Genetic Ancestry Inference: from AIMS to WGS application and challenges in admixed population

doi:10.21203/rs.3.rs-1824967/v1

Download PDF

Article

Genetic Ancestry Inference: from AIMS to WGS application and challenges in admixed population

https://doi.org/10.21203/rs.3.rs-1824967/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Increasingly, the inference of genetic ancestry plays a prominent role in clinical, population and forensic genetics studies. Over the last few decades, several genotyping strategies and analytical methodologies have been developed in order to assign individuals to specific biogeographic regions. However, despite all these efforts, the ancestry inference in populations with a recent history of admixture, such as those in America, is still a challenge. In admixed populations, proportion and components of genetic ancestry vary at different levels: (i) between populations; (ii) between individuals of the same population; (iii) throughout the individual's genome. In the present study we compared and evaluated different sets of markers, from those with small numbers of ancestry informative markers panels (AIMs), to high-density SNPs (HDSNP) and whole-genome-sequence (WGS) data. To this end, we evaluated 1,675 admixed samples from America, using a tetrahybrid admixture model (Native American, European, African and East Asian). Analyses show greater variation in the correlation coefficient of ancestry components within and between admixed populations, especially for minority ancestral components. We also observed, the higher the number of markers in the AIMs panel, the higher the correlation with HDSNP and WGS. In addition, the greater the number of markers, the more robust was the tetrahybrid admixture model.

Understanding how human genetic diversity is distributed and its implications has been a recurrent focus in clinical, population and forensic genetics [1, 2, 3]. Since the 1970s, with the pioneering studies of Richard Lewontin, it has been known that most human genetic variation occurs between individuals of the same population group, while genetic variation between individuals of distant populations is restricted to a small proportion of the human genome. This study is one of the first to refute the use of social races in biological studies and to draw attention to the fact that genetic information is more accurate for biological issues than social groups or ethno-racial self-declaration [4].

In this sense, the identification of an individual's biogeographic origin, based on genetic variations, seeks to address specific biological issues. For example, in forensic genetics, besides identifying the individual via CODIS (Combined DNA Index System), it is often necessary to provide additional information such as phenotypic characteristics and/or biogeographic origin [5]. In clinical genetics, some diseases are recognized as having different incidences among population group (e.g. Chronic kidney disease [6], hypertension [7], inflammatory bowel disease [8]) and the correct identification of individuals in these groups helps in the development of personalized medical care [9].

The inference of genetic ancestry has also been an important tool in the study of admixed populations. It has been widely used to control the effect of population stratification in association studies [10] and for the identification of disease-associated genes (review in [11]). In addition to its important role in revealing the past history of admixed populations, such as origin, number of migratory waves and the dating of migratory events [12, 13].

Over the last few decades, sets with a few dozen of ancestry informative markers (AIMs) have been optimized in order to compose informative ancestry panels with the purpose of detecting genetic differentiation in continental and subcontinental population groups [14, 15, 16]. The development of AIMs panels’ attempts to select a group of genetic markers that compose a small, accurate, low-cost and highly informative set. Usually, studies use only one AIM panel or even a complementary set (e.g. PIMA + 34plex [17], PACIFICPlex + 34plex [18], and KiddLab + Seldin + 34plex [19]).

In general, the AIMs panels vary in some characteristics such as: specific loci, number of loci, genotyping strategies and parental reference populations [20, 21, 22]. These observations raise questions about which panel is best suited for the purposes of each study. As well, they stimulate discussions about the need for collective efforts for the development of a panel for global use, which has a small number of markers, but with high informativeness [22, 23, 24].

On the other hand, the emergence of high-throughput genotyping technologies (e.g. high-density SNPs array (HDSNPs), whole-exome-sequence (WES), whole-genome-sequence (WGS)) has allowed high horizontal genome coverage studies for inference of genomic ancestry [25, 26]. In this scenario, there is not only a significant increase in the number of genetic markers evaluated, but also, the number of genotyped individuals per study grows from hundreds to thousands [27, 28, 29]. As a consequence, some analytical approaches have been adapted (e.g. [30, 31, 32]), as well new ones have been developed (e.g. inference of local ancestry at each locus of the genomes of admixed individuals (review in [33]).

Additionally, in either AIMs or high-throughput genotyping strategy, the proper inference of the genetic ancestry is dependent on the existence of a reference population’s panel for each ancestral component under study. The inference of genetic ancestry is still deficient in some population subgroups due to the lack of collection of reference data, in addition to some inconclusive or non-validated studies. For example, there is an effort in the search for development of reference panels for Asians and Native Americans, as well as reference genomes for Latin Americans [34].

Admixed populations, such as those in Latin America, are characterized by high genetic heterogeneity, resulting from the complex processes of admixture of Native Americans, with Europeans, Africans [35, 36] and more recently East Asians [37]. Therefore, the correct assignment of each genetic ancestry component is essential for studies with this population group. Concerned with this issue, some studies have compared AIMs panels in admixed American populations, and differences in ancestry proportions inference were observed between panels, which may be related to both the number of markers and the parental reference populations of each panel [17, 38, 39]. However, these studies are based on the trihybrid admixture model (Native American, European and African) and focus only on AIMs or pharmacogenomics’ panels. In the present study, we draw attention and evaluate the efficiency of ancestry inferences for the tetrahybrid model, which also takes into account the migrations from East Asia, intensified from the 20th century onwards, which took tens of millions of people, in particular, to the two largest countries in America: the United States and Brazil. Furthermore, in our analyses we sought to compare ancestry inferences from panels with small numbers of markers to analyses involving WGS.

To do so, we analyzed 1,675 admixed samples from the American continent (Supplementary Table S1) to which we compared: (i) 7 AIMS panels; (ii) a high-density array of SNPs, which was specially developed for population genetics studies, therefore without biased markers such as the SNP array of GWAS, and (iii) whole genome sequencing data.

SNP overlap between AISNP panels

As reported in other studies, we also observed the low overlap of SNPs for the set of AIMS panels evaluated (Fig. 1; Supplementary Table S2). Of the total set of 672 AISNPs, only 25 SNPs are overlapped (3.7%) with one or more AISNP panels (Fig. 1). The panel with the highest number of overlaps is 55AISNP (3 SNPs with 12AISNP; 4 with 34AISNP SNPs; 2 with PIMA SNPs; 13 with 128AISNP and 3 with 446AISNP).

Linkage disequilibrium

The comparison between AISNP panels and WGS data requires the same methodology for ancestry inference. Since the ADMIXTURE model assumes independent segregation between SNPs (Alexander et al., 2009), dense marker sets should be pruned to mitigate background linkage disequilibrium. To determine the optimal linkage disequilibrium coefficient for our WGS dataset, we evaluated several values (r² = 0.01, 0.05, 0.1, 0.3 and 0.5).

Pairwise comparative analysis using these different linkage disequilibrium coefficients in parental populations shows a high degree of correlation between them (r² = 0.997-1). When linkage disequilibrium coefficients were > = 0.1, it was observed a smaller dispersion for the whole ancestry inference distribution (Supplementary Figure S2). That was also observed when the correlation coefficients were equal to 1 between pairs of panel sets, especially for East Asian and Native American samples, which had slightly more discrepant values throughout the comparisons (Supplementary Figure S3B and S3D). Based on these results, subsequent analyses were performed with a linkage coefficient equal to 0.1 for the HDSNPs panel and WGS data.

Ancestry Inference in Parental Populations Groups

Correctly assigning individuals to their original continental groups is the proposal of panel sets, especially AISNPs. However, there are factors such as the number of markers and the set of populations used to establish allelic frequencies of the continental group, which may influence the accuracy of ancestry estimates. To better assess these issues we analyzed the influence of the number of markers when comparing the inference of ancestry in our panel sets for the four parental continental groups (Africa, Europe, East Asia, and Native American).

First, we evaluated the accuracy of the panel sets in correctly assigning individuals from each of the four continental groups. We found only a small proportion of samples not correctly assigned (Z-score >|3|; p-value 0.01; ranging from 3–1% in the HGDP and from 0.7–0.3% in the 1KGP) (Supplementary Tables S3-S5). The lowest performing panels were 12AISNP in the HGDP (~ 3% total; 5% in EAS inference) and 34AISNP in the 1KGP (0.7% total; 1.19% EUR).

Second, for each set of panels and continental group, we verified the distribution of the inferred proportion of individual ancestry. These results show a trend for the four continental groups, in which the greater the number of markers in the set, the smaller the dispersion of the inferred ancestry proportion and the higher the mean and median values for the group (Fig. 2).

Finally, a pairwise comparison of the inferred proportion of individual ancestry was performed between each panel set for each continental group (Supplementary Figure S4A-G). The 10 panel sets evaluated showed high correlation coefficient values (r² > 84%). For some groups such as African and European in the HGDP, the correlation coefficient was always close to 1 in all comparisons (Supplementary Figure S4A and S4C). On the other hand, it is possible to observe an increase in correlation as the number of markers in the panel increases for groups such as East Asian (Supplementary Figure S4B) and Native American (Supplementary Figure S4D). As for the 1KGP, EUR and EAS have higher correlation values (Supplementary Figure S4F and S4G) and AFR slightly lower correlation values, especially in comparisons involving the 12AISNP (Supplementary Figure S4E). Since the degree of correlation in marker sets may differ between populations and between data sets, this may be related not only to the number of markers, but also to parental groups used during the AIMS panels design process.

To assess the influence of parental population sets on ancestry inferences, we estimated the AISNP allele frequencies for each continental group in the HGDP and 1KGP dataset (Supplementary Table S6) and performed Fisher's exact test. For most markers, no significant differences were observed in the comparison of allelic frequencies of continental groups between two datasets. However, there are 13 SNPs with significant allele frequency differences (p-value _bonferroni = 0.00007): 5 between African groups; 3 among Europeans and 6 among East Asians (Table 1), most of these SNPs belong to the 446AISNP panel (8/13).

Table 1

List of markers with significantly different allelic frequencies between the parental groups of the HGDP and 1KGP dataset.
CHR	SNP	PANEL	Allele	Freq HGDP	Freq 1KGP	p-value
AFRICA
1	rs2814778	55AISP/128AISP	T	0.068	0.001	1.44E-10
2	rs2627037	446AISNP	A	0.663	0.453	3.33E-08
3	rs7630522	446AISNP	C	0.500	0.328	3.77E-06
15	rs746655	446AISNP	G	0.356	0.198	2.64E-06
16	rs8048832	128AISNP	T	0.194	0.350	8.98E-06
EUROPE
2	rs182549	446AISNP	T	0.345	0.509	4.89E-07
2	rs6754311	34AISNP	T	0.339	0.511	1.19E-07
6	rs2065664	128AISNP	T	0.062	0.368	6.13E-13
EAST ASIAN
4	rs1229984	446AISNP	C	0.482	0.303	7.96E-11
6	rs2065664	128AISNP	T	0.694	0.166	9.73E-12
12	rs59700332	34AISNP	G	0.042	0.002	1.27E-06
15	rs1800414	446AISNP	T	0.534	0.404	5.51E-06
15	rs1426654	446AISNP	A	0.058	0.012	2.06E-06
17	rs17642714	446AISNP	T	0.027	0.001	5.66E-06

Ancestry inference in admixed populations

To evaluate the ancestry inference profile in the American admixed populations, we used a tetrahybrid admixture model. As the HGDP is the only one that contains Native American samples, ancestry inferences in the admixed samples were performed with this dataset as the parental population. Based on our results (Supplementary Table S3-S5), we included as parental samples only AFR, EUR, EAS and NAM samples with z-score values<|3|.

As previously reported [26, 40], the admixed populations evaluated here have different proportions of parental ancestry, ranging from those with proportions of mostly African ancestry (ACB and AWS), mostly European ancestry (CLM, PUR, and SABE), and mostly Native American ancestry (MXL and PEL) (Table S7).

The pairwise comparison of individual ancestry inferences between the panels shows variation in the correlation coefficients, both between ancestry components in the admixed population and between the admixed populations. In Afro-Caribbeans the correlation coefficient for African ancestry component ranges from r² = 0.265 to 1 (Supplementary Figure S5A) and in African Americans from r² = 0.71 to 1 (Supplementary Figure S6A). In Colombians the correlation coefficient for European ancestry component ranges from r² = 0.52-1 (Supplementary Figure S7C), in Puerto Ricans from r² = 0.46–0.99 (Supplementary Figure S10C) and in Brazilians from r² = 0.70–0.99 (Supplementary Figure S11C). As for Mexicans and Peruvians, the correlation coefficient for Native American ancestry showed the greatest variation of r² = 0.15-1 (Supplementary Figure S8D e S9D). In general, the minority ancestry components in the admixed populations are the ones with the lowest correlation coefficient between the panels (e.g. Supplementary Figure S5C; S8A; S11B). Native American ancestry showed the worst correspondence between the panel estimates (e.g. Supplementary Figures S7D, S10D, S11D). Panels with fewer markers tend to have more divergent estimates (e.g. 12AISNP and 34AISNP) on the other hand, panels with more numerous markers (e.g. 446AISNP and 672AISNP) were the ones that showed greater correspondence with the HDSNP and WGS.

As a consequence of the variation in the estimates of individual ancestry between the panels, in the dispersion analyses, greater variation is observed in the average population ancestry (Fig. 3, Supplementary Table S7)

Finally, we evaluated the effect of using sets of different parental populations, especially in ancestry inferences for high-throughput data. Therefore, we performed the ancestry inference using as parental populations: (i) HGDP samples; (ii) 1KGP samples and (iii) HGDP + 1KGP samples. In order to evaluate the tetrahybrid admixture model, given that there are no Native American populations in the 1KGP, we selected 15 Peruvian samples (PEL), which had more than 95% Native American ancestry in our previous analyses.

Our results reveal high correspondence between the individual inferences of the 3 parental datasets (correlation coefficient 0.99-1). Showing that for high-density data, these two datasets have similar accuracy in attributing the continental ancestry of American admixed individuals.

In the present study, we evaluated 10 panel sets: eight AISPN, one HDSNP and one WGS and, using a tetrahybrid admixture model, we compared ancestry inferences in America admixed populations.

To verify the accuracy of the panels, we used samples from HGDP and 1KGP, whose geographic origin is known and without evidence of recent admixture. Despite the low marker overlap observed in the AISNPs panels, all panels showed a high rate of accuracy (error rate 0.1-3.0%; Supplementary Table S3). High levels of correlation were also observed in the pairwise comparisons of the panels (r² > 0.84; Supplementary Figures S4A-G). However, it is also possible to observe an increase in accuracy and correspondence in ancestry inferences, as the number of SNPs in the panels is higher.

These results reveal, in most cases, that the available AISNP panels meet the proposed role of correctly attributing ancestry according to the continental group to which the individual belongs. Several studies already compared the accuracy of panels and have reached similar results [19, 22, 24]. Therefore, currently many authors argue that there is no necessity for new AIMs panels to assign the 6 biogeographic regions (Sub-Saharan Africa, Europe, Southwest Asia, South Asia, East Asia and the Americas). Instead, efforts should be directed towards building panels for global use and with greater representation of population groups [22].

In this sense, many comparative studies are focused on minimizing marker redundancy (reducing costs and enabling reproducibility by different groups), selecting only the most informative and filling gaps in population representation [41]. In the same manner, there is a need to improve the selection of informative markers capable of revealing differences in subregions within continents or in regions with complex admixture patterns, such as populations from East, Southeast and South Asia, as well as those between Middle Eastern and European individuals [42].

Most AIMs panels use HGDP and 1KGP data as a reference population for the selection of their markers, including some evaluated in the present study [17, 43, 44]. These two public databases were essential for understanding the distribution of genetic diversity and affinity among human population groups [27, 45, 46]. However, they capture only a portion of human population diversity. Therefore, many AIMs panels, during their development process, endeavoured to include more populations for different population groups (e.g. 55 AISNP [22]; 128 AISNP [47]; 446 AISNP [48]).

Soundararajan et al [49] argues if there is low representativeness of data from reference populations, a greater number of markers becomes necessary for the robustness of allele frequencies for the definition of population groups of interest. Our results converge to this point, since we observed greater correspondence in individual ancestry inferences between panels with a greater number of markers, in addition to those of HDSNP and WGS data.

In the present study, we propose to focus on the American admixed populations. These populations emerged in the last half-century, especially from Native American, European, African sources. More recently, they have also received contributions from other regions such as East Asia and the Middle East.

Admixed populations need a closer look at ancestry inferences, as their genomic particularities trigger several challenges. Each admixture population has a peculiar evolutionary history, differing in parental sources, proportion and time of admixture. Furthermore, the admixture process produces variation at different levels: (i) in ancestry between admixed populations, (ii) between individuals in the same admixed population and (iii) throughout the genome of the same admixed individual [50].

For this reason, a method, model or panel that well captures the profile in one admixed population or admixed individual will hardly have the same performance for another. Our results showed this heterogeneity, especially in paired comparisons of individual ancestry inference between panels (Supplementary Figures S5-S11), where we observed variation in correlation coefficients both between ancestry components within the same admixed population and between admixed populations.

The inconsistencies observed in the ancestry inferences between the panels were even more evident for the minority ancestry components of the individuals in our results (e.g. Supplementary Figure S5C; S8A; S11B). This probably occurs because the genome of an admixed individual is a mosaic composed of segments from different parental sources. Over generations, due to the process of meiotic recombination, the components of distinct ancestry are shuffled between homologous chromosomes [11, 51]. Thus, the greater the number of generations since the admixture beginning, the smaller the size of the genomic segments of the ancestry will be. As well, the greater the proportion of an ancestry component, the greater the size of its segments in the genome, while on the other hand, the smaller the proportion of the ancestry component, the smaller the segments in the genome⁵¹. In this scenario, due to the lower density and genomic coverage, AISNP with few markers tend to infer more accurately the components of the majority ancestry than the minority ones of an admixed individual. Nevertheless, this problem should not be observed with data with higher SNP density and higher genomic coverage.

We also observed the Native American component as the one with the lowest correspondence in the ancestry inferences between the panels (Supplementary Figures S7D, S10D, S11D). Three factors likely explain these results: (i) two of the panels we used (12AISNP and 34AISNP) were not developed to capture Native American ancestry, which explains their low performance for this component. (ii) the Native American populations, due to their recent bottleneck history, are the most differentiated in the world [45] and the ones with the lowest number of representatives in the reference panels. As much as there are panels that were developed with the aim of enriching the Native American component (e.g. [22, 47, 48]), they do not always capture this component well in all admixed populations; (iii) the Native American component is the minority in most populations evaluated here (with the exception of MXL and PEL).

Here we focused on evaluating the tetrahybrid admixture model in American populations (NAM, EUR, AFR and EAS). We know that for most of the admixed populations assessed here, the East Asian contribution is less than 1%. However, there is a growing migratory flow of this population group to large urban centers in the USA and Brazil. East Asian immigration to Brazil began in 1908 with the Japanese and today, according to the Ministry of Foreign Affairs of Japan, more than 2 million descendants of Japanese people live in Brazil. In São Paulo, the city where the Brazilian samples of the present study were collected, there is one of the largest Japanese communities outside Japan. The Brazilian cohort has 33 samples 100% EAS that are direct descendants of the first Japanese immigrants [26]. Data from the 2010 Brazilian census reveal that in 10 years there was a 173.7% increase in individuals who declared themselves to be of Asian descent (Japanese, Chinese and Korean) [37]. Peru has the second largest ethnic Japanese population in South America after Brazil, and we observed in the 1KGP PEL samples at least two individuals, one with ~ 50% and the other with ~ 12% EAS ancestry. This scenario triggers, even with an average minority population ancestry, at the individual level, East Asian ancestry is majority (100%, 50%), as it represents recent admixture events. Therefore, it is recommended to adjust the admixture models according to the source number of each admixed population, even for minority sources, especially in cases of recent migratory and admixture events.

Although we did not perform the trihybrid model, we compared our results from admixed samples in 1KGP for 55AISNP, 128AISNP and 170AISNP with those from Pereira et al [38] (Supplementary Table S8). We found a variation in the differences between mean ancestry values from 0.13–5.72%. The biggest differences occurred in the inferences of European ancestry for the 128 AISNP in ACB (2.2%); ASW (3.6%); CLM (5.1%), MXL (5.7%) and PUR (4.9%) for Native American ancestry in PEL (3.6%). The directionality was that the tetrahybrid model underestimated that ancestry in relation to the trihybrid model. We are not suggesting that the tetrahybrid model should be universally used in the admixed populations of America. Again, it is necessary to know the history of populations to choose the most suitable model. In this sense, we are encouraging studies to use the tetrahybrid model even when the fourth component is a minority in the populations.

Finally, in the analyses with admixed populations, we found correlation coefficients greater than 0.9 between 446AISNP, 672AISNP, HDSNP and WGS data. This result shows that despite differences in genotyping methods and in the number of markers (hundreds to thousands), there are reliable panels for identifying the four components of ancestry in admixed populations. The choice of one or the other will depend on the purpose and needs of the study. For example, in forensic genetics, sometimes samples with quantity and quality are not available, which limits the genotyping methodology [52]. On the other hand, in clinical or genetic association studies, the accuracy of genomic ancestry is essential, and it is often necessary to go a step beyond the genomic mean and make inferences about ancestry in specific genomic segments [53, 54].

Increasingly, the admixed populations of America are protagonists in different studies (population history, clinical studies, forensics). Therefore, nowadays it is essential to discuss and understand how methodological advances, both in genotyping and in analysis, help to improve the inference of genetic ancestry in admixed populations. In the present study, we show that heterogeneity within and between admixed populations still poses methodological challenges. However, we have also shown that there are good panel sets and tools that can help address these challenges.

Datasets

Samples from three datasets were analyzed: (i) Human Genomic Diversity Panel (HGDP) [46]; (ii) Project 1000 genomes phase III (1000 Genome Project – 1KGP) [27], and (iii) Brazilian Cohort of Health, Well-being and Aging (Saúde e Bem Estar – SABE) [26].

Based on the American admixture history, analyses were performed with parental samples from African, European, East Asian and Native American populations of HGDP (543 individuals) and 1KGP (1511 individuals) as described in table S1. We also analyzed the 504 samples from the 6 admixed populations from the 1KGP and 1,171 Brazilian samples from the SABE cohort (Supplementary Table S1).

All samples were genotyped by WGS and are publicly available (https://www.internationalgenome.org/data - HGDP and 1KGP; https://ega-archive.org/studies/EGAS00001005052 – SABE). All individuals enrolled in the SABE cohort have agreed on participating in this study on written consent forms approved by local and national institutional review boards COEP/FSP/USP OF.COEP/23/10, CONEP 2044/2014, and CEP HIAE 1263-10.

Ancestry Informative Markers SNPs panel (AISNP panel)

Due to availability in the three datasets, we evaluated only SNPs as AIM. Based on this criterion, we selected 7 AIMs panels frequently used in studies with Latin American populations: 12 AISNP [43]; 34 AISNP [44]; PIMA [17]; 55 AISNP [22]; 128 AISNP [47]; 170 AISNP [55]; 446 AISNP [48]. In addition, we also evaluated the combination of the 7 panels, which we call 672 AISNP. The SNPs of the AIMS panels used in the present study are described in Supplementary Table S2.

High-density SNP chip array (HDSNPs panel)

As a representative of high-density SNP arrays, we chose the Axiom^™ Genome-Wide Human Origins (~ 600K SNPs – ThermoFisher Scientific). This genotyping panel was optimized for population geneticists studies and developed from genomic markers identified in 11 human populations (France, China, Papua New Guinea, San, Yoruba, Mbuti pygmies, Karitiana, Italy-Sardinia, Melanesia, Cambodia, Mongolia), avoiding confounding biases introduced using GWAS SNP arrays.

Merge Datasets

Based on the WGS data from the three datasets, the following sets of SNPs were selected: (i) AISNP panels: 672 SNPs make up the 7 AISNP panels selected for the present study. Of these, 5 SNPs (rs12402499; rs17287498; rs1321333; rs10954737; rs10071261) are not detected in all datasets (Supplementary Table S9), of which, three SNPs are informative of Native American ancestry and two of African ancestry. (ii) HDSNPs panel: ~600,000 SNPs that make up the Axiom Human Origins array. The overlap between the three datasets was 555,168 SNPs. (iii) WGS data: the original datasets have more than 60 million variants described. For the present study, we exclude SNPS: (a) MAF < 1%, (b) missing data per SNP > 1%, (c) Hardy-Weinberg p-value < 1x10^-8, (d) filter for LD coefficient (r² = 0.1 – see RESULTS section for more details). The final dataset contains 2,018,023 SNPs. For each set of markers, the three datasets (HGDP, 1KGP, SABE) were merged using vcftools v.0.1.15 [56] and plink v.1.9 [57]. To validate these merge data, a PCA analysis was performed (Supplementary Figure S1). Throughout the text we referred to AISNP, HDSNP and WGS as "panel sets".

Data analysis

Allelic frequency inferences, Hardy-Weinberg and Fisher's exact test was performed in plink v.1.9 software [57]. Correction for multiple testing was done according to Bonferroni correction.

Genetic ancestry inference

ADMIXTURE v.1.3 [58] was used to perform global ancestry inference. Analyses were performed in an unsupervised manner, when considering only parental populations, and in a supervised manner when considering admixed populations as well. As parameters we used 4 clusters (K = 4) and 2000 bootstrap replicates. Each analysis was repeated 10 runs and the results were combined using the CLUMPP v.1.222 software [59].

Information redundancy may occur in high-density SNP data (HDSNP and WGS). Therefore, to minimize and evaluate background linkage disequilibrium in the analyses, we also tested some disequilibrium linkage coefficients (r² = 0.01, 0.05, 0.1, 0.3 and 0.5), assuming distance not closer than 200Kb between adjacent markers [57, 58].

Comparisons of ancestry inferences were performed by correlation analysis and z-score test by R scripts with package stats v.4.1.1 [60].

Acknowledgements

This study was supported by Coordenação de Aperfeiçoamento de Pessoal de Nível Superior – CAPES; National Council for Scientific and Technological Development – CNPq (PhD scholarship - L.M.E.S.).

Author Contributions

S.O and K.N contributed to the conception and design of the study; M.S.N., M.O.S., Y.A.O.D. and M.Z provided SABE cohort samples and genomic data; L.M.E.S and K.N. performed the analyses; L.M.E.S, K.N and S.O wrote the article and incorporated inputs from other authors.

Competing interests

The author(s) declare no competing interests.

Data Availability

The datasets reported in this article are publicly available (https://www.internationalgenome.org/data- HGDP and 1KGP; European Genome-phenome Archive (EGA), under EGA Study accession number EGAS00001005052 – SABE).

Rohlfs, R. V., Fullerton, S. M. & Weir, B. S. Familial Identification: Population Structure and Relationship Distinguishability. PLoS Genet 8(2): e1002469; 10.1371/journal.pgen.1002469 (2012).
Garrison, N., Rohlfs, R. & Fullerton, S. Forensic familial searching: scientific and social implications. Nat Rev Genet. 14, 445; 10.1038/nrg3519 (2013).
Romero-Hidalgo, S. et al. Demographic history and biologically relevant genetic variation of Native Mexicans inferred from whole-genome sequencing. Nat Commun. 8, 1005; 10.1038/s41467-017-01194-z (2017).
Lewontin, R. C. The apportionment of human diversity. In Evolutionary biology (eds Steere W.C., Dobzhansky T., Hecht MK), 381–398 (Springer, 1972).
Bradbury, C. et al. Off-target phenotypes in forensic DNA phenotyping and biogeographic ancestry inference: A resource. Forensic science international. Genetics. 38, 93–104; 10.1016/j.fsigen.2018.10.010 (2019).
Reidy, K. J. et al. Genetic risk of APOL1 and kidney disease in children and young adults of African ancestry. Current opinion in pediatrics. 30, 252–259; 10.1097/MOP.0000000000000603 (2018)
Olczak, K. J. et al. Hypertension genetics past, present and future applications. Journal of internal medicine. 290, 1130–1152; 10.1111/joim.13352 (2021).
Mak, W. Y. et al. The epidemiology of inflammatory bowel disease: East meets west. Journal of gastroenterology and hepatology. 35, 380–389; 10.1111/jgh.14872 (2020).
Tatonetti, N. P. & Noémie E. Fine-scale genetic ancestry as a potential new tool for precision medicine. Nature medicine. 27, 1152–1153; 10.1038/s41591-021-01405-7 (2021).
Hellwege, J. N. et al. Population Stratification in Genetic Association Studies. Current protocols in human genetics. 95, 1.22.1–1.22.23; 10.1002/cphg.48 (2017).
Suarez-Pajes, E. et al. Genetic Ancestry Inference and Its Application for the Genetic Mapping of Human Diseases. International journal of molecular sciences. 22, 6962; 10.3390/ijms22136962 (2021).
Gouveia M.H. et al. Origins, Admixture Dynamics, and Homogenization of the African Gene Pool in the Americas. Mol Biol Evol. 1,1647–1656; 10.1093/molbev/msaa033 (2020).
Ongaro L., et al. Continental-scale genomic analysis suggests shared post-admixture adaptation in the Americas. Hum Mol Genet. 1, 2123–2134; 10.1093/hmg/ddab177 (2021).
Rosenberg, N. A. et al. Informativeness of genetic markers for inference of ancestry. American journal of human genetics. 73, 1402–22; 10.1086/380416 (2003).
Shriver, M. D. et al. Skin pigmentation, biogeographical ancestry and admixture mapping. Human genetics. 112, 387–99; 10.1007/s00439-002-0896-y (2003).
Phillips, C., Santos, C., Fondevila, M., Carracedo, Á. & Lareu, M. V. Inference of Ancestry in Forensic Analysis I: Autosomal Ancestry-Informative Marker Sets. Methods in molecular biology. 1420, 233–253; 10.1007/978-1-4939-3597-0_18 (2016).
Gontijo, C.C., et al. PIMA: A population informative multiplex for the Americas. Forensic science international. Genetics. 44, 102200; 10.1016/j.fsigen.2019.102200 (2020).
Santos, C., Phillips, C., Gomez-Tato, A., Alvarez-Dios, J., Carracedo, Á., & Lareu, M. V. Inference of Ancestry in Forensic Analysis II: Analysis of Genetic Data. Methods in molecular biology. 1420, 255–285; 10.1007/978-1-4939-3597-0_19 (2016).
Santos, H. C., et al. A minimum set of ancestry informative markers for determining admixture proportions in a mixed American population: the Brazilian set. European journal of human genetics: EJHG. 24, 725–731; 10.1038/ejhg.2015.187 (2016).
Phillips, C. et al. Inferring ancestral origin using a single multiplex assay of ancestry-informative marker SNPs. For. Sci. Int. Genet. 1, 273–280; 10.1016/j.fsigen.2007.06.008 (2007).
Pereira, R., et al. Straightforward inference of ancestry and admixture proportions through ancestry-informative insertion deletion multiplexing. PloS one. 7, e29684; 10.1371/journal.pone.0029684 (2012).
Kidd, K. K., et al. Progress toward an efficient panel of SNPs for ancestry inference. Forensic science international. Genetics. 10, 23–32; 10.1016/j.fsigen.2014.01.002 (2014).
Phillips C. Forensic genetic analysis of bio-geographical ancestry. Forensic science international. Genetics. 18, 49–65; 10.1016/j.fsigen.2015.05.012 (2015).
Guo, Y. X., Jin, X. Y., Xia, Z. Y., Chen, C., Cui, W., & Zhu, B. F. A small NGS-SNP panel of ancestry inference designed to distinguish African, European, East, and South Asian populations. Electrophoresis, 41(9), 649–656; 10.1002/elps.201900231 (2020).
Kehdy, F. S., et al. Origin and dynamics of admixture in Brazilians and its effect on the pattern of deleterious mutations. Proceedings of the National Academy of Sciences of the United States of America. 112, 8696–8701; 10.1073/pnas.1504447112 (2015).
Naslavsky, M. S., et al. Whole-genome sequencing of 1,171 elderly admixed individuals from São Paulo, Brazil. Nature communications. 13, 1004; 10.1038/s41467-022-28648-3 (2022).
The 1000 Genomes Project Consortium, A global reference for human genetic variation, Nature. 526, 68–74; 10.1038/nature15393 (2015).
Gurdasani, D., et al. The African Genome Variation Project shapes medical genetics in Africa. Nature. 517(7534), 327–332; 10.1038/nature13997 (2015).
Kelleher, J., et al. Inferring whole-genome histories in large population datasets. Nature genetics. 51(9), 1330–1338; 10.1038/s41588-019-0483-y (2019).
Lawson, D. J., Hellenthal, G., Myers, S., & Falush, D. Inference of population structure using dense haplotype data. PLoS genetics. 8(1), e1002453; 10.1371/journal.pgen.1002453 (2012).
Hellenthal, G., et al. A genetic atlas of human admixture history. Science. 343(6172), 747–751; 10.1126/science.1243518 (2014).
Liu, C. C., Shringarpure, S., Lange, K., & Novembre, J. Exploring Population Structure with Admixture Models and Principal Component Analysis. Methods in molecular biology. 2090, 67–86; 10.1007/978-1-0716-0199-0_4 (2020).
Wu, J., Liu, Y., & Zhao, Y. Systematic Review on Local Ancestor Inference From a Mathematical and Algorithmic Perspective. Frontiers in genetics, 12, 639877; 10.3389/fgene.2021.639877 (2021).
He, G., et al. Massively parallel sequencing of 165 ancestry-informative SNPs and forensic biogeographical ancestry inference in three southern Chinese Sinitic/Tai-Kadai populations. Forensic science international. Genetics. 52, 102475; 10.1016/j.fsigen.2021.102475 (2021).
Adhikari, K., Mendoza-Revilla, J., Chacón-Duque, J. C., Fuentes-Guajardo, M., & Ruiz-Linares, A. Admixture in Latin America. Current opinion in genetics & development. 41, 106–114; 10.1016/j.gde.2016.09.003 (2016).
Norris, E. T., et al. Genetic ancestry, admixture and health determinants in Latin America. BMC genomics. 19(Suppl 8), 861; 10.1186/s12864-018-5195-7 (2018).
Centro de Documentação e Disseminação de Informações (Brazil). Brazil, 500 years of settlement, (ed IBGE- Instituto Brasileiro de Geografia e Estatística) 197-213p. (Rio de Janeiro, 2007).
Pereira, V., et al. Evaluation of the Precision of Ancestry Inferences in South American Admixed Populations. Frontiers in genetics. 11, 966; 10.3389/fgene.2020.00966 (2020).
Debortoli, G., de Araujo, G. S., Fortes-Lima, C., Parra, E. J., & Suarez-Kurtz, G. Identification of ancestry proportions in admixed groups across the Americas using clinical pharmacogenomic SNP panels. Scientific reports. 11(1), 1007; 10.1038/s41598-020-80389-9 (2021).
Martin, A. R., et al. Human Demographic History Impacts Genetic Risk Prediction across Diverse Populations. American journal of human genetics. 100(4), 635–649; 10.1016/j.ajhg.2017.03.004 (2017).
Jin, X. Y., et al. Ancestry Prediction Comparisons of Different AISNPs for Five Continental Populations and Population Structure Dissection of the Xinjiang Hui Group via a Self-Developed Panel. Genes, 11(5), 505; 10.3390/genes11050505 (2020).
Phillips, C., et al. MAPlex - A massively parallel sequencing ancestry analysis multiplex for Asia-Pacific populations. Forensic science international. Genetics. 42, 213–226; 10.1016/j.fsigen.2019.06.022 (2019).
Pfaffelhuber, P., Grundner-Culemann, F., Lipphardt, V., & Baumdicker, F. How to choose sets of ancestry informative markers: A supervised feature selection approach. Forensic science international. Genetics. 46, 102259; 10.1016/j.fsigen.2020.102259 (2020).
Phillips, C., Fondevila, M., & Lareau, M. V. A 34-plex autosomal SNP single base extension assay for ancestry investigations. Methods in molecular biology. 830, 109–126; 10.1007/978-1-61779-461-2_8 (2012).
Rosenberg, N. A., et al. Genetic structure of human populations. Science. 298(5602), 2381–2385; 10.1126/science.1078311 (2002).
Bergström, A., et al. Insights into human genetic variation and population history from 929 diverse genomes. Science. 367(6484), eaay5012; 10.1126/science.aay5012 (2020).
Kosoy, R., et al. Ancestry informative marker sets for determining continental origin and admixture proportions in common populations in America. Human mutation. 30(1), 69–78; 10.1002/humu.20822 (2009).
Galanter, J. M., et al. Development of a panel of genome-wide ancestry informative markers to study admixture throughout the Americas. PLoS genetics. 8(3), e1002554; 10.1371/journal.pgen.1002554 (2012).
Soundararajan, U., Yun, L., Shi, M., & Kidd, K. K. Minimal SNP overlap among multiple panels of ancestry informative markers argues for more international collaboration. Forensic science international. Genetics. 23, 25–32; 10.1016/j.fsigen.2016.01.013 (2016).
Korunes, K. L., & Goldberg, A. Human genetic admixture. PLoS genetics. 17(3), e1009374; 10.1371/journal.pgen.1009374 (2021).
Tang, H., et al. Recent genetic selection in the ancestral admixture of Puerto Ricans. American journal of human genetics. 81(3), 626–633; 10.1086/520769 (2007).
Jordan, D. & Mills, D. Past, Present, and Future of DNA Typing for Analyzing Human and Non-Human Forensic Samples. Front. Ecol. Evol. 9,646130; 10.3389/fevo.2021.646130 (2021).
Blue, E. E., Horimoto, A., Mukherjee, S., Wijsman, E. M., & Thornton, T. A. Local ancestry at APOE modifies Alzheimer's disease risk in Caribbean Hispanics. Alzheimer's & dementia: the journal of the Alzheimer's Association. 15(12), 1524–1532; 10.1016/j.jalz.2019.07.016 (2019).
Horimoto, A., et al. Genome-Wide Admixture Mapping of Estimated Glomerular Filtration Rate and Chronic Kidney Disease Identifies European and African Ancestry-of-Origin Loci in Hispanic and Latino Individuals in the United States. Journal of the American Society of Nephrology: JASN. 33(1), 77–87; 10.1681/ASN.2021050617 (2022).
Pakstis, A. J., et al. Population relationships based on 170 ancestry SNPs from the combined Kidd and Seldin panels. Scientific reports. 9(1), 18874; 10.1038/s41598-019-55175-x (2019).
Danecek P., et al. The variant call format and VCFtools. Bioinformatics. 27(15): 2156–2158; 10.1093/bioinformatics/btr330 (2011).
Chang, C. C., et al. Second-generation PLINK: rising to the challenge of larger and richer datasets. GigaScience. 4, 7; 10.1186/s13742-015-0047-8 (2015).
Alexander, D. H., Novembre, J., & Lange, K. Fast model-based estimation of ancestry in unrelated individuals. Genome research. 19(9), 1655–1664; 10.1101/gr.094052.109 (2009).
Jakobsson, M., and Rosenberg, N. A. CLUMPP: a cluster matching and permutation program for dealing with label switching and multimodality in analysis of population structure. Bioinformatics. 23, 1801–1806; 10.1093/bioinformatics/btm233 (2007).
R Core Team (2021). R: language and environment for statistical computing. R Foundation for Statistical Computing. Viena. Austria. URL: https://www.R-project.org/.

No competing interests reported.

Download PDF

Editorial decision: Major revision
26 Jul, 2022
Reviews received at journal
20 Jul, 2022
Reviewers agreed at journal
11 Jul, 2022
Reviewers invited by journal
10 Jul, 2022
Editor assigned by journal
10 Jul, 2022
Editor invited by journal
10 Jul, 2022
Submission checks completed at journal
10 Jul, 2022
First submitted to journal
04 Jul, 2022

You are reading this latest preprint version

Genetic Ancestry Inference: from AIMS to WGS application and challenges in admixed population

Status:

Version 1

Abstract

Figures

Introduction

Results

SNP overlap between AISNP panels

Linkage disequilibrium

Ancestry Inference in Parental Populations Groups

Ancestry inference in admixed populations

Discussion

Materials And Methods

Datasets

Ancestry Informative Markers SNPs panel (AISNP panel)

High-density SNP chip array (HDSNPs panel)

Merge Datasets

Data analysis

Genetic ancestry inference

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1