You are on page 1of 5

Molecular Phylogenetics and Evolution 166 (2022) 107328

Contents lists available at ScienceDirect

Molecular Phylogenetics and Evolution


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

Short Communication

The value of updating GenBank accessions for supermatrix phylogeny: The


case of the New Guinean marsupial carnivore genus Myoictis
Matthew J. Phillips a, *, Michael Westerman b, Manuela Cascini a
a
School of Biology and Environmental Science, Queensland University of Technology, 2 George Street, Brisbane 4000, QLD, Australia
b
Department of Ecology, Environment and Evolution, La Trobe University, Melbourne 3086, VIC, Australia

A R T I C L E I N F O A B S T R A C T

Keywords: Erroneous taxonomic attributions in GenBank accessions can mislead phylogenetic inference and appear to be
GenBank widespread within genera. We investigate the influence of taxonomic misattributions for reconstructing the
Marsupials phylogeny of three-striped dasyures, which include four recognized Myoictis species (Marsupialia: Dasyuridae)
Myoictis
that are distributed across New Guinea and nearby islands. Molecular phylogenetic studies that have focused on
Phylogeny
dasyurids consistently resolve the interrelationships of these small carnivores, grouping M. leucura with
Taxonomic misattribution
M. wavicus, and placing M. wallacei and M. melas as successively deeper divergences from these. Two recent
marsupial and mammalian supermatrix phylogenies instead favour an alternative Myoictis topology that is
discordant with each of these relationships. We add new nuclear and mitochondrial sequences and employ
randomized accession resampling that shows the supermatrix topologies are an artefact of several outdated
taxonomic attributions in GenBank. Updating these accessions brings agreement across Myoictis phylogenies with
randomly resampled accessions. We encourage authors to update GenBank taxonomic attributions and we argue
that an option is needed for flagging accessions that are not demonstrably incorrect, but that provide anomalous
results. This would serve both as a caution for future supermatrix construction and to highlight accessions of
potentially significant biological interest for further study.

1. Introduction mammalian phylogenetics was substantially influenced by topological


artefacts stemming from a Canis sequence (Q35873) that is mislabelled
GenBank houses DNA sequences from more than 478,000 formally in GenBank as a bat (Sturnia).
described taxa (Sayers et al., 2018). How many of the millions of ac­ Taxonomic misattributions are rarely reported in the published
cessions have obsolete or erroneous taxonomic attributions is poorly literature and standard GenBank policy is not to update accessions
known, although increasing recent attention suggests >99% fidelity for without written agreement from the original authors/submitters.
generic and higher-level attributions (Leray et al., 2019), but often far Furthermore, the onus is on researchers to seek out and persuade the
more (>5% errors) at the species level (e.g. Locatelli et al., 2020; van original authors (who may no longer be contactable). We believe that
den Burg et al., 2020). The extent to which these taxonomic mis­ GenBank processes should be more conducive to correcting obsolete or
attributions mislead phylogeny reconstruction is little studied. Such incorrect taxonomic attributions. This is increasingly important with
errors may be perceived to be rare and overwhelmed by signal among taxonomically broad supermatrix analyses becoming commonplace and
the correctly attributed majority of loci in multi-locus matrices. Our the authors being further removed from the data acquisition and taxo­
recent work on mammal phylogeny leads us to suspect that erroneous nomic history of the included accessions. Indeed, blame for erroneous
taxonomic attributions may often mislead phylogeny reconstruction. taxonomic attributions should not be assumed to lie with end users or
This may involve minor errors within genera, as occurred with mis­ sequence submitters – they are just the last links along a chain of
attributed wallaby sequences (see Celik et al., 2019). In other cases the possible falliblity from collectors to museum curators and various
errors may have substantial downstream ramifications. We recently re­ intermediary researchers.
ported (Phillips and Shazwani Zakaria, 2021) that Morgan et al.’s Here we consider the case of the New Guinean marsupial genus
(2014) conclusion that mitochondrial DNA is unreliable for ordinal-level Myoictis. These rat-sized, apparently largely terrestrial carnivores are

* Corresponding author.
E-mail address: m9.phillips@qut.edu.au (M.J. Phillips).

https://doi.org/10.1016/j.ympev.2021.107328
Received 18 June 2021; Accepted 12 October 2021
Available online 16 October 2021
1055-7903/© 2021 Elsevier Inc. All rights reserved.
M.J. Phillips et al. Molecular Phylogenetics and Evolution 166 (2022) 107328

widely distributed across forested habitats in New Guinea and sur­ Guinea’s Bird’s Head Peninsula (UPNG3273), which was nested within
rounding islands (Woolley, 2005). Within the family Dasyuridae they M. leucura, and will be discussed below.
are most closely related to small Australian carnivores (Parantechinus,
Dasycercus, Dasykaluta, Dasyuroides) from more open, even arid habitats 2.2. Novel Myoictis DNA sequences
(see Westerman et al., 2016; García-Navas et al., 2020).
Molecular phylogenetic studies that have focused on dasyurids and Additional DNA sequences for two nuclear genes, interphotoreceptor
were undertaken by experts on Australo-Papuan marsupials have binding protein (IRBP) and von Willebrand factor (vWF), and for mito­
consistently resolved the interrelationships of the four Myoictis species, chondrial cytochrome b (Cytb) and 12S rRNA-coding genes were
grouping M. leucura with M. wavicus and placing M. wallacei and generated from tissue samples provided by the South Australian
M. melas as successively deeper divergences from these (e.g. Westerman Museum. These include Myoictis melas from the Torrecelli Mountains of
et al., 2006; Westerman et al., 2016; Kealy and Beck, 2017; see Fig. 1, Papua New Guinea (ABTC1336), Myoictis wallacei from the Aru Islands
Tree 1). In contrast, recent, taxonomically broad supermatrix phylog­ of Indonesia (ABTC50574) and the outgroup taxon, Parantechinus api­
enies across marsupials (May-Collado et al., 2015) and across all calis from Western Australia (ABTC7518).
mammals (Upham et al., 2019) reconstruct a Myoictis topology that is DNA was extracted using QIAGEN’s DNeasy Blood & Tissue Kits. PCR
discordant with both of the expected internal groupings (Fig. 1, Tree 2). reactions contained 9.5 µL of autoclaved H2O, 12.5 µL of Bioline MyFi
These supermatrix analyses include M. wavicus sequences that are mis­ mix, forward and reverse primers (10 pmol/µL) at 1 µL each, 0.25 µL of
attributed to M. melas on GenBank. Identifying and filtering out such Taq polymerase, 1 µL of DNA template, for a total volume of 25 µL. PCR
errors is less likely in supermatrix analyses, because they typically products were checked by electrophoresis in 1% TBE agarose gel stained
include only one exemplar of each species per gene, which in turn are with Gel Red (Biotium) alongside the BioLine HyperladderTM molecular
often obtained from different specimens. weight marker for comparison. PCR products were purified for
We investigate the basis for Myoictis misattributions and test sequencing using an UltraCleanTM PCR clean up kit (Bioline). PCR
whether, despite being only a small minority of accessions, they may primers and cycle conditions are provided in the supplementary mate­
mislead phylogeny reconstruction. This could be facilitated by errors rial (Table S1). Amplicons were Sanger sequenced using ABI, BigDye
being taxonomically consistent rather than randomly distributed. Upon chemistry at Macrogen (Korea). Base-calling ambiguities were visually
correcting erroneous GenBank accessions and adding additional se­ inspected and corrected as appropriate. GenBank accession numbers are
quences to fill nuclear and mitochondrial sampling gaps we clarify the provided in Tables S2 and S3.
phylogenetic relationships among Myoictis.
2.3. Phylogenetic analysis
2. Materials and methods
The pre-existing and novel DNA sequences for all loci that were
2.1. Taxonomic attribution and provenance of Myoictis GenBank attributed to at least three Myoictis species (thus, with potential to be
accessions directly phylogenetically informative) were concatenated for individual
specimens. These loci include nuclear IRBP, vWF, ε-globin, bfib7 and
Each GenBank accession for Myoictis was cross-checked against (1) mitochondrial tPhe-12S-tVal-16S, Cytb. Relevant dasyurine outgroup
sample records in the original published study, (2) extraction and taxa were added (Parantechinus apicalis, Dasykaluta rosamondae, Dasy­
sample donation notes made by one of us (MW) and sample provenance cercus cristicauda, Dasyuroides byrnei, Sarcophilus harrisii, Dasyurus hal­
data available in both Woolley (2005) and within the online zoological lucatus, Neophascogale lorentzii, Phascolosorex dorsalis). All sequences
collections of Australian museums (OzCam). This enabled us to assign were visually inspected and manually aligned in Se-Al 2.0a (Rambaut,
museum specimen numbers and reference numbers from Woolley 1996).
(2005) to each accession. Nine of 79 accessions were found to have We inferred tree topologies using maximum likelihood (IQ-TREE,
obsolete or otherwise erroneous taxonomic attributions. Phylogenies Nguyen et al., 2015) and Bayesian inference (MrBayes 3.2.7, Ronquist
were then reconstructed for each gene and this confirmed the species- et al., 2012). Alternative partitioning schemes were considered for the
level taxonomic attributions expected from our cross-checking. One nuclear data, either modelling genes separately or modelling codon
possible exception is M. melas from the Arfak Mountains of New positions separately. In each case the mtDNA was partitioned into RNA

Fig. 1. The affect of updating GenBank ac­


cessions on Myoictis phylogeny reconstruc­
tion. A. Myoictis phylogenies including Tree
1, the expected topology as inferred in mo­
lecular phylogenetic analyses focusing on
dasyurid marsupials, Tree 2, the topology
reconstructed in recent mammal and
marsupial-wide molecular supermatrix
studies, and Trees 3 & 4, which are two
alternative topologies. B. Percentage of trees
supported from among 100 maximum likeli­
hood analyses in which one of the available
GenGank accessions for each nuclear and
mitochondrial gene was randomly sampled
for each species, initially with the original
GenBank taxonomic attributions and then
with our updated taxonomic attributions.

2
M.J. Phillips et al. Molecular Phylogenetics and Evolution 166 (2022) 107328

stem and loop positions and into codon positions for Cytb, following 2.4. Simulated supermatrix phylogeny with randomized sampling of
Celik et al. (2019). Models for each partition were selected by Model­ GenBank accessions
Finder within IQ-TREE (see Table S4). Maximum likelihood analyses
were performed with branch lengths estimated independently (-sp op­ To investigate the effect of taxonomic misattributions on super­
tion) and with proportional branch lengths among partitions (-spp op­ matrix reconstruction of Myoictis phylogeny we simulated supermatrices
tion). Both corrected AIC and BIC favoured the codon-wise and -spp by randomly sampling one exemplar from each species for each of the
partitioning options. Ultrafast bootstrap approximation (1000 repli­ seven nuclear and mitochondrial genes (for consistency among acces­
cates) was employed for assessing confidence in the resulting topologies. sions, partial tPhe and tVal sequences were excluded, and 12S and 16S
For Bayesian inference in MrBayes, two independent analyses were rRNA were sampled separately). Outgroup sampling was retained from
each run with three Markov Chain Monte Carlo (MCMC) chains for five the primary phylogenetic analyses and held constant. In this way 100
million generations. The same gene-wise and codon-wise partitions were supermatrices with randomized accessions for Myoictis were built. Six of
used as described above for ML. Models were unlinked across all parti­ the 51 relevant Myoictis accessions were originally misattributed. The
tions for the substitution matrix, state frequencies (empirical), invariant maximum likelihood tree was inferred for each matrix, using IQ-TREE
sites and the shape parameter of the rates-across-sites gamma distribu­ (modelled as above) for two sets of analyses, one with the original
tion. Branch lengths were unlinked between the nuclear and mtDNA taxonomic misattributions and another with the updated taxonomic
data, but were proportionally scaled across partitions within each of attributions.
these genomes. Trees were sampled every 5000 generations, with the
first 25% discarded as burn-in. Clade frequencies across the two inde­ 3. Results and discussion
pendent runs reached convergence (clade frequency standard de­
viations < 0.01) and estimated sample sizes for likelihood, prior and 3.1. Myoictis phylogeny
substitution parameter estimates were all above 200 (Tracer v1.7.1,
Rambaut et al., 2018). We identified nine of the 79 Myoictis GenBank accessions as being
To examine the affinities of the Arfak Myoictis specimen (UPNG3273) attributed to the incorrect species. Our investigation was guided by the
additional 12S rRNA ML and Bayesian inference trees were recon­ gene trees we reconstructed, but these alone cannot distinguish between
structed, based on the above methods. These analyses included all misattribution, hemiplasy, erroneous taxonomy and sometimes phylo­
available Myoictis 12S rRNA accessions. We added two previously un­ genetic uncertainty for the less informative nuclear loci. Confident
published 12S rRNA sequences that were obtained as part of Westerman attribution of accessions to Myoictis species was further facilitated by a
et al.’s (2006) study. These were M. melas (ANWC M35780) and combination of sample numbers being included in accessions or asso­
M. leucura (QM J672), respectively from the Frieda River and Vanapa ciated publications (e.g., Krajewski et al., 2004; Westerman et al., 2006)
River Valley, both in Papua New Guinea. Several of the sequences are and record keeping by the original authors. The majority (6) of the
short (390–500 bp), so to ensure the appropriate phylogenetic context erroneous attributions were for older accessions that retained obsolete
the relationships among the outgroup and monohyly of each species taxonomy, with M. wavicus specimens listed as M. melas before the
were enforced in accordance with the multi-gene tree (Fig. 2), but with former taxon was erected as a full species from among the formerly
relationships within species unconstrained and the Arfak and new se­ paraphyletic M. melas. Three more recent M. wallacei accessions were
quences allowed to float unconstrained across the tree. mistakenly labelled as M. wavicus. All nine accessions have subsequently
been updated in GenBank.
Upon updating the taxonomic attributions our phylogenetic analyses

Fig. 2. Myoictis phylogeny inferred from the


combined nuclear (IRBP, vWF, ε-globin, bfib7) and
mitochondrial (tPhe-12S-tVal-16S, Cytb) dataset.
Individual specimens are denoted with collection
numbers (and W-reference numbers from Wool­
ley, 2005). Node support values are Bayesian
inference posterior probability (MrBayes, above)
and ML bootstrap (IQ-TREE, below). Asterisks
indicate full support for both analyses. These re­
sults are with codon-wise partitioning. Gene-wise
partitioning favoured the same topology. The
placement of the Arfak mountains Myoictis was
inferred in a separate analysis that included only
12S rRNA (Fig. S1).

3
M.J. Phillips et al. Molecular Phylogenetics and Evolution 166 (2022) 107328

reconstructed each Myoictis species as monophyletic, except one spec­ sequences and to encourage authors to check and update their acces­
imen attributed to M. melas from the Arfak Mountains (from only 392 bp sions. Any such system should also limit the potential for identifiers to be
partial 12S rRNA sequence, AY730705) falls within M. leucura, and is stigmatized and for vexatious bullying of original authors in the already
discussed below. Otherwise, the monophyly of each species and inter- competitive and often adversarial field of systematics. Linking acces­
specific relationships are all resolved with strong ML bootstrap and sions to collection specimen numbers is one potential solution (Kealy
Bayesian posterior probability (Fig. 2) and in agreement with Wester­ et al., 2019), but may be precluded by funding shortfalls at many
man et al. (2006, 2016). One caution here is that the signal for the institutions.
placement of M. leucura is dominated by mtDNA, since only the poorly We propose that accession anomalies could be brought to GenBank’s
informative ε-globin is included for this species from among the nuclear attention privately, by identified individuals. The evidence provided
loci. The dibbler (Parantechinus apicalis) from Western Australia is fav­ would be considered for veracity and worthiness for passing on to the
oured as sister to Myoictis (Fig. 2) in agreement with most recent studies original authors, who would either agree to alter the accession, decline
(e.g., Mitchell et al., 2014; Kealy and Beck, 2017). or alternatively, flag the accession. Flagging would not distinguish be­
tween errors and biological anomalies. Any counter evidence provided
3.2. Taxonomic misattributions mislead randomized accession by the original authors would be considered by GenBank upon deciding
supermatrix phylogeny whether or not to flag the accession. Other changes would not be made
without permission from the author(s). This process would be separate
Supermatrices were built by randomly sampling one accession per from the standard review process in which GenBank staff may label
species for each gene, to investigate whether being unaware of the small sequences as ‘UNVERIFIED’ until submitters can verify queries.
proportion (12%) of misattributed sequences was likely to mislead ‘UNVERIFIED’ sequences do not appear in BLAST searches, whereas
phylogeny reconstruction. Remarkably, a Myoictis topology that is flagged sequences would, given their highlighted potential biological
incongruent with the expected tree (see Fig. 1) was recovered from 71% interest. NCBI intend to discuss this and other models for sequence
of the randomised accession supermatrices. Almost all of these were the flagging (C. Schoch, pers. comm.).
putatively incorrect topology inferred by the recent supermatrix studies We believe that the “Arfak Myoictis (W130)” partial 12S rRNA
(May-Collado et al., 2015; Upham et al., 2019). Upon updating the accession (AY730705) is a candidate for flagging. It remains the only
taxonomic accessions the expected tree was recovered for all rando­ sequence generated to date from a specimen collected from Nenei in the
mised accession supermatrices. Arfak mountains of New Guinea. Although attributed to M. melas, our
The apparent paradox that a small proportion of misattributed se­ 12S rRNA ML and Bayesian analyses nest the sequence within M. leucura
quences can mislead a substantial majority of the random accession (Fig. S1). Despite its short length (392 bp) there is enough signal for KH
supermatrices may largely be explained by the misattributions being testing (in IQ-TREE) to reject placing the Arfak sequence within M. melas
non-randomly distributed among species. Five of the six misattributed at P = 0.0621. Notably, Menzies (1991) who collected the sample ap­
accessions from which the supermatrices were assembled involve pears to have considered all Myoictis to belong to M. melas, which might
M. wavicus sequences that were attributed to M. melas. Moreover, three help to explain propagation of obsolete taxonomy. However, if the
mtDNA loci (Cytb, 12S and 16S rRNA) disproportionately contribute sample is indeed M. leucura, it was collected more than 1000 km from its
phylogenetic signal due to their rapid evolution and coverage across all nearest conspecific records on the other side of New Guinea and would
Myoictis species. Prior to our updates one of the three ‘M. melas’ acces­ provide a considerable range extension for this species. Flagging sam­
sions for each of these three loci was in fact M. wavicus – giving a 70.4% ples such as this should serve both as a caution for supermatrix inclusion
probability of at least one of these misattributed mtDNA accessions and to encourage and target further biological research; in the present
being randomly included. Another factor in propagating error that de­ case, into the systematics and biogeography of Myoictis.
serves further consideration with simulations is that alternative topo­
logical hypotheses are often distinguished by short internal branches. CRediT authorship contribution statement
The concern here is that near-identical sequences between conspecifics Matthew J. Phillips: Conceptualization, Methodology, Formal
that are incorrectly assumed to be different species for one gene may analysis, Writing – original draft, Writing – review & editing, Visuali­
outweigh the sum of weaker signals for the true sister grouping across zation, Supervision. Michael Westerman: Data curation, Investigation,
several correctly attributed genes. Writing – review & editing. Manuela Cascini: Data curation, Investi­
Not only can a small proportion of erroneous (or obsolete) taxonomic gation, Writing – review & editing.
attributions mislead phylogeny reconstruction, but subsequent down­
stream inferences may also be mislead, such as tracing ancestral states
and molecular dating. We have not estimated divergence dates because Declaration of Competing Interest
there are no reliable fossil calibrations among Myoictis or its close out­
groups. However, we note that randomized accession supertrees with The authors declare that they have no known competing financial
misattributions that mislead the phylogeny have average tip to Myoictis interests or personal relationships that could have appeared to influence
crown lengths / stem length far lower (0.65) than the average for trees the work reported in this paper.
with correctly attributed accessions (1.07). This suggests that substan­
tial divergence estimation error is likely. Acknowledgements

3.3. Flagging anomalous accessions We thank Jim Menzies, Stephen Donnellan and the late Ken Aplin for
making Myoictis material available. Conrad Schoch at NCBI provided
Little attention has been paid in the literature to erroneous or sur­ valuable comments on the manuscript and additional guidance on
prising sequences, except in human and ancient DNA research, where GenBank processes. Two anonymous reviewers provided valuable
identifying contamination is fundamental (Llamas et al., 2017). NCBI do feedback on the manuscript.
provide a range of tools to assist taxomnomic assignment and to help
identify rogue sequence taxonomy (e.g. Ciufo et al., 2018; Schoch et al.,
2020). However, the requirement of GenBank to seek and receive Funding
agreement from original authors/submitters in order to correct or place
notes of caution on accessions is clearly a source of inertia. A more This work was supported by a Discovery Project grant to MJP from
flexible system is needed to make users aware of potential errors in the Australian Research Council (DP190103636).

4
M.J. Phillips et al. Molecular Phylogenetics and Evolution 166 (2022) 107328

Appendix A. Supplementary material Menzies, J.I., 1991. A Handbook of New Guinea Marsupials and Monotremes. Kristen
Press, Madang.
Mitchell, K.J., Pratt, R.C., Watson, L.N., Gibb, G.C., Llamas, B., Kasper, M., Edson, J.,
Supplementary data to this article can be found online at https://doi. Hopwood, B., Male, D., Armstrong, K.N., et al., 2014. Molecular phylogeny,
org/10.1016/j.ympev.2021.107328. biogeography, and habitat preference evolution of marsupials. Mol. Biol. Evol. 31,
2322–2330. https://doi.org/10.1093/molbev/msu176.
Morgan, C.C., Creevey, C.J., O’Connell, M.J., 2014. Mitochondrial data are not suitable
References for resolving placental mammal phylogeny. Mammal. Genome 25 (11-12), 636–647.
Nguyen, L.-T., Schmidt, H.A., von Haeseler, A., Minh, B.Q., 2014. IQ-TREE: a fast and
Celik, M., Cascini, M., Haouchar, D., Van Der Burg, C., Dodt, W., Evans, A.R., Prentis, P., effective stochastic algorithm for estimating maximum-likelihood phylogenies. Mol.
Bunce, M., Fruciano, C., Phillips, M.J., 2019. A molecular and morphometric Biol. Evol. 32, 268–274.
assessment of the systematics of the Macropus complex clarifies the tempo and mode Phillips, M.J., Shazwani Zakaria, S., 2021. Enhancing mitogenomic phylogeny and
of kangaroo evolution. Zool. J. Linn. Soc. 186, 793–812. https://doi.org/10.1093/ resolving the relationships of extinct megafaunal placental mammals. Mol.
zoolinnean/zlz005. Phylogenet. Evol. 158, 107082. https://doi.org/10.1016/j.ympev.2021.107082.
Ciufo, S., Kannan, S., Sharma, S., Badretdin, A., Clark, K., Turner, S., Brover, S., Rambaut, A., 1996. Sequence alignment editor. Available at: http://tree.bio.ed.ac.
Schoch, C.L., Kimchi, A., DiCuccio, M., 2018. Using average nucleotide identity to uk/software/seal.
improve taxonomic assignments in prokaryotic genomes at the NCBI. Int. J. Syst. Rambaut, A., Drummond, A.J., Xie, D., Baele, G., Suchard, M.A., 2018. Posterior
Evol. Microbiol. 68, 2386–2392. summarisation in Bayesian phylogenetics using Tracer 1.7. Syst. Biol. 67, 901–904.
García-Navas, V., Kear, B.P., Westerman, M., 2020. The geography of speciation in Ronquist, F., Teslenko, M., van der Mark, P., Ayres, D.L., Darling, A., Höhna, S.,
dasyurid marsupials. J. Biogeog. 47, 2042–2053. Larget, B., Liu, L., Suchard, M.A., Huelsenbeck, J.P., 2012. MrBayes 3.2: efficient
Kealy, S., Beck, R., 2017. Total evidence phylogeny and evolutionary timescale for Bayesian phylogenetic inference and model choice across a large model space. Syst.
Australian faunivorous marsupials (Dasyuromorphia). BMC Evol. Biol. 17, 240. Biol. 61, 539–542.
https://doi.org/10.1186/s12862-017-1090-0. Sayers, E.W., Cavanaugh, M., Clark, K., Pruitt, K.D., Schoch, C.L., Sherry, S.T., Karsch-
Kealy, S., Donnellan, S.C., Mitchell, K.J., Herrera, M., Aplin, K., O’Connor, S., Louys, J., Mizrachi, I., 2018. GenBank. Nucleic Acids Res. 49, D92–D96. https://doi.org/
2019. Phylogenetic relationships of the cuscuses (Diprotodontia: Phalangeridae) of 10.1093/nar/gkaa1023.
island Southeast Asia and Melanesia based on mitochondrial ND2 gene. Australian Schoch, C.L., Ciufo, S., Domrachev, M., Hotton, C.L., Kannan, S., Khovanskaya, R.,
Mammalogy 42, 266–276. https://doi.org/10.1071/AM18050. Leipe, D., Mcveigh, R., O’Neill, K., Robbertse, B., et al., 2020. NCBI Taxonomy: a
Krajewski, C., Moyer, G.R., Sipiorski, J.T., Fain, M.G., Westerman, M., 2004. Molecular comprehensive update on curation, resources and tools. Database 2020, baaa062.
systematics of the enigmatic “Phascolosoricine” marsupials of New Guinea. Aust. J. https://doi.org/10.1093/database/baaa062.
Zool.. 52 (4), 389. https://doi.org/10.1071/ZO04020. Upham, N.S., Esselstyn, J.A., Jetz, W., Tanentzap, A.J., 2019. Inferring the mammal tree:
Leray, M., Knowlton, N., Ho, S.-L., Nguyen, B.N., Machida, R.J., 2019. GenBank is a species-level sets of phylogenies for questions in ecology, evolution, and
reliable resource for 21st century biodiversity research. PNAS 116, 22651–22656. conservation. PLoS Biol. 17 (12), e3000494. https://doi.org/10.1371/journal.
https://doi.org/10.1073/pnas.1911714116. pbio.3000494.
Llamas, B., Valverde, G., Fehren-Schmitz, L., Weyrich, L.S., Cooper, A., Haak, W., 2017. van den Burg, M.P., Herrando-Pérez, S., Vieites, D.R., 2020. ACDC, a global database of
From the field to the laboratory: controlling DNA contamination in human ancient amphibian cytochrome-b sequences using reproducible curation for GenBank
DNA research in the high-throughput sequencing era. Sci. Technol. Arcaeol. Res. 3 records. Sci. Data 7, 268. https://doi.org/10.1038/s41597-020-00598-9.
(1), 1–14. Westerman, M., Krajewski, C., Kear, B.P., Meehan, L., Meredith, R.W., Emerling, C.A.,
Locatelli, N.S., McIntyre, P.B., Therkildsen, N.O., Baetscher, D.S., 2020. GenBank’s Springer, M.S., 2016. Phylogenetic relationships of dasyuromorphian marsupials
reliability is uncertain for biodiversity researchers seeking species-level assignment revisited. Zool. J. Linn. Soc. 176 (3), 686–701.
for eDNA. PNAS 117 (51), 32211–32212. https://doi.org/10.1073/ Westerman, M., Young, J., Donnellan, S., Woolley, P.A., Krajewski, C., 2006. Molecular
pnas.2007421117. relationships of New Guinean three-striped dasyures, (Myoictis, Marsupialia:
May-Collado, L.J., Kilpatrick, C.W., Agnarsson, I., 2015. Mammals from ‘down under’: a Dasyuridae). J. Mammal Evol. 13 (3-4), 211–222.
multi-gene species-level phylogeny of marsupial mammals (Mammalia, Metatheria). Woolley, P.A., 2005. Revision of the Three-striped Dasyures, Genus Myoictis
PeerJ. 3. (Marsupialia: Dasyuridae), of New Guinea, with a description of a new species. Rec.
Aust. Mus. 57, 321–340.

You might also like