Estimation of genomic ancestry in admixed populations

David A. Eccles; Geoffrey K. Chambers; Rodney A. Lea

doi:10.12688/f1000research.8319.1

Home Browse Estimation of genomic ancestry in admixed populations

ALL Metrics

-

Views

-

Downloads

Get PDF

Get XML

Export

▬

✚

Research Article

RETRACTED:

Estimation of genomic ancestry in admixed populations

[version 1; peer review: retracted]

David A. Eccles ^1,2, Geoffrey K. Chambers¹, Rodney A. Lea²

PUBLISHED 29 Apr 2016

Retraction

This paper has been retracted by the authors following advice from the Health and Disability Ethics Committees (HDECs). We have taken this step to respect ...

Author details Author details

¹ School of Biological Sciences, Victoria University, Wellington, New Zealand
² Environmental Science and Research, Porirua, New Zealand

OPEN PEER REVIEW

REVIEWER STATUS

Retraction

This paper has been retracted by the authors following advice from the Health and Disability Ethics Committees (HDECs). We have taken this step to respect the expressed wishes of iwi representatives who have stated that they do not want the results of this study to be published. In light of this, the content of the article has also been removed.
Regrettably this also means that the referee reports from Michael Barnes, Kaustubh Adhikari and Juan Camilo Chacón-Duque have had to be removed; however, we thank them for their help with the peer review process.

Keywords

bootstrap, sub-sampling, Māori population, GWAS, SNPchip, ancestry, admixture

Corresponding author: David A. Eccles

Competing interests: The authors declare that they have no competing interests.

Grant information: Research was partially funded by a Population and Environmental Health scholarship from ESR, Wellington New Zealand (DAE), as well as by an ESR programme grant in Envirogenomics (RAL).

Copyright: © 2016 Eccles DA et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Data associated with the article are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).

How to cite: Eccles DA, Chambers GK and Lea RA. RETRACTED: Estimation of genomic ancestry in admixed populations [version 1; peer review: retracted]. F1000Research 2016, 5:779 (https://doi.org/10.12688/f1000research.8319.1) First published: 29 Apr 2016, 5:779 (https://doi.org/10.12688/f1000research.8319.1) Latest published: 29 Apr 2016, 5:779 (https://doi.org/10.12688/f1000research.8319.1) Retracted: 31 Aug 2016

Comments on this article Comments (0)

Version 1

VERSION 1 PUBLISHED 29 Apr 2016

Author details Author details

¹ School of Biological Sciences, Victoria University, Wellington, New Zealand
² Environmental Science and Research, Porirua, New Zealand

Competing interests

The authors declare that they have no competing interests.

Grant information

Research was partially funded by a Population and Environmental Health scholarship from ESR, Wellington New Zealand (DAE), as well as by an ESR programme grant in Envirogenomics (RAL).

Article Versions (1)

version 1

Published: 29 Apr 2016, 5:779

https://doi.org/10.12688/f1000research.8319.1

Copyright

© 2016 Eccles DA et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Data associated with the article are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).

Download

Export To

metrics

	Views	Downloads
F1000Research	-	-
PubMed Central Data from PMC are received and updated monthly.	-	-

Citations

0

SEE MORE DETAILS

CITE

how to cite this article

Eccles DA, Chambers GK and Lea RA. RETRACTED: Estimation of genomic ancestry in admixed populations [version 1; peer review: retracted]. F1000Research 2016, 5:779 (https://doi.org/10.12688/f1000research.8319.1)

NOTE: If applicable, it is important to ensure the information in square brackets after the title is included in all citations of this article.

track

receive updates on this article

Track an article to receive email alerts on any updates to this article.

Open Peer Review

Retracted

Reviewer Status

Reviewer Reports

	Invited Reviewers
	1	2
Version 1 29 Apr 16	read	read

Michael Robert Barnes, Queen Mary University of London, London, UK
Kaustubh Adhikari, University College London, London, UK

Juan Camilo Chacón-Duque, University College London, London, UK

Comments on this article

All Comments(0)

Add a comment

Sign up for content alerts

Browse by related subjects

Back to all reports

Reviewer Report

13 Views

04 Aug 2016 | for Version 1

Kaustubh Adhikari, Department of Genetics, Evolution and Environment, UCL Genetics Institute, University College London, London, UK

Juan Camilo Chacón-Duque, Department of Genetics, Evolution and Environment, UCL Genetics Institute, University College London, London, UK

13 Views Cite this report Responses(0)

Not Approved

The objective of this manuscript by Eccles et al. is to propose a set of ancestry informative markers (AIMs) for the Maori population in New Zealand to distinguish the admixture between the Maori and European settlers into the country. The aboriginal people of the Australian continent, including the Maori people, have been studied less commonly, so such a study can be an important contribution to the literate.

The main workflow of the manuscript is as follows: The authors take a set of chip genotype data on self-reported Maori individuals, merge it with data from CEU Europeans, to select a set of differentiated SNPs between the two groups through bootstrap. In the selected subset of SNPs they remove SNPs not in HWE, order the SNPs based on allele frequency difference, then perform unsupervised k=2 clustering using Structure, increasing incrementally from top 1 to all 59 SNPs. Structure results with 59 SNPs were used as the benchmark for assessing performance and selecting SNP subsets. Using various criteria a set of 18 SNPs is retained, out of which 14 are available in the validation set consisting of self-reported Maori and European samples. Among these, the top 10 SNPs are selected as the set of recommended AIMs, which are used in the validation set to estimate ancestry fractions in the Maori and European individuals. In a separate analysis on admixed Maori individuals, ancestry estimated with those 10 AIMs is compared to self-reported ancestry.

However, the manuscript is very narrow and weak in its methodology and coverage: for example, the authors develop an unnecessarily complicated method of selecting the AIMs, do not provide any literature review to compare with existing methods, nor do they provide a proper benchmark for their ancestry estimation with the AIMs. Major and minor concerns are explained below in detail.

Major concerns:

The unsupervised Structure run for k=2 using 59 SNPs is taken as the benchmark for comparing the Maori & European individuals in the training set, as well as assessing the performance of the various SNP sets used as AIM candidates. This is a sort of circular reasoning as the set of 59 SNPs were selected as the top set of candidates for AIMs to begin with, so using itself to benchmark performance is inappropriate. Given that the authors started from chip genotype data with >200,000 SNPs on the Maori and Europeans, using an LD-pruned set of SNPs from this dataset to run unsupervised Structure for k=2, which will likely contain several thousand SNPs, is the proper benchmark.

All inferences about the training samples, e.g. which ones are outliers, as well as all statements about performance of SNP sets should be compared with this set of ancestry estimates. This is the common practice in literature regarding selection of AIMs, which the authors do not follow.

Self-reported ancestry is never a reliable benchmark for ancestry studies. It has been shown that reference groups in publicly available datasets assigned entirely to one ancestry group based on self-reporting often have some degree of admixture (e.g. Maples et al., 2013), while self-reported ancestry fractions are often widely inaccurate and depend on various perceived factors (e.g. Ruiz-Linares et al., 2014).
The authors provide no literature review on selection of AIMs, while there are many papers that describe methodology in selection and assessment of AIM sets. Galanter et al. (2012), Ruiz-Linares et al. (2014) are some examples of it in the Latin Americans, which is an even more challenging problem due to three-way admixture. The authors must discuss the methods available in the literature and compare the performance of their AIM selection method with such available methods.
The authors use the statistical method of bootstrapping to obtain their set of top SNPs that differentiate between the Maori and the Europeans. Bootstrapping is a modern and relatively computationally intensive method of repeated sub-sampling from a population, used to estimate the variance (or bias etc.) of a test statistic when such cannot be estimated directly. Considering this, the authors’ use of bootstrapping is unnecessary, as simpler statistical alternatives exist for the intended purpose.

Bootstrapping is used here to select a set of SNPs that consistently show large variation between average allele frequencies in Maori vs. European sub-samples. Since the intention is to achieve large difference in average allele frequency between the two groups, while keeping within-group variation in allele frequency to a minimum, a two-sample t-test for unequal variances is the most relevant test statistic, which is directly applicable to the full set of Maori and European individuals. After calculating the t-test statistic for each SNP, they can be ordered by magnitude and the top SNPs can be chosen. There is no need to perform a bootstrap when variances of allele frequency is directly computable from the genotypes.
The authors do not check if the Maori samples used for constructing the AIMs were admixed or not. The whole-genome LD-pruned SNP set should be used to estimate genetic ancestry, and identify and exclude admixed individuals, as described earlier in point 1. Having admixed individuals in the training set will bias the estimates of ancestry differences. There are several plots in the manuscript that show alarming results which the authors do not identify or address fully:

In Figure 3 showing ancestry proportions in the training samples using 59 SNPs, the Maori individuals have up to 30% European ancestry. More importantly, some CEU individuals seem to have up to 10% Maori ancestry, which is highly unlikely. The results therefore are likely artifacts and indicate potential problems with the method.

In Figure S6 the estimated ancestry values with 10 SNPs are shown for Maori and European individuals, and there is even more variation in estimated ancestry in both Maori and European ancestry. This again points to the possibility of problems in the method.

The self-reported Hawkes Bay Maori (HBM) individuals are taken as the reference group during both training and validation. Yet in both these figures the range of Maori ancestry for the HBM individuals is quite variable. This either indicates that many HBM individuals are admixed and therefore are not usable as reference individuals, or there are problems with the method.

Similarly, the considerable presence of Maori ancestry in the European individuals in those two figures also indicate a potential problem.
As mentioned in points 1 & 4 above, a proper benchmark for the assessment of the 59 starting AIMs or the proposed 10 AIMs is lacking. But comparing the results in the validation set with 10 AIMs from Figure S6 to the results in the training set in Figure 3 with 59 AIMs, Figure S6 shows a much larger variation in ancestry estimates, which indicates that the 10 AIMs subset is possibly not performing very well. But specific performance results similar to Figure S3 are not provided for the proposed set of 10 AIMs.
It is unacceptable that the final list of proposed AIMs is not shared with the public for their use. The main purpose of developing a list of AIMs is usually their sharing with the public so that they can be put to use in other studies. Other papers in the literature who developed a set of AIMs, such as those mentioned in point 2, share the list of AIMs so that others can use or compare them. This saves other researchers from a considerable amount of work of not having to repeat all the steps in genotyping and analysis to obtain their own set of AIMs. So the authors must release the set of SNPs to the public using identifiable IDs such as rs numbers. Such a list will only indicate that there is considerable difference in allele frequency between the Maori and European populations at those SNPs, without divulging further information, so there should be no conflict with individual participant confidentiality.

In fact, genotype or sequence data from nearby aboriginal populations are already publicly available, such as in the Simons Genetic Diversity Project (SGDP), so the extreme redaction by the authors will not be very effective anyway.

Minor concerns:

The reason for selection of CEU as the reference European population is not clear, as no explanations are given. The authors mention “relatively recent admixture with northern Europeans”, though a large number of European settlers were from UK & Ireland which doesn’t count as northern Europe. There are a large number of reference datasets available for various European populations, such as the 1000 Genomes project containing GBR (British) samples. A bigger reference European set of samples would be more appropriate for this study considering the history of European migration into the region.
The manuscript uses many non-standard vocabulary: for example, Group-Specific Marker (GSM) is not a standard term, Ancestry Informative Marker (AIM) or Population-Specific Allele (PSA) should be used depending on the context.
The word ‘genomic’ is ‘genomic ancestry’ is ambiguous. It is more appropriate to use ‘genetic ancestry’, or even better to use ‘autosomal ancestry’.
The authors’ use of the phrase “genomic signature” contradicts with its use in the literature: this term is commonly used for different variants associated to specific phenotypes or biochemical pathways; it is rarely used in a population genetics’ context. What they are trying to describe is a set of SNPs to estimate ancestry proportions, which is quite different from its common use. Also the whole paragraph is rather irrelevant.
Similarly, the use of the phrase “genomic test” is meaningless and inappropriate too.
The authors consistently use the words ‘estimated ancestry’, ‘ancestral fraction’ or ‘Q value’ without referring to the specific ancestry component they are using. The Maori ancestry fraction is 1 minus the European ancestry fraction, therefore it needs to be clarified which specific ancestry component is being discussed.
The use of the word “multi-marker approach” is also incorrect. The authors are simply describing a set of AIMs, which by nature will contain multiple SNPs. This is therefore nothing to highlight about – there are no ancestry estimation methods that use only one SNP, therefore the mention of multiple SNPs is trivial and unnecessary. There are specific cases where ‘multi-marker’ makes a meaningful distinction, e.g. in linkage studies or in GWAS, where the classical way is to analyze one SNP at a time.
There is no point in repeatedly mentioning genome-wide association studies (GWAS), since there is no direct connection between selecting a set of AIMs and performing whole-genome association analysis. The authors do not provide any connection either.
The authors’ use of the statistical term ‘overfitting’ is inappropriate. It refers to problems in prediction arising by using too many unrelated variables, so that the prediction includes a large amount of noise due to the inclusion of those unrelated variables. It has no relevance in the context of GWAS and AIMs selection mentioned here, which themselves are unconnected, as mentioned above.
This sentence is very unclear: “These data were then combined with matching data from the HapMap Project for the CEU population (90 individuals), and filtered to only include SNPs that had different profiles across all individuals in both populations (i.e. no two SNPs had the same reported genotype in all individuals)”
Reference 14 has a typo: Schriver should be Shriver.

References

1. Maples BK, Gravel S, Kenny EE, Bustamante CD: RFMix: a discriminative modeling approach for rapid and robust local-ancestry inference.Am J Hum Genet. 2013; 93 (2): 278-88 PubMed Abstract | Publisher Full Text
2. Ruiz-Linares A, Adhikari K, Acuña-Alonzo V, Quinto-Sanchez M, et al.: Admixture in Latin America: geographic structure, phenotypic diversity and self-perception of ancestry based on 7,342 individuals.PLoS Genet. 2014; 10 (9): e1004572 PubMed Abstract | Publisher Full Text
3. Galanter JM, Fernandez-Lopez JC, Gignoux CR, Barnholtz-Sloan J, et al.: Development of a panel of genome-wide ancestry informative markers to study admixture throughout the Americas.PLoS Genet. 2012; 8 (3): e1002554 PubMed Abstract | Publisher Full Text

Competing Interests

No competing interests were disclosed.

Respond to this report

Responses (0)

Back to all reports

Reviewer Report

24 Views

26 May 2016 | for Version 1

Michael Robert Barnes, William Harvey Research Institute, Queen Mary University of London, London, UK

24 Views Cite this report Responses(0)

Approved With Reservations

Summary of study

Eccles et al present a study of genetic classifiers of Maori ancestry, a population group that represent one of the last major migrations to previously uninhabited islands (about 600-800 years ago) and a rare opportunity to study very recent ancestral population admixture in a large group of genetically homogeneous individuals. They employed a bootstrap sub-sampling method to generate a set of group-specific markers to study admixture within the iwi (Māori tribe) from Hawkes Bay, New Zealand. They identified a panel of 10 markers with good validation and used these to estimate the individual ancestral fraction. The marker panel was used to determine the variation of Māori/European ancestral fraction within the iwi, together with an estimate of European ancestry in the population (28.7%).

Specific Comments
The population in question here is a fascinating and unique opportunity to study recent human history in a population isolate. Overall the study seems sounds, but it fails to address several questions which it might be expected to cover in more details considering the importance of a study of this nature.

Firstly it would be beneficial to compare the Maori data to a wider range of populations other than Caucasians, particularly Polynesians and Asian populations. A PCA or similar visualisation of such a comparison would be of value, and would inform on the likely heritage of the Maori population. It would also be interesting to compare the panel of 10 markers with other markers, such as mitochondrial markers.
The Maori panel is compared against the CEU hapmap panel and a European Parkinson’s disease control population. It is unclear why the later population was used for such a study. This gives the impression that the authors used populations which were conveniently available rather than most suited to the study. If this is not the case then it should be made clear what the rationale was for selection of comparator populations?
The paper fails to reference a number of recent and similar studies in other populations. It should reference these studies or similar and also comment on a comparison of the results seen in these studies with regard to the population structure of the iwi. Three examples of such studies that come to mind are:
- Libiger O, Schork NJ. A Method for Inferring an Individual's Genetic Ancestry and Degree of Admixture Associated with Six Major Continental Populations. Front Genet. 2013 3:322.
- Galanter JM, et al Development of a panel of genome-wide ancestry informative markers to study admixture throughout the Americas. PLoS Genet. 2012;8(3):e1002554.
- Churchhouse C, Marchini J. Multiway admixture deconvolution using phased or unphased ancestral panels. Genet Epidemiol. 2013 Jan;37(1):1-12.
From a Maori ancestry perspective this study might also be considered for comparison with the results here
- Edinur HA, Dunn PP, Hammond L, Selwyn C, Velickovic ZM, Lea RA, Chambers GK.Using HLA loci to inform ancestry and health in Polynesian and Maori populations.Tissue Antigens. 2012 Dec;80(6):509-22.

General comments

Figure 2b, is sound in content, but is somewhat distorted in the manuscript and would benefit from more similar sizing of the identically scaled x and y axis

Technically the paper is reasonably well structured. The title and abstract are accurate and reflect the content of the paper well.

The raw datasets could not be made available due to participant confidentiality, however Anonymised intermediate data is available. This is acceptable considering the nature of the data.

References

1. Libiger O, Schork NJ: A Method for Inferring an Individual's Genetic Ancestry and Degree of Admixture Associated with Six Major Continental Populations.Front Genet. 2012; 3: 322 PubMed Abstract | Publisher Full Text
2. Galanter JM, Fernandez-Lopez JC, Gignoux CR, Barnholtz-Sloan J, et al.: Development of a panel of genome-wide ancestry informative markers to study admixture throughout the Americas.PLoS Genet. 2012; 8 (3): e1002554 PubMed Abstract | Publisher Full Text
3. Churchhouse C, Marchini J: Multiway admixture deconvolution using phased or unphased ancestral panels.Genet Epidemiol. 2013; 37 (1): 1-12 PubMed Abstract | Publisher Full Text
4. Edinur HA, Dunn PP, Hammond L, Selwyn C, et al.: Using HLA loci to inform ancestry and health in Polynesian and Maori populations.Tissue Antigens. 2012; 80 (6): 509-22 PubMed Abstract | Publisher Full Text

Competing Interests

No competing interests were disclosed.

Respond to this report

Estimation of genomic ancestry in admixed populations

Retraction

Retraction

Keywords

Comments on this article Comments (0)

Open Peer Review

Retracted

Reviewer Status

Reviewer Reports

Comments on this article

Browse by related subjects

Competing Interests Policy

Stay Updated