Content uploaded by Mannis van Oven
Author content
All content in this area was uploaded by Mannis van Oven
Content may be subject to copyright.
ORIGINAL ARTICLE
Haplogrouping mitochondrial DNA sequences in Legal
Medicine/Forensic Genetics
Hans-Jürgen Bandelt &Mannis van Oven &Antonio Salas
Received: 24 March 2012 /Accepted: 6 August 2012 /Published online: 1 September 2012
#Springer-Verlag 2012
Abstract Haplogrouping refers to the classification of
(partial) mitochondrial DNA (mtDNA) sequences into hap-
logroups using the current knowledge of the worldwide
mtDNA phylogeny. Haplogroup assignment of mtDNA
control-region sequences assists in the focused comparison
with closely related complete mtDNA sequences and thus
serves two main goals in forensic genetics: first is the a
posteriori quality analysis of sequencing results and second
is the prediction of relevant coding-region sites for confir-
mation or further refinement of haplogroup status. The latter
may be important in forensic casework where discrimination
power needs to be as high as possible. However, most
articles published in forensic genetics perform haplogroup-
ing only in a rudimentary or incorrect way. The present
study features PhyloTree as the key tool for assigning
control-region sequences to haplogroups and elaborates on
additional Web-based searches for finding near-matches
with complete mtDNA genomes in the databases. In con-
trast, none of the automated haplogrouping tools available
can yet compete with manual haplogrouping using Phylo-
Tree plus additional Web-based searches, especially when
confronted with artificial recombinants still present in fo-
rensic mtDNA datasets. We review and classify the various
attempts at haplogrouping by using a multiplex approach or
relying on automated haplogrouping. Furthermore, we re-
examine a few articles in forensic journals providing
mtDNA population data where appropriate haplogrouping
following PhyloTree immediately highlights several kinds
of sequence errors.
Keywords Mitochondrial DNA .Forensics .mtDNA
database .Haplogroup .PhyloTree .Multiplex genotyping .
mtDNAmanager .MitoTool .HaploGrep .Sequence error .
Quality control
Introduction
Mitochondrial DNA (mtDNA) analysis has become a rou-
tine target in forensic casework where standard nuclear
markers cannot be applied. The EMPOP project (http://
www.empop.org) [1] provides the leading mtDNA database
of control-region sequences for comparison with partial
mtDNA sequences from forensic traces. Release 6 now
offers 10,841 entire control-region sequences plus some
additional partial sequences. Comparison of a trace se-
quence with the sequences from that database can provide
rough estimates of match probabilities in certain human
population groups.
From a given entire control-region sequence or its hyper-
variable segments I (HVS-I) and/or II (HVS-II) and/or III
(HVS-III), one can often determine the approximate location
of the mtDNA lineage in the current tree estimate of the
global mtDNA phylogeny. This process of allocating partial
(or entire) mtDNA genomes to haplogroups is called
Electronic supplementary material The online version of this article
(doi:10.1007/s00414-012-0762-y) contains supplementary material,
which is available to authorized users.
H.-J. Bandelt (*)
Department of Mathematics, University of Hamburg,
20146 Hamburg, Germany
e-mail: bandelt@yahoo.com
M. van Oven
Department of Forensic Molecular Biology, Erasmus MC,
University Medical Center Rotterdam,
3000 CA Rotterdam, The Netherlands
A. Salas
Unidade de Xenética, Instituto de Ciencias Forenses, Facultade de
Medicina, and Departamento de Anatomía Patolóxica e Ciencias
Forenses, Facultade de Medicina, Universidade de Santiago de
Compostela,
15782 Galicia, Spain
Int J Legal Med (2012) 126:901–916
DOI 10.1007/s00414-012-0762-y
haplogrouping. The term refers to the fact that the basal part
of the known mtDNA phylogeny is encoded as a nested
system of haplogroups (alias monophyletic clades). A readily
accessible global mtDNA tree based on complete mtDNA
genomes is provided by the PhyloTree project (http://
www.phylotree.org)[2], which was initiated in August 2008
and is updated at regular intervals. In fortunate cases, very few
mutational variants alone can unambiguously determine a
specific clade of the mtDNA phylogeny; e.g., the combined
presence of variants 131C and 183G pinpoint a novel subha-
plogroup of B2 (to be named B2j in Build 14) according to [3].
Despite the ease with which mutations and motifs can be
searched in PhyloTree, the use of this Web information is
apparently not yet routinely used bymost forensic laboratories.
In the worst case, no haplogrouping at all is performed.
More often, some sort of haplogrouping is aimed at by using
fixed partial motifs of mutational variants relative to the
revised Cambridge Reference Sequence (rCRS; [4]) that
need to be matched entirely and therefore do not allow for
occasional back mutations. In other cases, obsolete hap-
logroup schemes are still in use, which would no longer be
supported by up-to-date knowledge of the mtDNA phylog-
eny. This is also the case for the on-going anthropological
Genographic project, which employed a coarse classifica-
tion tree, which was outdated even at the time (2005) when
the project was launched [5]; see Supplementary File 1.
Haplogrouping is useful because it can help to detect
sequencing and documentation errors. For instance, optimal
haplogrouping of a given mtDNA sequence allows to parti-
tion the nucleotide variants into shared and private ones and
could therefore indicate the oversight of variants or the
presence of phantom mutations and artificial recombination
[6–9]. When executing a database search, an “innocent”
documentation error that, e.g., transforms a transition into
a transversion could convert a very common haplotype into
a very rare one. Simple errors could also lead to mismatches
when comparing two profiles coming from a forensic stain
and a suspect. On the other hand, the potential haplogroup
status of a control-region sequence suggests which addition-
al coding-region sites could be screened in order to increase
the discrimination power of the mtDNA test [9].
In the present study, we outline strategies for performing
haplogroup allocation of control-region sequences via per-
tinent Google (as performed in [10]) and PhyloTree
searches. We then go through the mtDNA population data
published in forensic journals (Legal Medicine,Journal of
Forensic and Legal Medicine,Journal of Forensic Sciences,
Forensic Science International: Genetics, and International
Journal of Legal Medicine) within the period 2007–2011.
We show that attempts at haplogrouping, either manual or
automated, may fall short of the goal for several reasons.
The shortcomings of the various attempts at haplogrouping
carried out in those articles are highlighted in brief; see
Supplementary File 1. We also demonstrate by way of a
case selected from the field of medical genetics that employ-
ing HVS-I sequences plus a major portion of the coding-
region sequence will normally predict a fine-grained hap-
logroup status. Similarly, a tailored multiplex SNP analysis
of coding-region sites could in principle improve haplo-
grouping inferred from HVS-I&II variation—provided that
the choice of SNPs has been made prudently, that is, by
selecting sites as conservative as possible for elucidating
pivotal parts of the mtDNA phylogeny that are poorly
determined by control-region variants.
Methods
Reference mtDNA tree
The reference tree for the mtDNA phylogeny employed in
our study for testing automated classification tools is
mtDNA tree Build 13 of 28 December 2011 (http://
www.phylotree.org/builds/mtDNA_tree_Build_13.zip).
This basal classification tree of haplogroup motifs is now
based on 10,627 complete genome sequences and thus
appears to be quite robust and detailed. Each haplogroup
is characterized by a set of nucleotide variants that are
shared by all its members due to common ancestry—unless
some sporadic back mutation has partially erased this sig-
nature. These shared variants are commonly referred to as
the “sequence motif”or “mutation(al) motif”. If one wishes
to know the motif (expressed as differences from the rCRS)
characterizing a particular haplogroup, one has to record the
mutations appearing along the branches of the tree all the
way from the rCRS to the node of the targeted haplogroup.
Supplementary File 2provides the control-region motifs for
all haplogroups included in PhyloTree Build 13.
Mutation weighting
The sequence motif of some mtDNA profile could be par-
tially incomplete (i.e., lacking variants that would be
expected from the phylogeny) due to oversight or natural
back mutation. Knowledge of positional mutation rates is
useful to evaluate how likely a back mutation would occur
at a given position. Therefore, in order to explore how
frequently a site is hit by a particular mutation, we refer to
the analysis carried out by Soares et al. [11] (based on 2,196
selected complete mtDNA sequences). In what follows we
will briefly refer to the Soares et al. score as the number of
occurrences (back or forth) inferred from the test dataset of
[11]. The top ten coding-region mutations in [11] are the
transitions at sites 709, 11914, 5460, 15924, 1719, 13708,
13105, 5147, 1598, and 8251, with scores ranging from 59
down to 22.
902 Int J Legal Med (2012) 126:901–916
This may also be compared to the number of occurrences
in Build 13 of PhyloTree, where the top ten coding-region
mutations are at sites 709, 13708, 11914, 5460, 15924,
1719, 5147, 1598, 10398, and 13105, with scores from 36
to 15, quite well agreeing with the previous list (site 8251 is
number 12 here with score 13). The mutations recorded
along this tree are shared mutations on basal parts of the
estimated mtDNA phylogeny, except for a few singular
lineages with their full spectrum of mutations. Mutations
beyond the current haplogroup level (“private mutations”)
are thus not shown. Therefore, the mutations in PhyloTree
constitute only a minority of all mutations that could be
retrieved from an exhaustive global tree of all published
mtDNA sequences that seem to be of reasonable quality.
Mutation weighting is even more important for the
control-region which includes the most extreme hotspots.
The top nine control-region mutations (excluding the hot-
spot transition at 16519 and some length polymorphisms) in
[11] are the transitions at sites 152, 16311, 146, 195, 16189,
16129, 16093, 16362, and 150 with scores ranging from 157
down to 63. In PhyloTree, the top nine are nearly the same:
152, 16311, 195, 146, 16189, 16129, 16362, 150, and
16172 with scores from 139 to 40.
To illustrate how the use of those scores can enhance
haplogrouping, consider the control-region profile 16221T
16291T 263G 309.1C 315.1C (sample BIO-03 Pa from the
Basque province Biskaia [12]). Except for hotspot insertions
at site 309, this profile matches the control-region variation
of two complete mtDNA sequences (GenBank accession
numbers GQ888728 and HQ675034), which are allocated
to haplogroup HV4a1a. On the other hand, the closely
related control-region profile 16291T 263G 309.1C
315.1C with an extra change 311T is found within hap-
logroup H2a2b (GenBank accession number AY339426).
Nonetheless, the classification of the former profile as be-
longing to haplogroup HV4a1a is on safe grounds: transi-
tions at 16291 occur very often (26 times in PhyloTree and
34 times recorded in [11]), whereas the 16221T variant
seems to be confined to haplogroup H4a (with one back
mutation recorded in PhyloTree).
Haplogrouping
As a first attempt towards a formal procedure, one might
regard haplogrouping of a single sequence as the minimum
(weighted) length extension of the mtDNA tree by adding
the new sequence to the reference tree—or much better—to
the total tree of complete mtDNA sequences underlying the
reference tree. In other words, one would in principle seek
for a location in the tree where the new sequence branches
off so that the weighted sum of private mutations gets
minimized. This process just constitutes the familiar sequen-
tial step in the Wagner tree procedure erst proposed in [13].
The obvious drawback is that the new sequence would
typically get misplaced due to homoplasy (long-branch at-
traction) in the rather rare case that it belongs to an entirely
new major haplogroup that is not yet represented in the
reference tree.
The control-region is the mtDNA segment that is most
commonly analyzed in forensic and population genetic stud-
ies. Haplogrouping control-region segments is often—but
not always—straightforward using PhyloTree, where the
optimization criterion can be fulfilled almost by visual in-
spection, at least guided by some experience. In any case,
the strategy would be to screen PhyloTree for the mutational
variants characterizing the sequence of interest and using the
tools available in the preferred Web browser. The search
strategies proposed here extend those previously used in
[14,15] for finding articles in which certain mtDNA muta-
tions were mentioned; see Table 1.
To exemplify these strategies, consider the control-region
profiles PK-187, PK-189, PK-249 of [16], all sharing the
rare variant 16179del, which were allocated by the authors
to haplogroup M30 but not classified further. In order to
explore the control-region variation within this haplogroup,
one could enter the Website http://www.ianlogan.co.uk/
sequences_by_group/m30_genbank_sequences.htm and im-
mediately find a complete mtDNA sequence (GenBank
accession number AY922256) within haplogroup M30d1
listed with this particular deletion. This justifies to classify
those three samples as M30d1 haplotypes. Alternatively,
although more laboriously, one could scan the M node of
the mtDNA tree provided by a Russian company (http://
mtdna.gentis.ru/tx/hg/M/tx_M_304991.htm#) to find first
the M4″67 node (“Taxon M_326510”) in which the M30
node (“Taxon M30_326525”) is nested and then explore the
entire M30 subtree branch by branch until one eventually
arrives at the desired sequence bearing 16179del (“Taxon
M30d1_81654”).
In a similar way, one could fine-classify the control-region
profile PK-044 from [16] allocated to haplogroup W3. This
sample shows four variants 16209C, 16255A, 119C, and
185A not covered by the motif of haplogroup W3. Ian
Logan’s Website for haplogroup W lists two complete
mtDNA sequences (GenBank accession numbers GU147938
and HM214761) allocated to haplogroup W3a which share
the first three of those seemingly private mutations. The same
pair of sequences is then eventually retrieved on the Gentis
Webs ite und er “Taxon N2W3a_322142”. For further instan-
ces of fine-classification beyond PhyloTree, see “Results”and
Supplementary File 1.
Multiplex genotyping design
Since control-region variation alone cannot settle the desired
haplogroup assignment in many cases, the most natural
Int J Legal Med (2012) 126:901–916 903
approach is to gain more information from the coding-
region, for instance, either by directly sequencing several
fragments [17] or by SNP screening via multiplex analysis
[18–22]; see Table 1in [23] for a list of articles on SNP
multiplex analyses. The design of a multiplex SNP array
would depend on whether it is supposed to (1) be used on a
(sub-)continental scale without prior sequence information
(e.g., [18–21]) or (2) be restricted to a specific haplogroup
(e.g., [24,25]) or (3) complement control-region informa-
tion (e.g., [19,26]). Further, the purpose of a multiplex
analysis matters: either haplogroup assignment is aimed at
(as in [18]) or discrimination power between samples in
forensic casework is to be increased (1) without prior
knowledge of haplogroup status (2) or with haplogroup
allocation presumed, or (3) with control-region information
given [27]. For the purpose of discriminating otherwise
identical mtDNA control-region sequences one would pref-
erably choose highly mutable sites, which however are
typically ignored or considerably downweighted for hap-
logroup assignment.
There is no realistic solution for a worldwide universal
SNP design, even when only discrimination power is targeted
[23], given the limitations of casework samples with respect to
availability of mtDNA for such additional analyses. Neither
would any worldwide SNP array be of much use even for
coarse haplogrouping since there are simply too many basal
branches in the Eurasian mtDNA phylogeny (especially with-
in haplogroup M). For the purpose of just identifying African,
West Eurasian and Native American matrilineal ancestry,
three multiplexes of small or medium sizes could suffice
[21]. In contrast, South and Southeast Asian or Oceanian
ancestry is difficult to determine through multiplex analysis
since the knowledge of complete mitochondrial variation is
yet quite limited; see, e.g., [28]. In this situation, one would
best proceed with a large number of parallel multiplex analy-
ses, in order to allocate profiles to main branches of the
phylogenetic tree, and some subsequent nested multiplexes
for refined classification. A multiplex design involving large
numbers of SNPs [29] could then assist in general haplogroup
assignment but never to the level provided by complete ge-
nome sequencing, given that a pre-designed multiplex array
can only target known variation (inferred from the known part
of the phylogeny and population data).
Then the ideal properties of any multiplex design for
haplogrouping are the following: (1) it should either have
a sub-continental focus or subclassify only a single major
haplogroup, (2) it should shield against mtDNA outliers
from other parts of the mtDNA phylogeny, (3) it should
use phylogenetically meaningful default categories (such as
L3*, M*, N*, R*, etc.), (4) it should avoid including sites of
very high mutation rates (if possible). Two further points are
that (5) the decision tree for haplogroup assignment should
constitute a truncation of the entire mtDNA tree from Phy-
loTree relative to the tested variants of the entire mtDNA
molecule, and that (6) the control-region information should
be integrated in order to arrive at a haplogroup allocation at
some finer level whenever (partial) control-region sequenc-
ing has been carried out.
As to points (1), (2), and (3), every sub-continent-specific
multiplex that is designed for allocating mtDNA samples to
the according haplogroups should provide information on the
important multi-furcation pivots of the mtDNA phylogeny.
Therefore, at least the following sites should be included:
1018 (recognizing haplogroup L3), 10400 (haplogroup M),
either 9540 or 10873 (haplogroup N), and 12705 (haplogroup
R). All of these mutations appear only once in PhyloTree
Build 13 and have Soares et al. [11]scoresofatmost2,so
that recurrent mutations at these sites are not likely to blur a
multiplex analysis. In particular, a reliable recognition of
Native American mtDNA haplogroups in multi-ethnic popu-
lations cannot make do with just five mtDNA coding-region
sites [30].
Table 1 Quick Web-based searches for mtDNA haplogroup affiliation
Goal Website Action / Query
Determining haplogroup http://www.phylotree.org/tree/
main.htm
Download the entire tree and search
site numbers (e.g., “73”)
Exploring the subhaplogroup levels http://mtdna.gentis.ru/ Travel through the tree from its root to the targeted
haplogroup (according to PhyloTree) and then scan
the nested array of subhaplogroups (“child taxa”)
Finding closely related lineages
in GenBank
http://www.ianlogan.co.uk/sequences_
by_group/haplogroup_select.htm
Click haplogroup and search private mutations (e.g., A73G)
Finding perfect matches in fragments http://www.google.com/ Enter truncated profile (e.g., “‘73G 263G 315.1C 489C 573.1C
573.2C’” or “‘A73G A263G 315.1C T489C 573.1C 573.2C’”)
Finding related partial profiles http://www.google.com/ Enter partial profile (e.g., “A73G G94A PhyloTree”or “A73G
G94A mtDNA”or “73G 94A mtDNA”)
Finding specific variant http://www.google.com/ Search variant (e.g. “A73G PhyloTree”or “A73G mtDNA”or
“'73A >G mtDNA”or “73G mtDNA”)
904 Int J Legal Med (2012) 126:901–916
As to point (4), one cannot always avoid fast sites, such
as position 3010, which is for instance the only site distin-
guishing haplogroup H1 from H, or worse, position 709
(relevant for the distinction of D4g1c from D4g1). Even
the top-most variable nucleotide substitution at position
16519, which is ignored in PhyloTree, can be of use for
particular haplogroups. Enigmatically, this site is very stable
within haplogroup F, where 16519T (with the rCRS nucle-
otide) is virtually fixed in subhaplogroups F3 and F4,
whereas 16519C is characteristic for subhaplogroups F1
and F2, so that it could make sense to employ this in a small
multiplex design [31].
As to point (5), any “new strategy”for the discrimination
of mtDNA haplogroups should really use the most recent
update of PhyloTree and preferably employ additional in-
formation. A scheme solely derived from a truncated and
distorted version of an mtDNA tree that has become obso-
lete (see, e.g., [32]) is of no help for the practicing forensic
geneticist.
Results
Quality control enhanced by haplogrouping
Oversights or misdocumentation, phantom mutations, and
artificial recombination are the main kinds of errors in
mtDNA datasets; for a fine-classification of errors, see
[33]. When (nearly) the entire mtDNA genome of a sample
is sequenced, then at least with West Eurasian mtDNA
samples haplogrouping is quite pedestrian, so that potential
oversights could be seen immediately. With short stretches
of mtDNA sequences such as HVS-I&II it may be difficult
to distinguish oversights from natural back mutations since
many of the variable sites are mutational hot spots. None-
theless, if several expected mutations were not recorded,
then one can predict incomplete documentation of nucleo-
tide variants. This is the case with the two haplogroup U2e1
sequences (VP64 and VP65) from [34]: the former lacks
four expected variants (16051G, 16362C, 73G, and 340T)
and the latter three (16051, 16362C, 217C), which seems to
reflect a reading problem at the beginning and end of the
two sequenced parts rather than the effect of natural back
mutations.
Phantom mutations are still an issue in forensic popula-
tion studies despite the fact that they should be practically
absent if standard sequencing protocols were followed. Pri-
vate or seemingly private mutational variants of a given
mtDNA profile are those variants which are not matched
by the associated haplogroup motif. One could then search
for additional matches in mtDNA profiles belonging to the
targeted haplogroup. Otherwise, rare mutations indicate po-
tential phantom mutations when they are repeatedly
recorded as private variants on the background of different
haplogroups within a population sample set [6]. Notorious
control-region phantom mutations (artificially generated un-
der suboptimal sequencing conditions) have been investi-
gatedin[7]. In particular, extension of the C stretch
preceding site 310 typically led to seeming mutations further
down in the 3′direction at sites 317, 320, 330, 343, 345, etc.
(see Table 5 of [7]). For instance, among 15,195 sequences
recorded in EMPOP Release 6 that cover these sites, there is
not a single instance of 320T. In [11] this variant received
two hits, both coming from the study [35] that had a few
problems with phantom mutations [7]. In the GenBank
entries of complete mtDNA sequences, one additionally
finds this mutation in EF184595 (from the problematic
dataset [36]; see [37]) and in FJ383190 [38] as well as in
HM238198.
Now, in the dataset provided by [34] we are seeing five
samples (VP24, VP45, VP57, VP79, and VP90) of hap-
logroup status R0, J1c2, T1a, HV0, and U8b, respectively,
for which 320T was recorded. Moreover, in two cases, this
is followed by another unknown transversion variant, 324G
or 328C, respectively, for which the scores in EMPOP and
[11] are all zero. In a study of Japanese mtDNA variation
[39], 320T was recorded seven times on different hap-
logroup backgrounds (B4, B4a, B4b1b, B4b1b′c, B4f,
D5a2a1, and N9a2a). A common cause for phantom muta-
tions in the HVS-II segment is the time-saving approach of
reading only the light strand, which however is not recom-
mended if one wishes to achieve optimum sequence quality
in forensic DNA analysis [40].
Artificial recombination between HVS-I and HVS-II has
haunted large databases as well as single studies since many
years [41–43]. In the Japanese study [39], the classification
of HVS-I&II sequences incorporated the knowledge of the
mtDNA phylogeny at the time. Now, with a deeper knowl-
edge of the mtDNA phylogeny, one can be more specific in
view of Build 13 from PhyloTree. For instance, the HVS-
I&II profile no. 94 (16136C 16183C 16189C 16217C
16284G 73G 146C 199C 202G 207A 263G 315.1C), clas-
sified as B4b1 by the authors of [39], would now best be
traced in PhyloTree. The rare variant 202G (occurring as a
unique event in both PhyloTree and [11]) immediately
points to haplogroup B4b1a1, and in fact, all other variants
are captured by this classification as well, with no variant
from the profile left unmatched. So, in this case, the profile
exactly constitutes that particular haplogroup motif. How-
ever, confronted with the unclassified sample no. 739
(16189C 16190T 16193.1C 16193.2C 16261T 16362C
73G 146C 199C 202G 207A 263G 309.1C 315.1C), we
would be led to exactly the same haplogroup slot with HVS-
II alone, but the HVS-I part is at odds with this classifica-
tion, as it would need to postulate three back mutations plus
additional mutations. The HVS-I part alone would rather
Int J Legal Med (2012) 126:901–916 905
point to haplogroup B4a (http://www.ianlogan.co.uk/
sequences_by_group/b4a_genbank_sequences.htm). There
is a second case of artificial recombination in [39]. The HVS-
I&II profile no. 92 (16136C 16183C 16189C 16217C 16284G
73G 103A 263G 315.1C) gets allocated to B4b1a1 according
to its HVS-I part, but its HVS-II part is incompatible with this
clear-cut haplogroup allocation and suggests B5b status instead
(e.g., either B5b1 or B5b2c; see http://www.ianlogan.co.uk/
sequences_by_group/b5_genbank_sequences.htm).
As for another instance of artificial recombination from
[34], the recorded profile 16126C 16270T 16294T 16304C
73G 242T 263G 295T 309.1C 315.1C (sample VP61) testifies
to J1b1a status when just HVS-II is consulted, but the HVS-I
part is definitely incompatible with this assignment and sug-
gests haplogroup T2b status instead. Further subhaplogroup-
ing is not possible as 16270T is not found among 87 complete
T2b mtDNA sequences drawn from GenBank as listed on Ian
Logan’s Website for haplogroup T2b. Finally, a classic case of
sample crossover has occurred in [44]betweentwosamples
from haplogroup B2 (similar to haplotype LPAZ008) and
haplogroup C1 (similar to haplotype LPAZ080), which has
given rise to the artificial hybrid mtDNA sequences LPAZ092
and LPAZ094; see Table 2.
In conclusion, haplogrouping is an invaluable tool for
narrowing the focus on potential sequencing or documenta-
tion errors. Coarse-grained haplogrouping may be sufficient
for comparison of the mtDNA pools of different geographic
or ethnic populations but not so for forensic databases. For
example, if one wished to investigate the profile 16129A
16180G 16223T 16234T 16288C 16298C 16327T 16518T
16519C 16527T 54A 73G 249del 263G in the sequence
range 16020–300 (drawn from [45]), the obvious assign-
ment to haplogroup C would not be of much help since this
would leave us uninformed about the rare variants 54A and
16518T, which do not occur in the list of Soares et al. [11]
and therefore constitute candidates for scrutinizing. A search
of 16518T in PhyloTree immediately leads to haplogroup
C5c, and Ian Logan’s Website for C5 subsequently provides
us with a near-match (in subhaplogroup C5c1, GenBank
accession FJ951452) of the entire profile with differences
only at positions 16093, 16129, and 16180, all of which are
unsuspicious.
Multiplex haplogrouping
In order to demonstrate the haplogrouping problems via
multiplexing, we reassessed in detail the haplogrouping
performed with the 33 mtSNP typing assay (33-plex)
designed by [46]. The same assay seems to have been used
later in [47] (without reference to [46]). None of the major
West Eurasian superhaplogroups (M, N, and R) were ex-
plicitly targeted by these mtSNPs. As a result, one arrives at
odd default categories, so that a mixed bag of unassigned
lineages in- and outside haplogroup N arises, misleadingly
referred to as “R”. This in part explains suboptimal results in
the case of the Turkish sample since mtDNA lineages of
ultimate African or East Asian ancestry could not be iden-
tified (apart from some Near-Eastern haplogroups which are
rare in Europe). Moreover, a number of minor basal West
Eurasian haplogroups (such as M1, R1, R2, R0a, HV1,
HV2, etc.) were not interrogated. Due to this lack of basal
distinction, some specific lineages outside haplogroup R
would be erroneously allocated to particular R subha-
plogroups because of recurrent mutations occurring at the
selected SNP sites. This is the case with sample mt150*,
where the HVS-I&II variation points to the Southeast Asian
haplogroup E1a1a. The multiplex analysis of [47] can also
yield misleading results when dealing with haplogroup HV.
Thus, their multiplex design does not distinguish between
subhaplogroups E of M and HV of R, but E1a1 has 14766C
as a single characteristic variant within haplogroup E, just as
haplogroup HV does within R0. As for point (4) (see
“Methods”), five of the top ten mutational hotspots in the
coding-region (viz. sites 709, 5460, 1719, 13708, and 8251)
were actually selected in [46,47], among which is the top
hypervariable coding-region site 709. Finally, the single
default category named “R”is too ambiguous because it
embraces most African and Asian mtDNA lineages from
various major haplogroups. Since 709 was chosen as the
single characteristic site for “N2 status”every complete
mtDNA profile with haplogroup defined according to the
decision tree would automatically fall into the category
“N2”whenever it bears the 709 variant and is outside
haplogroups HV, JT, U, and X. For example, most of the
African haplogroup L1b would thus be shunted into “N2”.
As for point (5) (see “Methods”), the decision tree of [46]
and its apparent update in [47](where“2709”should read
“2706”throughout) as a truncation of Build 8 of PhyloTree
may miss minor haplogroups due to mutations that cause a
primer mismatch for a targeted SNP. Consider, for instance,
the haplogroup U5b2a1b sample mt180 (see no. 6 in Table 2),
which was allocated to haplogroup R by the HVS-I&II anal-
ysis [48] and received the status “Not full profile”by the
multiplex analysis [47]. However, the latter analysis shows
that two SNP sites positively supported U and U5 status and
only the information for site 1719 was missing. The latter site
is rather irrelevant for the allocation to haplogroup U5. On the
other hand, the failure of the 1719 reaction (i.e., due to a
primer mismatch) in this case even gives a hint as to which
subhaplogroup of U5 can best be searched for near-matches
with the HVS-I&II profile: site 1721 is characteristic for
haplogroup U5b2. In fact, the optimal near-match is found
within a particular subhaplogroup of U5b2.
Now, with more recent updates of PhyloTree, one could
discern more haplogroups than those that were listed in
Table 1of [47]. For instance, the haplogroup J1c samples
906 Int J Legal Med (2012) 126:901–916
Table 2 Instances of (partial) control-region sequences with haplogroup classification according to original publication, mtDNAmanager, MitoTool, HaploGrep, and manual Web searches
No. Sample ref. Range
a
Haplotype
b
HG status
in article
“Expected”vs “Estimated HG”
by mtDNAmanager [50]
c
MitoTool [52]
d
HaploGrep [53]
e
Likely HG [2] Related sample
from GenBank
1[61] CR 16086 16129 16209 16223 16272 73 152
225 249d 263 315 + C 316 489 523-524d
M5 L3f* vs L3f* (Afr, WeuA, Ad);
M20 vs M20 (EA, Oc)
M20 M20 √M20 HM030505
2[62] CR 16172 16183C 16189 16209 16223 16258T
16311 16362 73 185A 189 195 234 263
309 + C 315 + C 523-524d
N9a5 L3f2 vs L3f2 (Afr, WeuA, Ad,
EA); M10a (Oc)
N10a N10a ~ N10a HM030542
3[62] CR 16069 16172 16223 16278 16291A 16298
16362 73 150 152 199 263 309 +
CC 315 + C
N* L3b1a vs L3b1a (Afr, Ad,
WeuA); J2 vs L3b1a (EA);
N10b vs N10b (Oc)
N10b N10b √N10b HM030500
4[48] HVS-I/II 16114A 16126 16218 16223 16275 16291
16356 16390 16391 73 263 309 + C
315 + C
L3 M3 vs M3 (Afr); E1a1a vs M3
(OC, Ad); M3 vs M3
(WeuA, EA)
H2a2b; H2a5;
H34; H20;
M52 ? M52a EF093557
5[48] HVS-I/II 16223 16291 16362 16390 73 263
309 + C 315 + C
L3 E1a1a vs E1a1a E1a1a E1a1a √E1a1a EF093544
6[48] HVS-I/II 16189 16319 16325 73 150 152 263 315 + C R U5b2a1 vs U5b2a1 (Afr, WeuA);
G1a1 vs U5b2a1 (EA, Oc, Ad)
U5b1a1b U5b2a1b √U5b2a1b GU296545
7[48] HVS-I/II 16051 16162 16213 16266 73 146
263 315 + C
U2 H1a vs H1a H1a3; H1h;
H1n; U2b
H1a3 ~ H1a3 EU979418
8 Family Tree
DNA
CR 16069 16126 16145 16231 16261 73 150
152 195 215 263 295 310 + T 315 + C
319 489 513
n.a. J2a vs J2a J2a1a1 J2a1a1 √J2a1a1 GU903270
9 Family Tree
DNA
CR 16086 16222 16224 16270 16311 16519
73 146 263 315 + C
n.a. K2b1 vs K2b1 K2b1a K2b1a √K2b1a EU770310
10 [63] CR 16086 16239 16311 16320 73 150 263
315 + C
n.a. L3e2 vs ? HV9; H13a2b; HV7;
HV10; HV11
U5b2a1a2 ? U5b2a1a1 GU296544
11 [64] CR 16224 16519 73 152 204 263 315 + C
497 524 + AC
? K1a vs ? H16a; H9; H3g;
K1a3a1b
K1a4c ~ K1a (K1a4a1) EU597496
12 [64] CR 16069 16261 73 185 189 263 295 315 + C
462 489
? J1c7 vs R8a2 J1c7 J1c + 16261 ~ J1c6 AY495209
13 [64] CR 16224 16519 73 152 204 263 272 315 + C
497 524 + AC
? K1a vs ? H16a; H9; H3g;
K1a3a1b
K1a4c ~ K1a (K1a4a1) EU597496
14 [64] CR 16183C 16189 16193 + C 73 262 263 285
309 + CC 315 + C 323 385 523-524d
? U1a1 vs ? H1g; H1q; U1a1 U1a1 √U1a1 AY882396
15 [64] CR 16069 16145 16207 16222 16231 16261
73 150 152 195 215 246 263 295
309 + CC 315 + C 319 489 513
? J2a vs ? J2a1a J2a1a ~ J2a1a FJ348157
16 [64] CR 16183C 16189 16193 + C 73 262 263 285
309 + C 315 + C 323 385 523-524d
? U1a1 vs ? H1g; H1q; U1a1 U1a1 √U1a1 AY882396
17 [64] CR 16153 16298Y 72 73 93 95C 263 309 + C
315 + C
? ? vs ? H2a2; H2a; H2; H;
HV0; V; V7a; R0
V7a ? V7a AF347006
18 [64] CR 16153 16298 72 73 93 95C 263 309 + C
315 + C
? ? vs ? H2a2; H2a; H2; H;
HV0; V; V7a; R0
V7a ~ V7a AF347006
19 [64] CR 16183C 16189 16193 + C 16217 16519
73 263 309 + CCC 315 + C 498d 499
B4b B4b vs B4b B4b; B2 B4b √B4b (B2d) EU095550
20
f
[64] CR D4/G D4/G vs D4/G M74; D4j11 D4e2a √M(M43a) FJ770954
Int J Legal Med (2012) 126:901–916 907
Table 2 (continued)
No. Sample ref. Range
a
Haplotype
b
HG status
in article
“Expected”vs “Estimated HG”
by mtDNAmanager [50]
c
MitoTool [52]
d
HaploGrep [53]
e
Likely HG [2] Related sample
from GenBank
16223 16311 16362 16519 73 263 315 + C
489 573 + CCC
21 [64] CR 16093 16172 16223 16297 16311 16362
16400 16519 73 146 185 263 315 + C 489
D4/G N10a vs D4/G (Afr, WeuA ,
Ad) M7e vs M7e (Oc, EA)
M7e M7e ? M74 HM030520
22 [64] CR 16153 16298 72 93 95C 263 309 + C
315 + C 523-524d
HV0 HV0 vs HV0 H2a2; H2a; H2; H;
HV0; V; V7a; R0
V7a √V7a AF347006
23 [64] CR 16298 16519 263 315 + C HV0* HV0* vs HV0* H2a2; H2a; H2; H;
HV0; V; R0
HV0 ~ HV (V) AY495306
24 [64] CR 16067 16311 152 195 263 315 + C HV1 H11 HV1 H11 vs HV1 H11 H11; H2b HV1b + 152 ~ HV1 (HV1b) HQ165756
25 [64] CR 16129 16185 16223 16224 16260 16298
16519 73 151 152 249d 263 315 + C 489
M37a Z vs Z Z1a Z1a √Z1a1a AY339515
26 [64] CR 16069 16126 16214 16311 16362 73 150 195
235 263 295 309 + C 315 + C 326 489
R0a J A4 vs J2 J2a2a J2a2a ~ J2a2a EF660967
a
Reported sequence range: CR entire control-region; HVS-I/II HVS-I and HVS-II
b
Mutations are recorded with respect to the rCRS [4]. All mutations are transitions unless a letter specifies a transversion (A, C, G, T) or a deletion (d); an insertion is signaled as “+”followed by the
inserted nucleotide(s)
c
The haplogroup status is identical for the five metapopulations considered by mtDNAmanager with the exception of those cases specified in round brackets where population codes are as follows:
Afr African; WeuA West Eurasian, EA East Asian, Oc Oceanian, Ad Admixed
d
The primary classification provided by MitoTool reflects completely matched motifs of specific haplogroups
e
HaploGrep provides three levels of security for the haplogroup classification (designated here by symbols ?, ~, √: haplogroup assignment not reliable probably due to insufficient data (?),
haplogroup assignment critical (~), and high amount of polymorphisms explained by the sample’s haplogroup affiliation (√))
f
If 573 + CCC is changed to 573 + CC in haplotype #20, then mtDNAmanager yields D4/G vs D4/G
908 Int J Legal Med (2012) 126:901–916
mt183 was allocated to J on the basis of the HVS-I&II but
received status “Hg not defined”(because the presence of
the rCRS-variant at site 2706) according to the multiplex
analysis [47]. However, PhyloTree highlights the recurrent
mutation at site 2706 in the definition of haplogroup J1c3c.
Therefore, the SNP analysis is consistent with the HVS-I&II
haplotype and in conjunction now yields this finer
classification.
For the Danish sample, haplogroup assignments based on
HVS-I&II [48] was poorly contrasted with the multiplex
analysis [47]. The use of the same sample numbering/codes
permits the reader to combine the results from the three
studies [46–48]. In some cases, the multiplex analysis of
[47] would support the haplogroup allocation obtained from
PhyloTree. For instance, consider sample mt098 (no. 7 in
Table 2) which bears the motif 73G 16051G 16163G char-
acteristic of haplogroup H1a3: in fact, the multiplex analysis
correctly identifies this sample as belonging to H1, so that
an initial guess of haplogroup status “U2”on the basis of
HVS-I/II can firmly be rejected.
A particularly interesting case in this regard constitutes
the outlier sample mt055* from [48] which received the
status “Hg not defined”in [47]. In fact, the HVS-I&II
profile points to haplogroup M52a (of North Indian and
Nepalese provenance) in view of the presence of transitions
16275G and 16390A (despite the lack of 16327A). The first
hint at this unexpected allocation comes from the rare tran-
sition at site 16275 (with Soares et al., score 1), which
occurs only once in PhyloTree Build 13, viz. as one of the
three control-region mutations defining M52a. A Google
search with “A16275G PhyloTree”provides (via Ian
Logan’s Website) two additional GenBank sequences that
can be allocated to other haplogroups (W1f and G1a1a,
respectively). A further search with “16275G mtDNA”
points to Family Tree DNA (FTDNA) control-region pro-
files from haplogroup T2b and from haplogroup M52b
(matching the complete mtDNA sequence with Genbank
accession number FJ383488). More interestingly, the latter
Google search also directs one to the Release history for
EMPOP 2, where the user gets informed that the sequence
information for the Danish sample has been expanded to the
entire control-region, so that the variants 16519C 489C
573.1C 573.2C 573.3C 573.4C 573.5C 573.6C come on
top of the HVS-I&II profile. This is in line with haplogroup
status M and covers also the insertion of three cytosines at
site 573, which is characteristic of haplogroup M52a. An
EMPOP query for the two variants 16275G and 16390A
together, however, yields yet another hit with a forensic
sample from Thailand, which was previously published by
[49] and has haplogroup R9b1b status. The partial profile
16223T 16275G 16390A 489C 573.1C 573.2C 573.3C is
distinctive enough to guarantee M52a status, although this
entails that the definition of M52a may need some
modification in the future by not necessarily requiring
16327A, for example. If one would yet be in slight doubt
about this haplogroup allocation, one could of course con-
sult the coding-region. The study [47] has actually done
that: one of the two characteristic sites of the larger hap-
logroup M52′58 is 5460, which is exactly the site that gave
rise to the status “Hg not defined”in [47].
Automated haplogrouping
The Web-tool mtDNAmanager [50] is supposed to deliver
for any partial (HVS-I/HVS-II/HVS-III) or entire control-
region sequence two kinds of haplogroup status: “expected”
and “estimated”. Any discrepancy between expected and
estimated status is supposed to be regarded as dubious,
warranting some verification by the user (if possible). The
exact processes by which these inferences are made are just
vaguely indicated in the accompanying article [50]. There-
fore the implemented algorithm can only be treated as a
black box, for which some features might be uncovered
through some series of focused tests. To this end we took
some published entire control-region sequences and modi-
fied some of them by single changes.
To begin with, consider the haplogroup J2a1a1 sequence
no. 15 of Table 2. This haplogroup, as well as its nested
superhaplogroups J2a1a and J2a1, are all unknown to
mtDNAmanager, although each of them has diagnostic
control-region sites and J2a1a1 has essentially been in place
since PhyloTree Build 3 (1 March 2009). Thus, when the
entire control-region is offered as input, then “J2a”is
returned for both expected and estimated haplogroup.
Now, if one removes the variant nucleotide at 152, then
“J2a”versus “J2”will be the answer to the corresponding
query, and if one drops 150 instead, then “J2a”versus “J”
will be returned. Finally, if just 295 is removed, then “J2a”
versus no haplogroup status proposed (to be interpreted as
“?”) will be the output. This shows that the “estimated
haplogroup”status does not tolerate single back mutations
at any of the diagnostic sites listed in the protocol of
mtDNAmanager; note also that in this particular example,
the mentioned transitions are mutational hotspots, so spo-
radic reversions may not be infrequent within these hap-
logroups. If this feature was properly understood by the user
in a laboratory, then the analyst could go back to the electro-
pherograms and check. But if everything appeared to be
correct in any of these hypothetical cases, then the only
reasonable output would be J2a1a1 in view of the wealth
of mutations left that signalize this quite specific haplogroup
status.
The strict requirement handled by mtDNAmanager that
all variants from a perceived haplogroup motif must be
present can be quite misleading. In particular, PhyloTree
itself or the listing of all complete mtDNA sequences from
Int J Legal Med (2012) 126:901–916 909
GenBank (see Ian Logan’s Website) falling into a particular
subhaplogroup may clearly demonstrate the reversal of
some basal nucleotide variant (relative to rCRS) or the
parallel gain of 73G which is apparently given a special
treatment by mtDNAmanager. For example, haplogroup
V7a bears the HVS-I&II motif 16153A 16298C 72C 93G
263G (and 309.1C 315.1C). The sequences nos. 17, 18, and
22 of Table 2are therefore clearly members of this hap-
logroup but are partially left unclassified by mtDNAman-
ager (Table 2). Moreover, the variant 95C also present in
these three sequences is confirmed to occur in some hap-
logroup V7a member (GenBank accession number
AF347006).
The next test case is a haplogroup K2b1a sequence (no. 9
in Table 2). This haplogroup has also been present in Phy-
loTree since Build 3, but mtDNAmanager recognizes only
the next superhaplogroup K2b1. Now, if one queries only
the HVS-I part (or together with the HVS-III part which
does not contain any variant here), then expected and esti-
mated haplogroup differ: “K2b1”versus “U5 K”. Although
the variant 16270T is diagnostic for haplogroup U5, its
parallel presence within haplogroup K2b is well confirmed
(giving rise to K2b1) and the additional mutation (16222T
for K2b1a) renders the assignment K2b1 quite reliable.
It is also instructive to see how mtDNAmanager deals
with obvious instances of artificial recombination. First, the
SWGDAM database [51] curated by the FBI has never been
fully revised and still contains two very clear cases of hybrid
profiles; see entries nos. 1 and 2 of Table 3. The SWGDAM
profiles were added to the reference data on which mtDNA-
manager operates. Table 3also includes a few other instan-
ces we have found in the recent forensic datasets (see above
and Supplementary File 1). In no case mtDNAmanager gave
a warning about the hybrid status of these artificial recombi-
nants, essentially because no precautions were taken to
contrast the haplogroup predictions from the separate seg-
ments HVS-I and HVS-II. And in no case were the two
different haplogroup sources identified. In one case (no. 5)
“expected”and “estimated”haplogroup status even coincid-
ed, so that the user would not be alerted at all. This shows
that mtDNAmanager is unhelpful for discovering clear
instances of artificial recombination.
In summary, (1) mtDNAmanager works with a restricted
and not properly updated list of haplogroup motifs, (2)
variants are not weighted according to their mutability but
instead according to polymorphism counts (in a cryptic
fashion), (3) the distinction between “expected”and “estimat-
ed”haplogroup is not well justified and can rather mislead the
average user, (4) mtDNAmanager does not incorporate the
information that can be drawn from the complete mtDNA
sequences stored in GenBank, (5) it handles ambiguous
“meta-population”categories (e.g., “Admixed groups”includes
Dubaians, Egyptians, “Hispanics”, Northern Africans, and
Table 3 Artificial mtDNA recombinants from forensic databases and their haplogroup classification according to mtDNAmanager, MitoTool, HaploGrep, and manual Web searches
No. Sample ID, source HVS-I
a
HVS-II
a
HG status [2]“Expected”vs “Estimated”
‘HG’” [50]
Best HG according
to MitoTool [52]
Best HG according
to HaploGrep [53]
1 USA.AFR.000942 [51] 16126 16187 16189 16223 16264
16270 16278 16293 16311 16519
73 249d 263 290d 291d
309 + C 315 + C 489
L1b × C1 C1c vs X M31a1 M31a1 M31a1 ?
2 FRA.CAU.000084 [51] 16298 73 185 188 228 263 295
315 + C
HV0× J1c ? vs ? H2a2, H2a, H2, H,
HV0, V, R0
J1c2 ~
3 VP61 [34] 16126 16270 16294 16304 73 242 263 295 309 + C
315 + C
T2b × J1b1a T2b vs T U5 T2b4 T2b + 16296! ~
4 739 [39] 16189 16190 16193 + CC 16261 16362 73 146 199 202 207 263
309 + C 315 + C
B4a × B4b1a1 M7c vs R30a* H1b; H1g; H1q; H1h; H1n;
H7a1; B4a1a; B4a1c4
R31 ?
592[39] 16136 16183C 16189 16217 16284 73 103 263 315 + C B4b1a1 × B5b B4b1 vs B4b1 H1g; H1q; B4; B4b1 B4b1 ~
6 LPAZ092 [44] 16181 16189 16217 73 185 249d 263 290d
291d 309 + C 315 + C
B2 × C1 B4 vs ? H1g; H1q; B4 B4 ?
7 LPAZ094 [44] 16182C 16183C 16189 16223 16298
16325 16327 16344
73 263 309 + CC 315 + C C1 × B2 U5b2a1 vs M8 H1g; H1q; N9b C4c1b ?
a
Mutations are recorded with respect to the rCRS [4]. All mutations are transitions unless a letter specifies a transversion (A, C, G, T) or a deletion (d); an insertion is signaled as “+”followed by the
inserted nucleotide(s). A reversal of a mutation along a branch of the mtDNA tree is highlighted by the suffix “!”(as in PhyloTree)
910 Int J Legal Med (2012) 126:901–916
Venezuelans) which can misdirect otherwise clear-cut hap-
logroup allocation, and (6) does not shield against serious errors
in the employed reference database or input sequences.
The MitoTool software [52] has a wider scope than
mtDNAmanager since it provides several applications of
interest in the biomedical field and it allows dealing with
entire genomes (not merely control-region segments). Here,
we only consider the module that allows for haplogroup
classification. The strategy followed by this web interface
application seems to be based on a score system that takes
into account the number of exact matches between the tested
profile and the diagnostic variants that define the different
haplogroups of the worldwide phylogeny, in a similar way
as mtDNAmanager does, but here too the user is not in-
formed about the exact optimality criterion. The classifica-
tion tool performs substantially better than mtDNAmanager
as can be concluded from the results of Table 2. However,
like mtDNAmanager, the procedure implemented in Mito-
Tool does not allow any detection of signals of artificial
recombinants (Table 3).
Recently, a more sophisticated tool for automated haplo-
grouping has been proposed [53]: HaploGrep allows for the
classification of mtDNA segments of any length into hap-
logroups using the current build of PhyloTree as the refer-
ence classification tree and an algorithm that involves the
weighting of diagnostic sites according to their relative
positional mutation rates. The weighting scheme employs
a simple linear-affine scaling of the absolute number of
occurrences of a particular mutation in the current build of
PhyloTree. For example, given that the maximum score of a
mutation recorded in PhyloTree Build 13 equals 139, all
(and only those) mutations that occur in Build 13 get
weighted by 140 minus their respective numbers of occur-
rences. Thus the top mutation has weight 140−13901,
whereas the next hotspot mutation has weight 140−890
51. A factor of 51 however does not properly reflect the
relative likelihoods of these two hotspot mutations. At the
other extreme, the difference in weights (139 versus 134)
between a mutation recorded only once in Build 13 and a
medium-frequent control-region mutation with PhyloTree
score 6, say, is far too small in order to have discriminative
power.
Table 2includes the results of HaploGrep (using Build 13
of Phylotree) for the same test haplotypes used for mtDNA-
manager and MitoTool. A total of 19 out of the 26 test
samples (73 %) from this table were classified as one could
achieve by using PhyloTree and scanning individual com-
plete mtDNA sequences manually. An interesting feature of
HaploGrep is that it implements a system that alerts the user
about the performance of the haplogroup classification,
indicating for instance that the haplogroup assignment
might not be reliable probably due to insufficient data, or
that the haplogroup assignment was uncertain. HaploGrep
however has not been designed to detect artificial recombi-
nants; nonetheless, it seems to perform better than mtDNA-
manager or MitoTool in the sense that HaploGrep detects at
least in some instances the major component of the artificial
mixture. In these cases however, the user is not alerted about
the presence of artificial recombination. For instance, the
first case sample #1 in Table 3is a recombinant L1b × C1,
while HaploGrep (Build 13) indicates a haplogroup M31a1
status.
Given that HaploGrep performs better than other avail-
able bioinformatics tools for haplogrouping, it is worth to
investigate it more deeply. By way of examples (see Tables 2
and 3), we evaluate the algorithm implemented in Haplo-
Grep. From the parsimony point of view, the mismatch
distance between a sample sequence and a haplogroup motif
would be relevant; whereby a moderate scaling relative to
the phylogenetic position of a haplogroup and the number of
seemingly private mutations in the sample may not be
unreasonable. However, in HaploGrep, the phylogenetic
position of a haplogroup is measured by the distance to a
particular extant (contemporary) mtDNA sequence, viz. the
rCRS. As a consequence, the drop from a 100 % rank for
perfect haplogroup allocation to the rank for a slightly
imperfect allocation can be quite drastic for sample sequen-
ces phylogenetically close to the rCRS but very minor for
African haplogroup L0 mtDNA sequences. To illustrate this
reference-associated bias, we have selected six peripheral
haplogroups from the mtDNA phylogeny, each character-
ized by exactly two control-region mutations. We have then
deleted exactly one of the two characteristic variants, first
the one with the higher mutational score and then the other
one. The deletion of one variant had no or very little effect
on the haplogroup assignment returned by HaploGrep
(Build 13) but a considerable effect on ranks (alias quality
scores) of the HV subhaplogroups in contrast to the L0 and
L1 subhaplogroups; see Table 4. A similar effect is caused
by the addition of a private variant (16303A), which occurs
in Build 13 exactly once, namely in the motif of the Aus-
tralian haplogroup O1. However, the addition of a variant
(e.g., 16307G) that is not recorded in PhyloTree has no
effect as such mutations are generally ignored in HaploGrep
(data not shown).
Apart from the examples given in Tables 2and 3, one can
also test how the software implemented in HaploGrep
behaves when dealing with more ambiguous profiles. For
instance, if one tries to classify the profile 16093C (se-
quence range: 16024 to 16400), which could perfectly fit
within, e.g., haplogroup H or many other closely related
haplogroups, it is classified by HaploGrep (Build 13) in
the first place as belonging to the South Asian haplogroup
“R8a1+16093”or equally probable as R8a1b, both with
quality rank 100 %. One can indeed find profiles 16093C
16519C in India (recall that the variation at 16519 is
Int J Legal Med (2012) 126:901–916 911
disregarded in PhyloTree and hence in HaploGrep), but 24
out of 3+29 matches of either profile 16093C or 16093C
16519C in the EMPOP database are from West Eurasia and
therefore not belonging to haplogroup R8. Since 16093 is a
mutational hotspot one avoids the formal erection of sub-
haplogroup status within haplogroups HV or H or H1 etc. on
the basis of this mutation alone, so that 16093 is usually
recorded only as a co-motif mutation which cannot stand by
its own. In fact, in no case is 16093 the single motif muta-
tion of a named haplogroup in PhyloTree; quite often it is
not even clear whether 16093 is part of the motif as indi-
cated by the bracketed notation “(16093)”. For this reason,
R8a1 + 16093 cannot be regarded as a haplogroup; the
mtDNA sequence(s) allocated to this point of the mtDNA
phylogeny would still await characterization of a subha-
plogroup; it is then not clear a priori whether or not
16093C will enter the corresponding motifs (as 16093C
appears on virtually all haplogroup backgrounds). A
100 % match with a motif can therefore not be interpreted
as a successful classification of high quality without further
considerations.
The classification tasks that are most difficult (if not
frustrating) are (partial) control-region profiles nearly iden-
tical to the ancestral (partial) control-region profile of a
haplogroup which has dozens of subhaplogroups without
characteristic mutation in that segment. In these cases, one
would definitely need some coding-region information for
any successful haplogroup allocation. For instance, the an-
cestral motif of haplogroup L3 has only rCRS-variant
16223T within the (HVS-I) range 16024–16400 and this
motif was retained in the descendant haplogroups M and
N. In fact, there are several subhaplogroups of the latter ones
with the same ancestral motif. If no extra mutations were
observed in a sample sequence, then classification would
become a lottery, whether automated or not. The same is
true for the two descendant motifs 16223T 16362C (match-
ing, e.g., haplogroups D, G, M6, and M9) and the rCRS
profile (matching haplogroups R, HV, H, etc.) within that
HVS-I range. When these three ancestral motifs are entered
into HaploGrep, then huge lists of potential haplogroups are
returned for 16223T and 16223T 16362C, but the empty
input motif in the case of the rCRS is rejected, although this
real sequence is found 956 times among 13467 sequences in
EMPOP. In the case of input sequences without any scored
variant, rank calculations are infeasible or meaningless be-
cause of the problem of denominator zero in the formula
employed by HaploGrep.
A general problem with most (if not all) approaches to
automated haplogroup assignment is the restriction to the
current build of PhyloTree, so that the closely related com-
plete mtDNA sequences that are allocated to a specific
haplogroup do not get screened one by one for additionally
matched variants, which would strengthen the proposed
haplogroup assignment. That is, the near-matching strategy
(see, e.g., [54]) is in general not employed with automated
approaches. For instance, the incorrect assignment of sam-
ple no. 10 of Table 2to haplogroup U5b2a1a2 by Haplo-
Grep (Build 13) is due to the conflicting signals at the two
hypervariable sites 16192 and 16311 in the sample for
subclassification within the larger haplogroup U5b2a1a.
Table 4 Quality scores (ranks)
of haplogroup assignment in
HaploGrep [53] for modified
haplogroup control-region
motifs
a
Relative to the rCRS according
to PhyloTree Build 13
(see Supplementary File 2)
Profile with
HG motif
Length of HG
motif string
a
Loss of one motif mutation Gain (and loss) of mutation
Variant Rank Variant(s) Rank
H1b 4 +16356! 90.1 +16303 86.9
+16362! 90.9 +16362! +16303 75.8
HV12a 3 +16220C! 82.8 +16303 87.2
+16292! 84.2 +16292! +16303 67.5
R0a1a2 9 +95C! 93.7 +16303 94.1
+93! 94.0 +93! +16303 87.4
N1b1d 10 +188! 94.0 +16303 94.5
+185! 94.4 +185! +16303 88.3
L3f1b2 10 +16266A! 95.0 +16303 94.7
+16172! 95.6 +16172! +16303 89.9
L2b1a 17 +16355! 98.4 +16303 96.5
+146!! 98.5 + 146!! +16303 95.2
L1b1a1 20 +264! 97.0 +16303 97.0
+16215! 97.0 +16215! +16303 94.0
L0a1c 20 +16287! +1612 98.1 +16303 97.1
9!! 98.2 +16129!! / +16303 95.5
912 Int J Legal Med (2012) 126:901–916
Scanning the dozen complete mtDNA sequences from
U5b2a1a stored in GenBank and conveniently displayed on
the U5b2 Website of Ian Logan reveals a perfect match with
the control-region profile of a member of U5b2a1a1 (Gen-
Bank accession number GU296544), thus solving the issue.
A semi-automated approach is not per se a guarantee for
achieving an appropriate haplogroup allocation, especially
when the initial algorithm sorts mtDNA samples belonging
to single minor haplogroup into distinct slots. This might
have happened, for example, to the three haplogroup R2
members in the dataset of [55], which were differently
labeled as “R”,“R0”, and “R2”, despite the fact that the
characteristic haplogroup R2 variant 16071T has score 1 in
both Soares et al. [11] and Build 13 of PhyloTree (and is
additionally accompanied by the hotspot variant 152C). The
coarse haplogrouping of several haplogroup J2a2a samples
to “J”and one haplogroup L3h1a2a sample to “L3”also
indicates that the manual search strategies were possibly not
yet sufficiently elaborate.
Conclusions
For a non-expert or a novice in mtDNA analysis, it may be
tempting to employ automated tools for haplogroup assign-
ment as an all-in-one solution without carrying out addition-
al analyses. None of the tools tested in the present study are
free of shortcomings in design and scope. In the case of
mtDNAmanager, the classification is based on an algorithm
that was not published in detail and whose performance is
very poor. There is no justification for using this software as
it returns results that are clearly misleading. MitoTool per-
forms better than mtDNAmanager but worse than Haplo-
Grep; in contrast to mtDNAmanager, MitoTool can run
complete mtDNA genomes. When entire genomes are used
in MitoTool instead of partial control-region sequences, the
classification algorithm performs better. The more sophisti-
cated algorithm of HaploGrep performs reasonably well
with a number of control-region profiles but it is currently
not exempt of problems because it delivers haplogrouping
ranks as quality scores that show a strong bias against
sequences phylogenetically close to the rCRS.
Therefore, given the in part questionable performance of
present algorithms for haplogroup classification when using
difficult short-segment profiles, one cannot yet recommend
automated classification as the single strategy for sorting
partial mtDNA sequences; instead, for the time being, a
combined automated and manual approach is more success-
ful. A general drawback of most implemented classification
tools is the restriction to matching haplogroup motifs in one
way or the other. An alternative could be a formal near-
matching strategy operating on a large database of allocated
partial mtDNA sequences. For HVS-I sequences this has
been realized within the Genographic Project but is limited
by insufficient information content of the short HVS-I re-
gion and a far too coarse haplogroup scheme [5]. A combi-
nation of both approaches that incorporates reconstructed
ancestral mtDNA sequences (haplogroup motif sequences)
together with near-matching to a reliable expanded database
would be desirable.
In any case, the user should go through PhyloTree and Ian
Logan’s Web pages manually and check if the haplogroup
status delivered by an automated software or some initial
manual sorting is plausible. Although there may be some dis-
cussion about the way how branching patterns are being rear-
ranged and haplogroups renamed in PhyloTree, this resource
servesasthegoldstandardtowhichanyattemptsofhaplo-
grouping are to be compared. Although the large database of
published complete sequences is not totally error-free, the most
problematic datasets have been disregarded by PhyloTree [56,
57]. If necessary, single dubious sequences in otherwise un-
problematic datasets can be excluded for the near-matching
search [37]. Any missing mutations or hybrid motifs can be
highlighted with the help of this classification tree and the
strategies outlined in [10,58].
The main difficulty is then to decide whether an alloca-
tion of the given sequence to some branch of the mtDNA
tree is either sufficiently convincing or somewhat ambigu-
ous or infeasible as some novel haplogroup may be involved
which has not yet been described. A fully automated hap-
logroup allocation could hardly map expert knowledge into
simple rules in an adequate way, given the fact that available
mtDNA databases, though constantly expanding, just pro-
vide a portion of all published mtDNA variation, which
anyway constitutes only a micro-extract of the real-world
mtDNA variation. Nonetheless, expert knowledge should
gradually evolve into guided semi-automated expert sys-
tems that will assist the user in exploring complete mtDNA
databases more skillfully and evaluating the phylogenetic
position of given partial mtDNA sequences more
thoroughly.
The forensic geneticist should check if all the diagnostic
variants of a targeted haplogroup are present in the sequence
(as highlighted in the output of HaploGrep) that is being
classified; if some are but others are not, one should either
reconsider the haplogroup allocation (in the case that the
positive evidence was weak) or go back to the raw data
(electropherograms) and check if these variants are actually
present or were omitted by mistake or wrongly documented.
If more than one diagnostic mutation is absent, even when
the electropherograms are apparently correct, one should
also suspect a mistake at some step of the analysis. For
instance, artificial recombination cannot be detected by just
inspecting the original electropherograms, because the cause
of the error could be sample mix-up that occurred before the
sequencing stage [8]. If two different segments of the profile
Int J Legal Med (2012) 126:901–916 913
(such as HVS-I and HVS-II) are allocated to different parts
of the phylogeny, then artificial recombination has most
likely occurred [42,43], and therefore, re-analysis of the
sample from the very beginning (DNA extraction) with
forward and reverse sequencing is mandatory. Separate
and non-overlapping sequencing of HVS-I and HVS-II is
prone to sample mixing which results in hybrid profiles, as
is again testified by our re-analysis of two forensic studies.
Therefore such a sequencing scheme should be avoided in
forensic casework [9]. The incidence of artificial recombi-
nation could be reduced by analyzing the same sample
independently by two analysts, and also by using a posteri-
ori phylogenetic monitoring as explained above.
Haplogrouping in the forensic context thus serves
two goals. First, it enables the forensic scientist to
extrapolate from the control-region motif to the
expected coding-region motif which, if necessary, could
be tested site by site for verification. Therefore, it
enables wider comparisons with other closely related
sequences, for which more coding-region information
is available. Second, but perhaps most importantly for
the average forensic laboratory, haplogrouping serves as
an element of an a posteriori quality control. In partic-
ular, it assists in highlighting private mutations, among
which phantom mutations would accumulate and there-
by could come into focus for re-inspection. Even a
single minor error can have important consequences
for database searches or for the comparison of mtDNA
profiles coming from evidentiary samples and a suspect.
Laboratories should not ignore haplogrouping as a tool
that can eventually help avoiding wrong decisions in
court.
Two recent editorials [59,60] in the International
Journal of Legal Medicine and Forensic Science Inter-
national Genetics have set guidelines for future submis-
sion of papers with mtDNA population data. These rules
are important to guarantee a minimum level of quality. Our
reassessment of studies in low-rank forensic journals stresses
that the notorious errors, which could easily be avoided, are
committed over and over again. Unfortunately, in neither
editorial in these journals was the importance of haplogroup-
ing stressed. This may also reflect the fact that only a minority
of articles from these (and other forensic) journals have ex-
plicitly cited PhyloTree in the past 4 years.
Acknowledgements The Ministerio de Ciencia e Innovación
(SAF2008-02971 and SAF2011-26983) and the European project “A
European Initial Training Network on the history, archaeology, and
new genetics of the Trans-Atlantic slave trade (EUROTAST)”(EU
project: 290344) gave support to AS. MvO was supported in part by
the Netherlands Forensic Institute (NFI) and by a grant from the
Netherlands Genomics Initiative (NGI)/Netherlands Organization for
Scientific Research (NWO) within the framework of the Forensic
Genomics Consortium Netherlands (FGCN).
References
1. Parson W, Dür A (2007) EMPOP—a forensic mtDNA database.
Forensic Sci Int Genet 1(2):88–92
2. van Oven M, Kayser M (2009) Updated comprehensive phylogenetic
tree of global human mitochondrial DNA variation. Hum Mutat 30
(2):E386–E394
3. Gómez-Carballa A, Ignacio-Veiga A, Álvarez-Iglesias V, Pastoriza-
Mourelle A, Ruiz Y, Pineda L, Carracedo Á, Salas A (2012) A
melting pot of multicontinental mtDNA lineages in admixed
Venezuelans. Am J Phys Anthropol 147(1):78–87
4. Andrews RM, Kubacka I, Chinnery PF, Lightowlers RN, Turnbull
DM, Howell N (1999) Reanalysis and revision of the Cambridge
reference sequence for human mitochondrial DNA. Nat Genet 23
(2):147
5. Behar DM, Rosset S, Blue-Smith J, Balanovsky O, Tzur S, Comas
D, Mitchell RJ, Quintana-Murci L, Tyler-Smith C, Wells RS
(2007) The Genographic Project public participation mitochondrial
DNA database. PLoS Genet 3(6):e104
6. Bandelt HJ, Quintana-Murci L, Salas A, Macaulay V (2002) The
fingerprint of phantom mutations in mitochondrial DNA data. Am
J Hum Genet 71(5):1150–1160
7. Brandstätter A, Sanger T, Lutz-Bonengel S, Parson W, Beraud-
Colomb E, Wen B, Kong Q-P, Bravi CM, Bandelt HJ (2005)
Phantom mutation hotspots in human mitochondrial DNA.
Electrophoresis 26(18):3414–3429
8. Salas A, Carracedo Á, Macaulay V, Richards M, Bandelt HJ
(2005) A practical guide to mitochondrial DNA error prevention
in clinical, forensic, and population genetics. Biochem Biophys
Res Commun 335(3):891–899
9. Salas A, Bandelt HJ, Macaulay V, Richards MB (2007) Phylogeo-
graphic investigations: the role of trees in forensic genetics. Forensic
Sci Int 168(1):1–13
10. Bandelt HJ, Salas A, Taylor RW, Yao YG (2009) The exaggerated
status of “novel”and “pathogenic”mtDNA sequence variants due
to inadequate database searches. Hum Mutat 30(2):191–196
11. Soares P, Ermini L, Thomson N, Mormina M, Rito T, Röhl A, Salas
A, Oppenheimer S, Macaulay V, Richards MB (2009) Correcting for
purifying selection: an improved human mitochondrial molecular
clock. Am J Hum Genet 84(6):740–759
12. Cardoso S, Villanueva-Millan MJ, Valverde L, Odriozola A, Aznar
JM, Pineiro-Hermida S, de Pancorbo MM (2012) Mitochondrial
DNA control region variation in an autochthonous Basque population
sample from the Basque Country. Forensic Sci Int Genet 6(4):
e106–e108
13. Farris JS (1970) Methods for computing Wagner Trees. Syst Biol 1
(19):83–92
14. Bandelt H-J, Salas A, Bravi CM (2006) What is a 'novel' mtDNA
mutation―and does 'novelty' really matter? J Hum Genet 51
(12):1073–1082
15. Bandelt HJ, Yao YG, Salas A (2008) The search of 'novel' mtDNA
mutations in hypertrophic cardiomyopathy: MITOMAPping as a
risk factor. Int J Cardiol 126(3):439–442
16. Rakha A, Shin KJ, Yoon JA, Kim NY, Siddique MH, Yang IS,
Yang WI, Lee HY (2011) Forensic and genetic characterization of
mtDNA from Pathans of Pakistan. Int J Legal Med 125(6):841–848
17. Yao YG, Kong QP, Man XY, Bandelt HJ, Zhang YP (2003)
Reconstructing the evolutionary history of China: a caveat
about inferences drawn from ancient DNA. Mol Biol Evol
20(2):214–219
18. Umetsu K, Tanaka M, Yuasa I, Adachi N, Miyoshi A, Kashimura
S, Park KS, Wei YH, Watanabe G, Osawa M (2005) Multiplex
amplified product-length polymorphism analysis of 36 mitochon-
drial single-nucleotide polymorphisms for haplogrouping of East
Asian populations. Electrophoresis 26(1):91–98
914 Int J Legal Med (2012) 126:901–916
19. Álvarez-Iglesias V, Jaime JC, Carracedo Á, Salas A (2007) Coding
region mitochondrial DNA SNPs: targeting East Asian and Native
American haplogroups. Forensic Sci Int Genet 1:44–55
20. Quintáns B, Álvarez-Iglesias V, Salas A, Phillips C, Lareu MV,
Carracedo Á (2004) Typing of mitochondrial DNA coding region
SNPs of forensic and anthropological interest using SNaPshot
minisequencing. Forensic Sci Int 140(2–3):251–257
21. van Oven M, Vermeulen M, Kayser M (2011) Multiplex genotyping
system for efficient inference of matrilineal genetic ancestry with
continental resolution. Investig Genet 2(1):6
22. Álvarez-Iglesias V, Barros F, Carracedo Á, Salas A (2008) Mini-
sequencing mitochondrial DNA pathogenic mutations. BMC Med
Genet 9:26
23. Salas A, Amigo J (2010) A reduced number of mtSNPs saturates
mitochondrial DNA haplotype diversity of worldwide population
groups. PLoS One 5(5):e10218
24. Brandstätter A, Salas A, Niederstätter H, Gassner C, Carracedo Á,
Parson W (2006) Dissection of mitochondrial superhaplogroup H
using coding region SNPs. Electrophoresis 27(13):2541–2550
25. Álvarez-Iglesias V, Mosquera-Miguel A, Cerezo M, Quintáns B,
Zarrabeitia MT, Cuscó I, Lareu MV, García O, Pérez-Jurado L,
Carracedo Á, Salas A (2009) New population and phylogenetic
features of the internal variation within mitochondrial DNA macro-
haplogroup R0. PLoS One 4(4):e5112
26. Álvarez-Iglesias V, Salas A, Cerezo M, Ramos-Luis E, Jaime JC,
Lareu MV, Carracedo Á (2006) Genotyping coding region mtDNA
SNPs for Asian and Native American haplogroup assignation. Int
Congress Series 11(1288):4–6
27. Crespillo M, Paredes MR, Prieto L, Montesino M, Salas A, Albarrán
C, Álvarez-Iglesias V, Amorin A, Berniell-Lee G, Brehm A, Carril JC,
Corach D, Cuevas N, Di Lonardo AM, Doutremepuich C, Espinheira
RM, Espinoza M, Gómez F, González A, Hernández A, Hidalgo M,
Jimenez M, Leite FP, López AM, López-Soto M, Lorente JA, Pagano
S, Palacio AM, Pestano JJ, Pinheiro MF, Raimondi E, Ramon MM,
Tovar F, Vidal-Rioja L, Vide MC, Whittle MR, Yunis JJ, Garcia-
Hirschfel J (2006) Results of the 2003-2004 GEP-ISFG collaborative
study on mitochondrial DNA: focus on the mtDNA profile of a mixed
semen-saliva stain. Forensic Sci Int 160(2–3):157–167
28. Ballantyne KN, van Oven M, Ralf A, Stoneking M, Mitchell RJ,
van Oorschot RA, Kayser M (2012) MtDNA SNP multiplexes for
efficient inference of matrilineal genetic ancestry within Oceania.
Forensic Sci Int Genet 6(4):425–436
29. Cerezo M, Černý V, Carracedo Á, Salas A (2009) Applications of
MALDI-TOF MS to large-scale human mtDNA population-based
studies. Electrophoresis 30(21):3665–3673
30. Zuccarelli G, Alechine E, Caputo M, Bobillo C, Corach D, Sala A
(2011) Rapid screening for Native American mitochondrial
and Y-chromosome haplogroups detection in routine DNA
analysis. Forensic Sci Int Genet 5(2):105–108
31. Köhnemann S, Pfeiffer H (2011) Application of mtDNA SNP
analysis in forensic casework. Forensic Sci Int Genet 5(3):216–221
32. Yang Y, Zhang P, He Q, Zhu Y, Yang X, Lv R, Chen J (2011) A
new strategy for the discrimination of mitochondrial DNA
haplogroups in Han population. J Forensic Sci 56(3):586–590
33. Bandelt HJ, Parson W (2004) Fehlerquellen mitochondrialer DNS-
Datensätze und Evaluation der mtDNS-Datenbank D-Loop-BASE.
Rechtsmedizin 14:251–257
34. Zgonjanin D, VeselinovićI, Kubat M, FuračI, Antov M, Lončar E,
TasićM, VukovićR, Omorjan R (2010) Sequence polymorphism
of the mitochondrial DNA control region in the population of
Vojvodina Province, Serbia. Leg Med (Tokyo) 12(2):104–107
35. Mishmar D, Ruiz-Pesini E, Golik P, Macaulay V, Clark AG,
Hosseini S, Brandon M, Easley K, Chen E, Brown MD,
Sukernik RI, Olckers A, Wallace DC (2003) Natural selection
shaped regional mtDNA variation in humans. Proc Natl Acad
Sci U S A 100(1):171–176
36. Gonder MK, Mortensen HM, Reed FA, de Sousa A, Tishkoff SA
(2007) Whole-mtDNA genome sequence analysis of ancient African
lineages. Mol Biol Evol 24(3):757–768
37. Yao Y-G, Salas A, Logan I, Bandelt H-J (2009) mtDNA data
mining in GenBank needs surveying. Am J Hum Genet 85
(6):929–933; author reply 933
38. Chandrasekar A, Kumar S, Sreenath J, Sarkar BN, Urade BP,
Mallick S, Bandopadhyay SS, Barua P, Barik SS, Basu D, Kiran
U, Gangopadhyay P, Sahani R, Prasad BVR, Gangopadhyay S,
Lakshmi GR, Ravuri RR, Padmaja K, Venugopal PN, Sharma MB,
Rao VR (2009) Updating phylogeny of mitochondrial DNA
macrohaplogroup M in India: dispersal of modern human in
South Asian corridor. PLoS One 4(10):e7447
39. Sekiguchi K, Imaizumi K, Fujii K, Mizuno N, Ogawa Y, Akutsu T,
Nakahara H, Kitayama T, Kasai K (2008) Mitochondrial DNA
population data of HV1 and HV2 sequences from Japanese
individuals. Leg Med (Tokyo) 10(5):284–286
40. Parson W, Bandelt HJ (2007) Extended guidelines for mtDNA
typing of population data in forensic science. Forensic Sci Int
Genet 1:13–19
41. Bandelt HJ, Kong QP, Parson W, Salas A (2005) More evidence
for non-maternal inheritance of mitochondrial DNA? J Med Genet
42:957–960
42. Bandelt HJ, Salas A, Bravi CM (2004) Problems in FBI mtDNA
database. Science 305(5689):1402–1404
43. Bandelt HJ, Salas A, Lutz-Bonengel S (2004) Artificial recombi-
nation in forensic mtDNA population databases. Int J Legal Med
118(5):267–273
44. Afonso-Costa H, Carvalho M, Lopes V, Balsa F, Bento AM, Serra A,
Andrade L, Anjos MJ, Vide MC, Pantoja S, Vieira DN, Corte-Real F
(2010) Mitochondrial DNA sequence analysis of a native Bolivian
population. J Forensic Leg Med 17(5):247–253
45. Karachanak S, Carossa V, Nesheva D, Olivieri A, Pala M, Hooshiar
Kashani B, Grugni V, Battaglia V, Achilli A, Yordanov Y, Galabov
AS, Semino O, Toncheva D, Torroni A (2012) Bulgarians vs the
other European populations: a mitochondrial DNA perspective. Int J
Legal Med 126(4):497–503
46. Mikkelsen M, Rockenbauer E, Sørensen E, Rasmussen M,
Børsting C, Morling N (2008) A mitochondrial DNA SNP
multiplex assigning Caucasians into 36 haplo- and subha-
plogroups. Forensic Sci Int: Genet Supplement Series 1
(1):287–289
47. Mikkelsen M, Rockenbauer E, Demir H, Børsting C, Morling N
(2011) Frequencies of 33 coding region mitochondrial SNPs in a
Danish and a Turkish population. Forensic Sci Int Genet 5(5):559–560
48. Mikkelsen M, Sørensen E, Rasmussen EM, Morling N (2010)
Mitochondrial DNA HV1 and HV2 variation in Danes. Forensic
Sci Int Genet 4(4):e87–e88
49. Zimmermann B, Bodner M, Amory S, Fendt L, Röck A, Horst D,
Horst B, Sanguansermsri T, Parson W, Brandstätter A (2009)
Forensic and phylogeographic characterization of mtDNA lineages
from northern Thailand (Chiang Mai). Int J Legal Med 123
(6):495–501
50. Lee HY, Song I, Ha E, Cho SB, Yang WI, Shin KJ (2008)
mtDNAmanager: a Web-based tool for the management and quality
analysis of mitochondrial DNA control-region sequences. BMC
Bioinforma 9:483
51. Monson KL, Miller KWP, Wilson MR, DiZinno JA, Budowle B
(2002) The mtDNA Population Database: an integrated software
and database resource for forensic comparison. Forensic Sci Commun
4:no 2
52. Fan L, Yao YG (2010) MitoTool: a web server for the analysis and
retrieval of human mitochondrial DNA sequence variations.
Mitochondrion 11(2):351–356
53. Kloss-Brandstätter A, Pacher D, Schonherr S, Weissensteiner H,
Binna R, Specht G, Kronenberg F (2011) HaploGrep: a fast and
Int J Legal Med (2012) 126:901–916 915
reliable algorithm for automatic classification of mitochondrial
DNA haplogroups. Hum Mutat 32(1):25–32
54. Kong QP, Bandelt HJ, Sun C, Yao YG, Salas A, Achilli A, Wang CY,
Zhong L, Zhu CL, Wu SF, Torroni A, Zhang YP (2006) Updating the
East Asian mtDNA phylogeny: a prerequisite for the identification of
pathogenic mutations. Hum Mol Genet 15(13):2076–2086
55. Scheible M, Alenizi M, Sturk-Andreaggi K, Coble MD, Ismael S,
Irwin JA (2011) Mitochondrial DNA control region variation in a
Kuwaiti population sample. Forensic Sci Int Genet 5(4):e112–e113
56. van Oven M (2010) Revision of the mtDNA tree and corresponding
haplogroup nomenclature. Proc Natl Acad Sci U S A 107(11):
E38–E39, author reply e40-31
57. Zhao M, Kong QP, Wang HW, Peng MS, Xie XD, Wang WZ,
Jiayang DJG, Cai MC, Zhao SN, Cidanpingcuo TYQ, Wu SF, Yao
YG, Bandelt HJ, Zhang YP (2009) Mitochondrial genome evidence
reveals successful Late Paleolithic settlement on the Tibetan Plateau.
Proc Natl Acad Sci U S A 106(50):21230–21235
58. Kong QP, Salas A, Sun C, Fuku N, Tanaka M, Zhong L, Wang CY,
Yao YG, Bandelt HJ (2008) Distilling artificial recombinants from
large sets of complete mtDNA genomes. PLoS One 3(8):e3016
59. Parson W, Roewer L (2010) Publication of population data of
linearly inherited DNA markers in the International Journal of
Legal Medicine. Int J Legal Med 124(5):505–509
60. Carracedo Á, Butler JM, Gusmão L, Parson W, Roewer L,
Schneider PM (2010) Publication of population data for forensic
purposes. Forensic Sci Int Genet 4(3):145–147
61. Nur Haslindawaty AR, Panneerchelvam S, Edinur HA, Norazmi
MN, Zafarina Z (2010) Sequence polymorphisms of mtDNA HV1,
HV2, and HV3 regions in the Malay population of Peninsular
Malaysia. Int J Legal Med 124(5):415–426
62. Irwin JA, Saunier JL, Beh P, Strouss KM, Paintner CD,
Parsons TJ (2009) Mitochondrial DNA control region variation in a
populationsample fromHong Kong,China. Forensic Sci Int Genet 3
(4):e119–e125
63. Malyarchuk B, Derenko M, Grzybowski T, Perkova M, Rogalla U,
Vanecek T, Tsybovsky I (2010) The peopling of Europe from the
mitochondrial haplogroup U5 perspective. PLoS One 5(4):e10285
64. Tillmar AO, Coble MD, Wallerstrom T, Holmlund G (2010)
Homogeneity in mitochondrial DNA control region sequences
in Swedish subpopulations. Int J Legal Med 124(2):91–98
916 Int J Legal Med (2012) 126:901–916
