Download On the effects of sampling, analysis and interpretation strategies for

UvA-DARE (Digital Academic Repository)
On the effects of sampling, analysis and interpretation strategies for complex forensic
DNA research with focus on sexual assault cases
Benschop, C.C.G.
Link to publication
Citation for published version (APA):
Benschop, C. C. G. (2012). On the effects of sampling, analysis and interpretation strategies for complex
forensic DNA research with focus on sexual assault cases
General rights
It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s),
other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).
Disclaimer/Complaints regulations
If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating
your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask
the Library: http://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam,
The Netherlands. You will be contacted as soon as possible.
UvA-DARE is a service provided by the library of the University of Amsterdam (http://dare.uva.nl)
Download date: 02 Aug 2017
Chapter 6
Assessment of mock cases involving
complex low template DNA mixtures:
A descriptive study
Manuscript submitted
Corina C.G. Benschop
Hinda Haned
Tanja J.P. de Blaeij
Alexander J. Meulenbroek
Titia Sijen
Abstract
Assessment of complex mock cases
Cases that rely on complex DNA mixtures with low template (LT) components are
among the most challenging cases to interpret and report. In this study, we designed
such mixtures and we describe how reporting officers (ROs) at the Netherlands
Forensic Institute (NFI) assess these when embedded in a mock case setting. DNA
mixtures containing LT DNA from two to four contributors, sporadic contamination
and/or DNA of relatives, were amplified four-fold using the Next Generation Multiplex
(NGM) kit. Consensus profiles were then generated which included the alleles
detected in at least half of the replicates. Four mock cases were created by including
reference profiles of a hypothetical victim and suspect. The mock cases were assessed
by eight ROs following the stepwise interpretation approach currently in use at the
NFI. With this approach, the results of the comparisons between the DNA profiles of
the evidentiary trace and the reference profiles are classified into four categories of
evidential value [1]. The interpretations by the ROs were compared to the likelihood
ratios (LRs) obtained from a probabilistic model that allows the interpretation of LT
DNA evidence and both were compared to the true composition of the designed
mixtures.
Chapter 6
131
Chapter 6
Chapter 6
Introduction
Forensic DNA analysis and interpretation of samples containing multiple
contributors is especially challenging when some components are low template (LT).
A decade ago, forensic DNA typing mainly dealt with (as what we would now call) high
template DNA samples. Since the development of low copy number (LCN) DNA
typing methods (e.g. [2-8]), and the release of highly sensitive STR typing kits [9-11]
we are approaching analysis down to the level of single cells. Drawbacks of STR typing
of minimal amounts of DNA are the occurrence of stochastic amplification artefacts
and alleles of sporadic contaminant(s). Stochastic amplification artefacts are welldefined and include allele drop-in, elevated stutter, heterozygote peak imbalance, allele
drop-out and locus drop-out [2,12]. A common strategy to deal with LT artefacts is to
use replicate analyses coupled with a consensus approach to infer the genotype(s) of
the individual(s) that contributed to the DNA profile, though the inferences may be
imperfect (e.g. have drop-outs) [2,13-16]. The consensus profile is then compared to
reference profiles provided within the case. In some instances, the profile is searched
against a DNA database. Both profile interpretation and profile comparison become
more complicated when dealing with LT DNA components due to the stochastic
amplification artefacts. Such artefacts may further provoke interpretation bias by ROs.
Therefore, stepwise guidelines were proposed that ensure that profile interpretation
occurs without prior knowledge of the reference DNA profile of e.g. the suspect
[1,17-21].
At the Netherlands Forensic Institute (NFI) the stepwise approach described
in Meulenbroek et al. [1] is used to ensure unbiased interpretation of DNA-based
evidence. This approach distinguishes four successive steps that are modified from
Clayton et al. [17]: (1) STR profiling and analysis of the peaks, (2) interpretation of
DNA profiles, (3) comparison of DNA profiles with categorisation of the evidential
value, and (4) considering the findings in the context of other facts within the criminal
case. In step 1, the alleles in the DNA profile are assigned by applying a detection
threshold and removing technical artefacts, such as spikes (bleed-through signals) and
blobs (dye residues). In step 2, the coherence of the alleles is established by estimating
the (minimum) number of contributors and the contribution of different individuals
and deducing, if possible, the genotype of the major and minor contributors. In step
3, the interpreted DNA profile of the evidentiary trace is compared to the reference
DNA profile of a person of interest (if available), which in many cases concerns
the suspect. A database search may have preceded this manual comparison. To avoid
interpretation bias, step 3 starts when step 2 is completed. In practice, the evidentiary
profile(s) may be re-evaluated based on the reference DNA profile of an expected
contributor prior to comparison to the reference DNA profile of the suspect. At
the NFI, the results of comparisons between the profile of the evidentiary trace and
132
Assessment of complex mock cases
133
Chapter 6
the reference profile(s) are classified into four categories of evidential value, i.e.: A:
exclusion; B: match with statistical evaluation; C: match without statistical evaluation;
D: cannot be included or excluded (Fig. 1). Ideally, the results are placed in one
category only. However, for complex DNA profiles and comparisons there may be
arguments for two categories. This holds specifically for categories B and C (‘match
with statistical evaluation’ and ‘match without statistical evaluation’), C and D (‘match
without statistical evaluation’ and ‘cannot be included or excluded’), but also A and
D (‘exclusion’ and ‘cannot be included or excluded’). There can be clear differences
in the juridical impact of for instance ‘exclusion’ or ‘cannot be included or excluded’.
Nevertheless, complex LT STR profiles can hold arguments for both these categories.
Basically the expert opinion in this evaluation process concerns the question ‘are the
alleles of the suspect that are not seen in the DNA profile of the evidence related to
absence of DNA of this donor or a result of allele drop-out?’. With steps 1 to 3 the
reporting officers (ROs) aim to infer which donor(s) contributed to the evidentiary
trace. Thus, these steps are source level-driven. The fourth step is activity level-driven;
the ROs consider the results of the profile comparisons in the context of the case and
evaluate two competing propositions invoking actions that have led to deposition of
the cell material. Qualitative probability assignments are generally used to present the
results of this evaluation [22,23].
In this study, we report on the assessment of complex mock cases involving LT
DNA mixtures by ROs of the NFI. We prepared four challenging DNA mixtures,
amplified these in four-fold using the Next Generation Multiplex (NGM) kit and
generated consensus profiles that included alleles detected in at least half of the
replicates (denoted the ‘n/2 consensus approach’ [15]). Based on these DNA profiles,
four mock cases were created by including reference DNA profiles of (an) expected
contributor(s) (e.g. victim, partner) and a hypothetical suspect. The four mock cases
were handed to eight ROs at the NFI. They individually assessed the complex DNA
profiles and examined whether the suspect may have contributed to the DNA mixture.
Then they classified the outcomes in the above described categories. In addition, for
each case likelihood ratios (LRs) were obtained with LRmix [24], which implements a
probabilistic model that allows the interpretation of LT DNA evidence [25]. Since we
used designed samples that have the advantage that the individual contributors to the
mixtures are known, the ROs conclusions and the LRs can be verified. The outcomes
are reported case by case.
Chapter 6
Materials and methods
Samples
DNA mixtures were prepared using established amounts of pristine or diluted
high template single donor DNA extracts of unrelated or related (brothers) donors
with known STR profiles [16]. The volunteering brothers gave informed consent to
use their DNA for this study, but not to transfer their profiles to third parties or to
publish their complete DNA profiles.
Chapter 6
STR typing, profile analysis and consensus approach
DNA mixtures were amplified in four-fold using the Next Generation Multiplex
(NGM) kit (AB) at 29 cycles following the manufacturer’s recommendations. As a
negative control, we used 10 µl dH2O, and as a positive control we used 3 µl (300 pg)
DNA007 (AB). All amplifications were performed on the same 9700 PCR apparatus
(AB). For capillary electrophoresis (CE) one ABI3130xl instrument was used applying
3 kV 15 s injection settings. CE mixtures contained 8.7 µl HiDi Formamide, 0.3 µl
LIZ500, and 1 µl of PCR products, or 1 µl of NGM allelic ladder. STR profiles were
analyzed using GeneMapper ID-X version 1.1.1 software (AB). The analysis was
performed using a detection (allele calling) threshold of 50 relative fluorescence
units (rfus) while a stochastic threshold of 400 rfu was used. Signals below detection
threshold cannot be reliably called a true allele.When peaks occur below the stochastic
threshold, this serves as an alert that not all of the DNA typing information may
have been detected. In house experiments generated different stochastic thresholds
under different conditions: 175 rfu applies to 3 kV 5 s CE injection settings, 300 rfu
to 3 kV 10 s and 400 rfu to 3 kV 15 s (Westen and Sijen unpublished results). We
applied back stutter (–1 repeat unit) ratios specific for each locus as recommended
by the manufacturer (AB) and a general forward stutter (+1 repeat unit) ratio of 2.5%
except for D22S1045 for which 7.36% was taken. These +1 repeat unit stutter ratios
were determined during an in-house NGM validation: since D22S1045 carries a trinucleotide repeat unit, a higher stutter ratio applies (Westen and Sijen unpublished
results).
The four independently amplified NGM profiles of each DNA mixture were used
to generate consensus profiles following the n/2 consensus method [15]; alleles that
were called in at least two of the four NGM profiles were assigned to the consensus
profile.
134
Assessment of complex mock cases
Profile comparisons and evidential value
The four mock cases were presented case by case to eight NFI ROs who had
no prior knowledge of the sample preparation. The ROs independently assessed the
profiles and classified the results using the four categories described in Fig. 1.To prevent
bias, the interpretation of the presented DNA evidence and the profile comparisons
were performed in separate steps. First, we handed the individual amplifications and
the consensus profile. Secondly, we handed the reference profile of the expected
contributor(s). At that stage the ROs re-evaluated the evidentiary profile to assess
alleles (genotypes) of the unknown contributor(s). Finally the reference profile of a
hypothetical suspect was given, and the ROs performed the profile comparisons. This
exercise was somewhat different than real casework analysis, as the ROs were: (1) at
that moment not yet familiar with NGM profiles, (2) restricted in time, (3) not allowed
to discuss with one another and (4) the cases were not reconsidered by a second RO.
Chapter 6
Figure 1. General guidelines for the classi�ication of the categories of evidential value.
135
Chapter 6
Chapter 6
To calculate the likelihood ratio (LR), we used the LRmix module of the R
package Forensim [24]. The implementation of LRmix follows the method of Gill et
al. [25]. LRmix accommodates parameters for the probability of drop-out (prD), the
probability of drop-in (prC), the allele frequencies and the number of contributors.
We used LRmix to obtain LRs using the consensus profiles of the cases as the sample
profiles. Reference profile(s) include the genotypes of the expected contributor(s)
(e.g. victim) and suspect. In addition, we assessed the LRmix outcomes for an
‘assembled profile’ which refers to the profile that is assembled from all reference
profiles of all contributors and is therefore devoid of drop-out or drop-in alleles.
With LRmix we weighed the prosecution (Hp) and defence (Hd) hypotheses for the
trace sample. Both considered the same number of contributors but Hp regarded
the suspect and Hd an unknown unrelated individual (next to, for both hypotheses,
the expected contributor(s) and in some calculations an additional unknown person).
For the assembled profile the Hp regarded the suspect and all other contributors,
while the Hd considered one unknown unrelated individual and all other contributors.
For both the assembled DNA profile and the trace sample we explored the LRs at
prDs ranging from 0.01 to 0.99 (at 0.01 intervals). Since the mixtures were prepared
from pristine and clean DNA samples a low prC is appropriate and set at a typical
low value of 0.01. In addition, for the assembled profile the ‘high template LR’ was
calculated by using a prD and a prC of 0.0 (which is relevant for this assembled profile
as it is devoid of drop-outs and drop-ins). The LRs for the trace and the assembled
profiles are presented on a log10 scale.
The statistical method used to estimate the number of contributors was the
maximum likelihood estimator (MLE), which was carried out using the Forensim
package for the R statistical software [26]. MLE takes into account both the number
of alleles per locus and the population’s allele frequencies. MLE makes no use of
quantitative information (e.g. PHs), and has proven to be useful as an indicator of the
number of contributors [16,26,27]. The RO’s used their expert training to determine
the number of contributors.
Mock case design and analytical examination of the
profiles
The details for the four mock cases are presented in Table 1. Each case contains
a set of profiles from a complex DNA mixture (four replicates and a consensus),
reference profiles of expected contributors (hypothetical victim plus in one case the
partner of the victim) and a reference profile of a hypothetical suspect.
136
Assessment of complex mock cases
Table 1. Overview of the set-up of the four mock cases.
Mock case 1
Mock case 2
The DNA mixture for mock case 2 was generated by combining 150 pg and 30 pg
of DNA of two brothers (Table 1). For the major contributor (donor 1), all alleles were
detected. For the minor contributor (donor 2), two (heterozygote) alleles were not
reproduced and thus not included in the consensus profile (Table 3). As with mock case
1, many PHs were below 400 rfu, which is indicative of the occurrence of stochastic
amplification artefacts. One non-donor allele was present in one of the individual
137
Chapter 6
The allele calls of the four individual NGM profiles and the consensus profile of
mock case 1 are shown in Table 2 and the analytical results are summarized in Table 3.
All alleles of the major contributor (150 pg) were detected, while some alleles of the
30 pg donor were missing and for the 6 pg donor only a few alleles were detected.
From the peak heights (PHs) (Table 3) it was deduced that stochastic amplification
artefacts such as drop-outs are to be expected since many of the detected alleles are
below 400 rfu which we determined to be the appropriate stochastic threshold.
To estimate the number of contributors we used the MLE method which resulted in
estimates of two contributors for three of the four individual profiles and the consensus
profile; only for the fourth individual profile the correct estimate of three contributors
was obtained. This latter profile is the only profile that shows five alleles at a (single)
locus (D2S441) (Fig. 2); all other profiles show a maximum of four alleles per locus.
Severe allele drop-out in each of these profiles explains the inability to detect the third
contributor from the number of alleles per locus. An extra contributor may also be
indicated by PH ratios that deviate from what is expected with the estimated number
of contributors and mixture proportion [28]. However, in these NGM profiles the PHs
do not reveal the presence of the third contributor; the heights of the (few) alleles of
donor 3 (the 6 pg contributor) are in the same range as the (larger number of) alleles
of donor 2 (the 30 pg contributor) (Fig. 2). Although without prior knowledge of the
true contributions one cannot be certain whether two or three donors contributed,
one can ascertain that the minimum number of contributors of DNA to the mixture
is two. When setting up the mock case, two reference profiles were prepared: donor
1 was presented as the victim (the expected contributor) and donor 2 as the suspect.
Table 2. Overview of the alleles present in the NGM pro�iles of the evidentiary trace, the consensus pro�ile and the reference
pro�iles of donor 1 (victim), donor 2 (suspect) and donor 3 (unknown contributor for which only 6 pg DNA was added,
Table 1) of mock case 1. Yellow cells represent alleles that do not correspond to the victim and match an allele of the suspect.
Blue cells mark alleles that cannot be explained by the two reference pro�iles (victim and suspect) that were handed to the
ROs. Red cells represent alleles of the suspect that are missing in the consensus pro�ile.
Chapter 6
Chapter 6
138
Chapter 6
Table 3. Analytical results per contributor for the four NGM ampli�ications (amp. 1 to amp. 4) and consensus pro�ile for each
mock case.
Assessment of complex mock cases
139
Figure 2. PET-dye channel of the electropherograms obtained from four (A-D) NGM ampli�ication of the DNA mixture of mock
case 1. Closed peaks represent alleles of donor 1 (victim). Stars indicate alleles that are in accordance with the reference
pro�ile of the donor 2 (suspect). Due to the LT character of these mixtures (resulting in PHs below the stochastic threshold
of 400 rfu), several alleles of donor 2 have dropped out (indicated by dashes). Arrows point at alleles that are ampli�ied from
the third donor that contributed 6 pg of DNA.
Chapter 6
Chapter 6
140
Assessment of complex mock cases
profiles: a peak positioned between two ‘true’ alleles (thereby most likely representing
a stutter drop-in). Notwithstanding this non-donor allele, for all profiles the number of
contributors was correctly estimated at two when using MLE (Table 3). We prepared the
reference profile of a suspect who had not contributed any DNA to the mixture (next
to the reference profile of donor 1 which was denoted the victim’s profile). To increase
the complexity of the scenario, this suspect is a third brother of the two brothers present
in the mixture; he shares 23 out of 32 alleles with the true minor contributor.
Mock case 3
The mixture for mock case 3 was prepared by admixing DNA samples taken from
three brothers. Donor 1 contributed 150 pg of DNA, while donors 2 and 3 both
contributed 30 pg of DNA (Table 1). In the consensus profile all alleles of donor 1 were
detected (Table 3), but inference of the major donor’s profile is hardly possible due to
severe allele sharing. For both the 30 pg contributors, more alleles were detected than
for the 30 pg contributors in the other mock cases (Table 3), which is most likely due
to the high number of shared alleles in this three-brother DNA mixture. Actually, only
four alleles of one of the 30 pg donors (donor 3) were missing in the consensus profile)
(Table 3). These were all non-shared heterozygous alleles. As a result of the severe allele
sharing, the number of contributors was underestimated when using the MLE method:
for all profiles an estimate of only two contributors was obtained, which is an expected
result as with MLE PH ratios are not taken into account which is for a three brother
mixture the only indication for the presence of three contributors (Table 3). For mock
case 3 only reference profiles of two of the three brothers were prepared: donor 1 (as
the victim) and donor 2 (as the suspect).
Mock case 4
141
Chapter 6
For mock case 4, we mixed DNA of four unrelated contributors (three males, 1
female).The mixture consisted of one major (300 pg) and three LT (all 30 pg) components
(Table 1). All but one of the peaks that are attributed to the major (300 pg) contributor
had PHs that were above the stochastic threshold of 400 rfu (the exception is a peak of
381 rfu at the FGA locus). The peaks for the minor components all resided below this
threshold. The maximum number of alleles per locus was five for amplifications two and
three, six for amplifications one and four and seven for the consensus profile. Therefore,
using MLE the number of contributors was correctly estimated at four when using the
consensus profile, but underestimated (three contributors) for all four individual profiles.
For profile comparisons in case 4, three reference profiles were prepared, namely that
of the victim (the major donor), the partner of the victim (one of the 30 pg donors)
for which we stated that presence in the evidentiary trace is not crime-related, and the
suspect (another one of the 30 pg donors).
Chapter 6
Chapter 6
Case evaluation and assessment of the evidential
value
For each mock case the ROs started with examining the four amplification plots
and the consensus profile of the evidentiary trace. The ROs (all experienced in LT
profile interpretation) were aware that a stochastic threshold of 400 rfu applies to
the presented NGM profiles. Statistical tools were not employed by the ROs. First, the
ROs were asked to estimate the minimum number of contributors needed to explain
the profile. Therefore, they assessed several aspects, such as the number of alleles per
locus, the PHs, the position of alleles (stutter or not), the amplicon size and the number
of times an allele is detected in the four amplifications. Furthermore, they considered
the profiles in their entirety and examined the coherence between the profiles. The
ROs correctly estimated the minimum number of contributors for all cases (Table 4).
For two cases (cases 1 and 4) some ROs acknowledged that they were possibly giving
a conservative estimation as they commented that there could be an additional low
level contributor.
After profile interpretation, the ROs received the reference DNA profile(s) of the
expected contributor(s), and re-evaluated the evidentiary trace profiles to assess the
genotypes of the various contributors. Next, the ROs compared the reference DNA
profile of the suspect with the interpreted consensus DNA profile of the evidentiary
trace and classified the results for each case, using the four categories of evidential
value (Fig. 1) [1]. We then evaluated the LRs using LRmix on these consensus profiles.
In addition, the LRs for the assembled profiles were evaluated (these profiles contain all
alleles of all contributors without drop-in alleles) as these will represent the LRmix LRs
without the uncertainties residing in LT DNA profiles. While in future the evaluation
of LRs may be the basis for the classification of evidential values, this methodology is
not yet established and the LRs determined in the next section serve an exploratory
purpose.
Mock case 1
In case 1 (mixture of three contributors in a ratio of 150 : 30 : 6 pg, Table 1)
28 alleles of the suspect are observed in the consensus profile which means that
four alleles are missing. Furthermore, one extra allele is included in the consensus
(Table 4). When classifying the evidential value, the ROs used two categories: C (match
without statistical evaluation) and D (cannot be included or excluded) (Table 4). The
two ROs that reported category C considered the alleles of the suspect represented
in the trace profile. The missing alleles in the consensus were regarded as drop-out, an
opinion fed by the observation that all suspect’s alleles were seen in at least one of the
individual amplifications (Table 4). One of these ROs had even considered an RMNE
142
Assessment of complex mock cases
143
Chapter 6
or RMP statistical evaluation at the time of the exercise. However, when this was
attempted later on, it became clear that several assumptions not in agreement with
the DNA profiling results were required for this statistical weighing of the evidence
namely: (1) assume only two donors, of which one is the victim, (2) assume that
nobody but the donors contributed DNA and (3) assume that all alleles of the donors
are represented. Category D, cannot be included or excluded, was reported by six of
the eight ROs. The basis for their decision was the one allele in the consensus profile
(at D2S441, Table 2 and Fig. 2) that could not be explained by the reference profiles
of the victim and suspect, and that this allele was not likely to represent a stutter
drop-in as it was observed in three of the four amplifications. The absence of four
alleles of the suspect in the consensus profile (but all present in one of the individual
amplifications) had less impact on their decisions, as this finding is well expected with
LT samples. Thereby, the crucial locus for this case was D2S441: in the consensus one
allele of the suspect is missing while an allele of the 6 pg donor is included (Table 2
and Fig. 2). If this D2S441 genotype would have been obtained for a high template
sample, all ROs would have excluded the suspect (noting that a close relative of the
suspect could be involved). Now, all ROs recognised the LT effects and most of them
agreed on category D ‘cannot be included or excluded’. The log10 LRs (conditioned
on the prD and regarding either two or three contributors) that are estimated for
the evidentiary trace deviate substantially from the LR estimated for the assembled
profile (Table 4), due to the LT artefacts in the trace sample. Nevertheless, using
the scale suggested by Evett and Weir [29,30] also for the trace sample, very strong
support for Hp over Hd can be obtained (with a low or medium prD). Depending on
the assumption regarding the number of contributors to the sample, Hp is defined as
the victim and the suspect (assuming two donors) or the victim, the suspect and an
unknown individual (assuming three donors) have contributed to the DNA mixture,
while Hd is defined as the victim and an unknown unrelated individual (assuming
two donors) or the victim and two unknown unrelated individuals (assuming three
donors) have contributed to the DNA mixture. When examining the LRs per locus,
D2S441 had the lowest estimated log10 LR, namely below zero for all prDs, as the
genotype can only be explained by having both a drop-out and a drop-in. In contrast,
the log10 LR for locus D2S441 in the assembled profile was well above zero (data
not shown). The LR can vary largely from locus to locus within a profile [31], thus
a detailed examination of each locus is warranted, besides studying the overall LR.
For case 1 the composition of the designed mixture would have allowed a category
B or C classification, as the suspect’s DNA was included in the mixture.The LRs obtained
with LRmix are in agreement with both these categories, as they represent a relatively
strong evidential value. However, the ROs appear to report more conservative as
they opted for category C or D, predominantly because of their uncertainty about
the number of contributors as reflected by the genotyping results at locus D2S441.
Table 4. Results of pro�ile comparisons.
Chapter 6
Chapter 6
144
Mock case 2
Assessment of complex mock cases
145
Chapter 6
In case 2 (mixture of two brothers with a third brother presented as suspect,Table
1), 27 alleles of the suspect are detected in the evidentiary trace consensus profile,
while five of his alleles are missing (two of these alleles were detected once in the
individual amplifications, Table 4). Three alleles in the consensus do not correspond
to the victim or suspect (Table 4). It seems unlikely that these three alleles represent
increased stutters since firstly all are detected in three of the four amplifications and
secondly one of the three extra alleles does not reside at stutter position (Table 5).
Six of the ROs reported category A ‘exclusion’ and two ROs reported category
D ‘cannot be included or excluded’ (Table 4). The ROs that classified category A
regarded the mixture as a two-contributor mixture, allowed for severe allele sharing
and excluded the suspect, because of the five missing and three unexplained alleles
in the consensus profile. The ROs that reported category D based the classification
on the fact that there were unexplained alleles, but also a high resemblance between
the DNA profile of the evidentiary trace and the reference profile of the suspect.
The category D classification of the ROs is supported by the LRs for the evidentiary
trace. These LRs are low and range, conditioned on the prD, from a log10 LR of -3.2
to 2.7 (Table 4). A negative log10 LR means that the contribution of the suspect (as
formulated under Hp, and at a given prD) is not supported; a log10 LR of 2 means
that the probability of observing this evidence is 100 times greater when assuming
the suspect as a contributor than when assuming an unrelated unknown person
as a contributor. The low LRs and the category D classification are in agreement
and imply that the suspect may or may not have contributed to the DNA mixture.
The LR result of 0 for the assembled profile (at prC 0.0) shows that the reference
profile of the suspect does not fit this mixed DNA profile, which is in agreement
with the absence of this brother in the mixture. When uploading the reference
profile of the true contributor to LRmix high log10 LRs were obtained, namely 16.8
for the assembled profile and 12.6 to 8.8 for the trace sample (data not shown),
which shows that LRmix gives much higher log10 LR values when the true brother
is assessed.
Overall, for case 2 most ROs correctly reported exclusion, while the LRs for the
evidentiary trace were inconclusive and represented very weak support for both
the Hp and Hd when conditioned on the prD.This difference may be due to the fact
that the ROs regarded the PHs, acknowledged the possibility of severe allele sharing
between the contributors and considered the profiles in their entirety.These aspects
are not incorporated in the version of LRmix that we have used to compute the LRs.
Table 5. Overview of loci that are complex in the pro�ile comparisons of cases 2 to 4. For case 1 see Table 2. Yellow cells
represent alleles that are not of the victim and match to an allele of the suspect. Blue cells show alleles that cannot be
explained one of the reference pro�iles. Red cells represent alleles of the suspect that are missing in the consensus pro�ile.
Chapter 6
Chapter 6
146
Mock case 3
Assessment of complex mock cases
Mock case 4
In case 4 (mixture of one major and three minor components with one minor
presented as suspect, Table 1), 29 of the alleles of the suspect were detected in the
consensus profile. Three reference profiles were provided (victim, victim’s partner,
suspect). Unknown to the ROs, only 11 of these 29 alleles are specific for the suspect.
Five alleles in the consensus profile could not be explained by the three reference
profiles of which three can neither be increased -1 stutters (see examples in Table 5).
One RO classified the results of the profile comparisons in category C, match without
statistical evaluation, while all the other ROs reported category D, cannot be included
147
Chapter 6
In case 3 (mixture of three brothers; two brothers presented, one as victim and
one as suspect) all the alleles of both the suspect and the victim are detected in
the consensus profile, but four unexplained alleles remain (Table 4). These four alleles
seem true alleles, as they do not reside at -1 repeat stutter position and are observed
in two to four of the individual amplifications (Table 5). Based on the four individual
amplifications and the consensus profile, the ROs estimated a minimum of two
contributors. When the DNA profile of the suspect was compared, all ROs reported
category C, match without statistical evaluation, since all alleles of the suspect were
observed in the consensus profile. The ROs added that an additional (related) third
contributor may be present as this would explain the four remaining alleles. One RO
considered to accompany the match by a statistical evaluation, which at a later time
point appeared not feasible; for RMNE unrealistic assumptions were needed (all alleles
of all donors are represented and there are no drop-outs or drop-ins in the sample),
and a calculation for the minor contributor only was not possible, because the mixture
could not be deconvoluted. When using LRmix, the results of the assembled profile and
trace sample are in the same range when the correct number of three contributors is
regarded for the trace sample (Table 4). When the probability of drop-out is set low,
a high log10 LR is estimated which implies very strong support for Hp (the victim,
the suspect and one unknown individual contributed to the DNA mixture) over Hd
(the victim, and two unknown unrelated individuals contributed to the DNA mixture).
When the prD increases, the log10 LRs go down and more weight is given to the Hd.
When only two contributors are regarded for the trace sample, the support for Hp
decreases, as the estimated log10 LR at a low prD is much reduced.
Case 3 has a degree of complexity that probably will appear in rare cases only.
Nevertheless, the results show the misleading effect that relatedness of donors has,
as the number of contributors is readily underestimated. Actually, both the results of
the ROs and LRmix show lower evidential values when the presence of the third
contributor is not taken into account.
Chapter 6
or excluded. Some ROs that reported category D noticed support for a minimum
of four donors and therefore refrained from conclusions regarding the presence or
absence of the suspect’s DNA in the evidentiary trace (also termed ‘inconclusive’).
Using LRmix for the trace sample, the estimated log10 LR was at most 4.9 (correct
number of four contributors regarded and at a prD of 0.29). This value deviates largely
from the log10 LR of 12.8 estimated for the assembled profile, indicating that much
information is missing for the trace sample (Table 4). Both the categories selected
by the ROs and the estimated LRs are of low evidential value. These results can be
expected with such a complex mixture of four individuals and numerous drop-outs,
even though the suspect was a true contributor to the mixture.
Concluding remarks
Chapter 6
Overall, this descriptive study on the assessment of complex LT DNA mixtures
by both ROs and a probabilistic model, gives the impression that the evaluation of LT
DNA profiles is feasible, as the results obtained in each case are in agreement with
the design of each of the mixtures. We aim for a future in which ROs can interact with
statistical models in order to evaluate evidence and back up their expert assessments.
Exploratory methods such as LRmix will assist this kind of interaction best. We aimed
to provide insight into the assessment of cases involving complex DNA mixtures as
undertaken at the NFI, as this may be informative to ROs dealing with complex LT
DNA profiles. Moreover, we feel that it is crucial to not only invest in methods that
sensitize the analysis of LT DNA samples, but also support the implementation of LR
approaches in casework and put effort in developing guidelines to report the cases in
an unbiased manner.
Acknowledgements
We thank the reporting officers that assessed the four mock cases. We also
thank the four brothers who provided samples and both Ankie van Gorp and Rolla
Voorhamme for critically reading the manuscript and many helpful discussions. This
study was supported by a grant from the Netherlands Genomics Initiative/Netherlands
Organization for Scientific Research (NWO) within the framework of the Forensic
Genomics Consortium Netherlands.
148
Supplementary material
Assessment of complex mock cases
Supplementary Table 1. LRmix results for the consensus pro�ile of the trace sample at
various prDs as summarised in Table 4.
References
1.
2.
3.
4.
6.
7.
8.
9.
10.
149
Chapter 6
5.
A.J. Meulenbroek, T. Sijen, C.C.G. Benschop and A.D. Kloosterman, A practical model to explain
results of comparative DNA testing in court. Forensic Sci. Int. Gene. Suppl. 3 (2011) e325-e326.
P. Gill, J. Whitaker, C. Flaxman, N. Brown and J. Buckleton, An investigation of the rigor of
interpretation rules for STRs derived from less than 100 pg of DNA. Forensic Sci. Int. 112 (2000)
17-40.
J.P. Whitaker, E.A. Cotton and P. Gill, A comparison of the characteristics of profiles produced with
the AMPFlSTR SGM Plus multiplex system for both standard and low copy number (LCN) STR
DNA analysis. Forensic Sci. Int. 123 (2001) 215-223.
M.L. Gaines, P.W. Wojtkiewicz, J.A. Valentine and C.L. Brown, Reduced volume PCR amplification
reactions using the AmpFlSTR Profiler Plus kit. J. Forensic Sci. 47 (2002) 1224-1237.
A.D. Kloosterman and P. Kersbergen, Efficacy and limits of genotyping low copy number (LCN)
DNA samples by multiplex PCR of STR loci. J. Soc. Biol. 197 (2003) 351-359.
L. Forster, J. Thomson and S. Kutranov, Direct comparison of post-28-cycle PCR purification and
modified capillary electrophoresis methods with the 34-cycle “low copy number” (LCN) method
for analysis of trace forensic DNA samples. Forensic Sci. Int. genet. 2 (2008) 318-328.
A.A. Westen, J.H. Nagel, C.C.G. Benschop, N.E.C. Weiler, B.J. de Jong and T. Sijen, Higher capillary
electrophoresis injection settings as an efficient approach to increase the sensitivity of STR typing.
J. Forensic Sci. 54 (2009) 591-598.
N.E.C. Weiler, A.S. Matai and T. Sijen, Extended PCR conditions to reduce drop-out frequencies in
low template STR typing including unequal mixtures. Forensic Sci. Int. genet. 6 (2012) 102-107.
A. Matai, J. Harteveld and T. Sijen, The Next Generation Multiplex (NGMTM Kit) in a Forensic
Setting. Forensic News (2010) July.
V.C. Tucker, A.J. Hopwood, C.J. Sprecher, R.S. McLaren, D.R. Rabbach, M.G. Ensenberger, J.M.
Thompson and D.R. Storts, Developmental validation of the PowerPlex((R)) ESI 16 and
PowerPlex((R)) ESI 17 Systems: STR multiplexes for the new European standard. Forensic Sci. Int.
Chapter 6
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
Chapter 6
22.
23.
24.
25.
26.
27.
28.
29.
150
genet. 5 (2011) 436-448.
TV.C. Tucker, A.J. Hopwood, C.J. Sprecher, R.S. McLaren, D.R. Rabbach, M.G. Ensenberger, J.M.
Thompson and D.R. Storts, Developmental validation of the PowerPlex((R)) ESX 16 and
PowerPlex((R)) ESX 17 Systems. Forensic Sci. Int. genet. 6 (2012) 124-131.
R.A. van Oorschot, K.N. Ballantyne and R.J. Mitchell, Forensic trace DNA: a review. Investig Genet
1 (2010) 14.
T. Caragine, R. Mikulasovich, J. Tamariz, E. Bajda, J. Sebestyen, H. Baum and M. Prinz, Validation of
testing and interpretation protocols for low template DNA samples using AmpFlSTR Identifiler.
Croat. Med. J. 50 (2009) 250-267.
S. Cowen, P. Debenham, A. Dixon, S. Kutranov, J. Thomson and K. Way, An investigation of the
robustness of the consensus method of interpreting low-template DNA profiles. Forensic Sci. Int.
genet. 5 (2011) 400-406.
C.C.G. Benschop, C.P. van der Beek, H.C. Meiland, A.G.M. van Gorp, A.A. Westen and T. Sijen, Low
template STR typing: Effect of replicate number and consensus method on genotyping reliability
and DNA database search results. Forensic Sci. Int. genet. 5 (2011) 316-328.
C.C.G. Benschop, H. Haned and T. Sijen, Consensus and pool-profiles to assist in the analysis and
interpretation of complex low template DNA mixtures. Int. J. Legal Med. doi:10.1007/s00414-0110647-5.
T.M. Clayton, J.P. Whitaker, R. Sparkes and P. Gill, Analysis and interpretation of mixed forensic stains
using DNA STR profiling. Forensic Sci. Int. 91 (1998) 55-70.
P. Gill, R. Sparkes, R. Pinchin, T. Clayton, J. Whitaker, J. Buckleton, Interpreting simple STR mixtures
using allele peak areas. Forensic Sci. Int. 91 (1998) 41-53.
P. Gill, C.H. Brenner, J.S. Buckleton, A. Carracedo, M. Krawczak, W.R. Mayr, N. Morling, M. Prinz, P.M.
Schneider and B.S. Weir, DNA commission of the International Society of Forensic Genetics:
Recommendations on the interpretation of mixtures. Forensic Sci. Int. 160 (2006) 90-101.
D.E. Krane, S. Ford, J.R. Gilder, K. Inman, A. Jamieson, R. Koppl, I.L. Kornfield, D.M. Risinger, N. Rudin,
M.S. Taylor and W.C. Thompson, Sequential unmasking: a means of minimizing observer effects in
forensic DNA interpretation. J. Forensic Sci. 53 (2008) 1006-1007.
B. Budowle, M.C. Bottrell, S.G. Bunch, R. Fram, D. Harrison, S. Meagher, C.T. Oien, P.E. Peterson,
D.P. Seiger, M.B. Smith, M.A. Smrz, G.L. Soltis and R.B. Stacey, A perspective on errors, bias, and
interpretation in the forensic sciences and direction for continuing advancement. J. Forensic Sci. 54
(2009) 798-809.
R. Cook, I.W. Evett, G. Jackson, P.J. Jones and J.A. Lambert, A model for case assessment and
interpretation. Sci. Justice 38 (1998) 151-156.
I.W. Evett, P.D. Gill, G. Jackson, J. Whitaker and C. Champod, Interpreting small quantities of DNA:
the hierarchy of propositions and the use of Bayesian networks. J. Forensic Sci. 47 (2002) 520-530.
H. Haned and P. Gill, Analysis of complex DNA mixtures using the Forensim package. Forensic Sci.
Int. Gene. Suppl. (2011) e79-e80.
P. Gill, A. Kirkham and J. Curran, LoComatioN: a software tool for the analysis of low copy number
DNA profiles. Forensic Sci. Int. 166 (2007) 128-138.
H. Haned, L. Pene, J.R. Lobry, A.B. Dufour and D. Pontier, Estimating the number of contributors to
forensic DNA mixtures: does maximum likelihood perform better than maximum allele count? J.
Forensic Sci. 56 (2011) 23-28.
D.R. Paoletti, T.E. Doom, C.M. Krane, M.L. Raymer and D.E. Krane, Empirical analysis of the STR
profiles resulting from conceptual mixtures. J. Forensic Sci. 50 (2005) 1361-1366.
www.cstl.nist.gov/biotech/strbase/training.htm (2011)
I.W. Evett and B.S. Weir, Interpreting DNA evidence: Statistical Genetics for Forensic Scientists,
Assessment of complex mock cases
30.
31.
Sunderland, MA: Sinauer Associates (1998).
J.M. Butler, Forensic DNA typing, Burlington, MA: Elsevier Academic Press (2005).
D.J. Balding and J. Buckleton, Interpreting low template DNA profiles. Forensic Sci. Int. genet. 4
(2009) 1-10.
Chapter 6
151