N. Gregory, Crime and the family: like grandfather, like father, like son? Br. J.
Forensic Pract. 6 (4) (2004) 32–36.
[13] D.P. Farrington, G.C. Barnes, S. Lambert, The concentration of offending in families,
Leg. Crim. Psychol. 1 (1) (1996) 47–63.
[14] D.P. Farrington, S. Lambert, D.J. West, Criminal careers of two generations of fa-
mily members in the Cambridge study in delinquent development, Stud. Crime
Crime Prev. 7 (1) (1998) 85–106.
[15] M. Van De Rakt, P. Nieuwbeerta, N.D. De Graaf, Like father, Like son: the re-
lationships between conviction trajectories of fathers and their sons, Br. J. Criminol.
48 (4) (2008) 538–556.
[16] D.P. Farrington, J.W. Coid, Early Prevention of Adult Antisocial Behaviour,
Cambridge University Press, 2003.
[17] The hunt for the Saturday Night Strangler https://www.theguardian.com/uk/
2003/jan/18/wales.kevintoolis.
[18] Richard Willing, USA TODAY, Suspects get snared by a relative’s DNA, 7 June
(2005) https://usatoday30.usatoday.com/news/nation/2005-06-07-dna-cover_x.
htm.
[19] BBC News, How police found Gafoor, 4 July 2003 (2003) http://news.bbc.co.uk/2/
hi/uk_news/wales/3038138.stm.
[20] The Guardian, DNA traps brick thrower who killed lorry driver, 20 April, 2004
(2004) https://www.theguardian.com/uk/2004/apr/20/ukcrime1.
[21] The Yorkshire Post, Seven years for taxi rapist trapped by family DNA, 24 Feb 2006
(2006) https://www.yorkshirepost.co.uk/news/seven-years-for-taxi-rapist-trapped-
by-family-dna-1-2605538.
[22] The Telegraph, Trapped after 20 years when sister had a visit from the cold case
detectives, 18 July 2006 (2006) https://www.telegraph.co.uk/news/uknews/
1524208/Trapped-after-20-years-when-sister-had-a-visit-from-the-cold-case-
detectives.html.
E. Karantzali, et al.
[23] F.R. Bieber, C.H. Brenner, D. Lazer, Finding criminals through DNA of their re-
latives, Science. 312 (5778) (2006) 1315–1316 Jun 2.
[24] S.P. Myers, M.D. Timken, M.L. Piucci, G.A. Sims, M.A. Greenwald, J.J. Weigand,
K.C. Konzak, M.R. Buoncristiani, Searching for first-degree familial relationships in
California’s offender DNA database: validationof a likelihood ratio-based approach,
Forensic Sci Int Genet. 5 (5) (2011) 493–500 Nov.
[25] K.L. O’Connor, E. Butts, C.R. Hill, J.M. Butler, P.M. Vallone, Evaluating the effect of
additional forensic loci on likelihood ratio values for complex kinship analysis,
Proceedings of the 21st International Symposium on Human Identification. San
Antonio, TX, 10-14 October, (2010).
[26] C.N. Maguire, L.A. McCallum, C. Storey, J.P. Whitaker, Familial searching: a spe-
cialist forensic DNA profiling service utilising the National DNA Database to iden-
tify unknown offenders via their relatives–the UK experience, Forensic Sci Int
Genet. 8 (1) (2014) 1–9 Jan http://media.wix.com/ugd/4344b0_
46b5263cab994f16aeedb01419f964f6.pdf.
[27] T. Hicks, F. Taroni, J. Curran, J. Buckleton, V. Castella, O. Ribaux, Use of DNA
profiles for investigation using a simulated national DNA database: Part II.
Statistical and ethical considerations on familial searching, Forensic Sci Int Genet. 4
(5) (2010) 316–322 Oct.
[28] J. Ge, R. Chakraborty, A. Eisenberg, B. Budowle, Comparisons of familial DNA
database searching strategies, J. Forensic Sci. 56 (6) (2011) 1448–1456.
[29] D.J. Balding, M. Krawczak, J.S. Buckleton, J.M. Curran, Decision-making in familial
database searching: KI alone or not alone? Forensic Sci Int Genet. 7 (1) (2013)
52–54 Jan.
[30] M. Kruijver, R. Meester, K. Slooten, Optimal strategies for familial searching,
Forensic Sci Int Genet. 13 (2014) 90–103 Nov.
[31] C.J. Gershaw, A.J. Schweighardt, L.C. Rourke, M.M. Wallace, Forensic utilization of
familial searches in DNA databases, Forensic Sci Int Genet. 5 (1) (2011) 16–20 Jan.
[32] R.V. Rohlfs, E. Murphy, Y.S. Song, M. Slatkin, The influence of relatives on the
efficiency and error rate of familial searching, PLoS One. 8 (8) (2013)
e70495Aug 14.
[33] J.M. Butler, R. Schoske, P.M. Vallone, J.W. Redman, M.C. Kline, Allele frequencies
for 15 autosomal STR loci on U.S. Caucasian, African American, and Hispanic po-
pulations, J Forensic Sci. 48 (4) (2003) 908–911 Jul.
[34] AmpFlSTR®NGM SelectTM PCR amplification kit, User guide, Life Technologies
Corporation, Carlsbad, CA, http://tools.thermofisher.com/content/sfs/manuals/
Forensic Science International: Genetics 43 (2019) 102129
cms_089008.pdf.
[35] C.R. Steffen, M.D. Coble, K.B. Gettings, P.M. Vallone, Corrigendum to’ U.S.
Population Data for 29 Autosomal STR Loci’ [Forensic Sci. Int. Genet. 7 (2013) e82-
e83], Forensic Sci. Int. Genet. 31 (2017) e36–e40.
[36] AABB annual report 2003, (2019) ptannrpt03.pdf (Accessed March 4 2019) https://
www.aabb.org/sa/facilities/Documents/.
[37] C. Clopper, S. Pearson, The use of confidence or fiducial limits illustrated in the case
of the Binomial, Biometrika 26 (1934) 404–413.
[38] B.V. Gnedenko, I.A. Ushakov, I.V. Pavlov, Statistical Reliability Engineering, Wiley,
John & Sons, 1999 April.
[39] Scientific Working Group on DNA Analysis Methods Ad Hoc Committee on Partial
Matches, SWGDAM Recommendations to the FBI Director on the “Interim Plan for
the Release of Information in the Event of a ‘Partial Match’ at NDIS, Forensic Sci
Comm 11 (4) (2009).
[40] J.M. Butler, C.R. Hill, M.C. Kline, D.L. Duewer, C.J. Sprecher, R.S. McLaren,
D.R. Rabbach, B.E. Krenke, D.R. Storts, The single most polymorphic STR Locus:
SE33 performance in U.S. populations, Forensic Sci. Int. Genet. Suppl. Ser. 2 (2009)
23–24.
[41] P. Wiegand, B. Budowle, S. Rand, B. Brinkmann, Forensic validation of the STR
systems SE 33 and TC 11, Int. J. Legal Med. 105 (6) (1993) 315–320.
[42] B. Rolf, M. Schürenkamp, A. Junge, B. Brinkmann, Sequence polymorphism at the
tetranucleotide repeat of the human beta-actin related pseudogene H-beta-Acpsi-2
(ACTBP2) locus, Int. J. Legal Med. 110 (1997) 69–72.
[43] M.H. Polymeropoulos, D.S. Rath, H. Xiao, C.R. Merrill, Tetranucleotide repeat
polymorphism at the human beta-actin related pseudogene H-beta-Ac-psi-2
(ACTBP2), Nucleic Acids Res. 20 (1992) 1432.
[44] A. Lászik, P. Sótonyi, S. Rand, C. Hohoff, Frequency data for the STR locus ACTBP2
(SE33) in eight populations, Int. J. Legal Med. 115 (2001) 94–96.
[45] T.M. Reid, M.L. Baird, J.P. Reid, S.C. Lee, R.F. Lee, Use of sibling pairs to determine
the familial searching efficiency of forensic databases, Forensic Sci Int Genet. 2 (4)
(2008) 340–342 Sep.
[46] J. Ge, B. Budowle, R. Chakraborty, Choosing relatives for DNA identification of
missing persons, J Forensic Sci. 56 (Suppl. 1) (2011) S23–8 Jan.
[47] C.E. Pu, A. Linacre, Systematic evaluation of sensitivity and specificity of sibship
determination by using 15 STR loci, J. Forensic legal Med. 15 (5) (2008) 329–334.