Volunteers were recruited to the study if they complied with the following conditions: (1) self-reported healthy condition; (2) no biological kinship related to anteriorly recruited participants within at least three generations; (3) no immigration or inter-marriage events in their family histories. In total, 257 blood samples were collected from two ethnic minorities (Mongolian and Ewenki) of IMAR, China. The sample sizes were 157 Mongolian and 100 Ewenki volunteers, respectively. Blood samples were stored at –20 ${}^{\circ }$C before DNA extraction using the paramagnetic particle method. The concentration of DNA samples was estimated by measuring their absorbance at 260 nm using a NanoDrop ND-1000 Spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA) according to the manufacturer’s recommendation. DNA samples were then diluted to 1 ng/$\mu$L. Written informed consents were obtained from all participants before the study began. The study received approvals from the Ethics Committees of Southern Medical University, Guangzhou, China and Xi’an Jiao Tong University, Xi’an, China (No. XJTULAC201).

PCR amplification of the overall loci was carried out in a single PCR system using the reagents and reaction conditions as described in the manufacturer’s protocol for the AGCU InDel 50 kit (AGCU, Wuxi, China) and performed in a GeneAmp PCR 9700 Thermal Cycler (Applied Biosystem, Foster City, CA, USA). The 25 $\mu$L reaction volume for the multiplex amplification system included 10 $\mu$L of reaction mix, 5 $\mu$L of primer mix, 1 $\mu$L of heat-activated C-Taq DNA polymerase, 1 ng of genomic DNA, and filled to the total volume with sdH${}_2$O. Thermal cycling conditions were programmed as follows: the initial pre-degeneration step was 95 ${}^{\circ }$C for 2 min, followed by 29 cycles of denaturation at 94 ${}^{\circ }$C for 30 s, annealing at 60 ${}^{\circ }$C for 1 min, extension at 72 ${}^{\circ }$C for 1 min, and a final extension at 60 ${}^{\circ }$C for 30 min. The ABI 3500 xL Genetic Analyzer (Applied Biosystem, Foster City, CA, USA) was used to detect PCR product in a 1 $\mu$L volume added to 0.5 $\mu$L of AGCU Marker SIZ-500 and 12 $\mu$L of deionized formamide. Subsequently, InDel genotyping was performed by GeneMapper ID-X version 1.5 Software (Applied Biosystem, Foster City, CA, USA) with the analytical threshold of peak height at 50 relative fluorescence units (RFU) for allele calling. Positive and negative controls were included to ensure the accuracy of results for InDel genotyping.

Genotype data for the 47 autosomal InDel variations in 26 reference populations from the 1000 Genomes Project Phase 3 were downloaded from the online website Ensemble Genome Browser (http://grch37.ensembl.org/index.html) in order to conduct inter-population genetic analyses. Previously published datasets describing genetic diversities in the 47 polymorphic InDel loci from different geographic regions in China were also integrated. These included data for the 47 InDels from the Hainan Han (HAH), Hainan Li (HNL), Zunyi Gelao (ZGL), Beijing Han (BJH), Henan Han (HNH), Heilongjiang Han (HLJH), Shandong Han (SDH), Shanxi Han (SXH), Tibetan, Chengdu Han (CDH), Yi, and Guangdong Han (GDH) populations [20, 24, 25, 26]. We then integrated the genotype data for CHB from the 1000 Genomes Project Phase 3 and BJH from previous research to create a combined dataset of 256 Han individuals from Beijing city, referred to as BJH in this study. Finally, a combined dataset was generated for the subsequent analyses by combining Mongolian, Ewenki and other individuals from 37 reference populations. Detailed information for the 26 continental reference populations derived from the 1000 Genomes Project Phase 3 is shown in Supplementary Table 1.

Forensic statistical parameters for the 47 InDel variations in the Mongolian and Ewenki ethnic minorities were calculated using the browser-based STRAF application (http://cmpg.unibe.ch/shiny/STRAF/) [27]. These included match probability (MP) [28], discrimination power (DP), probability of exclusion (PE), polymorphism information content (PIC) [29] and observed heterozygosity (Hob). Linkage disequilibrium (LD) analyses of pairwise InDel loci in each group were carried out using SNPAnalyzer version 2.0 (Istech, Goyang, South Korea) Software [30]. An r${}^{\mathrm{2}}$ threshold of 0.8 was used to determine the strong linkage state for pairwise InDel loci. Genepop software (version 4.04, Rousset, France) was used to perform Hardy-Weinberg equilibrium (HWE) tests for all loci in the two ethnic minorities, and to calculate the pairwise F-statistic (F${}_{S\mathit{}T}$ ) index in the 39 populations [31]. The cumulative discrimination power (CDP) and the combined probability of exclusion (CPE) were calculated with Excel by reference to the relevant formulae.

The Dispan program was used to calculate Nei’s D${}_A$ genetic distances (D${}_A$ distances) among the overall populations. Principal component analysis (PCA) was performed at the individual level to reveal the spatial distribution pattern for all samples from the two ethnic minorities and from the reference populations. The efficiency of the 47 InDel loci in assigning these individuals to the corresponding biogeographic regions was not immediately intuitive. We therefore generated different clustering patterns at the population level for five continents (Africa, Europe, America, East Asia and South Asia) and three continents (Africa, Europe and East Asia) using the R program. Genetic differences between the two ethnic groups and the reference populations were more intuitively reflected by simultaneously analyzing the clustering patterns of the overall populations using different PCA plots. The Jensen–Shannon divergence values [32] for the 47 InDel variations were calculated using the built-in algorithm “Thorough analysis of population data of a custom Excel file” in the online Snipper site (http://mathgene.usc.es/snipper/analysispopfile2_new.html) [33]. The Jensen–Shannon divergence values were then converted to the “informativeness-for-assignment” metric (I${}_n$) by referring to the previously reported formula [34]. Arlequin software (version 3.5.1.2, Laurent Excoffier & Heidi Lischer, Switzerland) [35] was used to compute locus-by-locus p values of the pairwise populations within China in order to reveal intra-population differentiations within the same geographic region by the method of molecular variation analysis of variance (AMOVA) [36].

To investigate in more detail the substructure patterns for the two ethnic minorities from IMAR and to infer the proportions of different ancestral genetic components, STRUCTURE software (version 2.3.4, Pritchard & Stephens & Donnelly, United Kingdom) was used to perform unsupervised ancestral component prediction of the populations [37]. The length of burn-in period was 100,000 times followed by 100,000 MCMC repetitions. Initial runs were performed without any prior information about the sample origin and were based on the admixture model and correlated allele frequencies pattern. The number of hypothetical ancestry clusters (K) was set from 2 to 5, with 15 independent runs performed for each of the tested K values. The optimal K was identified using the online web Harvester program (http://taylor0.biology.ucla.edu/structureHarvester/) [38]. The average permutated individual and population Q-matrices for 15 replicates of each K were assessed by CLUMPP software (version 1.1.2, Jakobsson & Rosenberg, USA) [39]. Subsequent plotting (bar plot) was conducted using DISTRUCT software (version 1.1, Jakobsson & Rosenberg, USA) [40] with the input of CLUMPP results. Phylogenetic construction of the two ethnic groups and of the reference populations was generated based on D${}_A$ distances with the neighbor-joining method applied by MEGA software (version 6.06, Tamura, Japan) [41]. Subsequent tree visualization and management were performed with the online tool ITOL [42].

To better visualize the parameter distributions and gain a comprehensive view of the differences, Rstudio software (https://www.rstudio.com/products/rstudio/download/) was used to draw the boxplot of the forensic statistical parameters for the two ethnic minorities, the heat map of the insertion allelic frequencies in five continental populations, and the heat map of F${}_{S\mathit{}T}$ and D${}_A$ values for pairwise populations.

HWE exact tests for the 47 InDel loci in the Mongolian and Ewenki groups were performed to confirm the validity of sample collection. After Bonferroni’s correction, p value of 0.0011 (0.05/47 = 0.0011) was considered to represent deviation from the equilibrium state. All InDel loci were found to comply with HWE, with a minimum p value of 0.008 for the rs142221201 locus in Mongolians, and 0.014 for the rs151335218 locus in the Ewenki group.

Before using the product law in the CDP and CPE calculations, LD tests of pairwise InDel loci were performed to assess the independence of each InDel locus. The results of LD analyses were mirrored by data matrix of inverted triangles. As shown in Supplementary Fig. 1, none of the small block was covered in crimson and no area was encircled by the thick black line with the r${}^{\mathrm{2}}$ threshold established at 0.8. This showed that no LDs existed between any of two different InDel loci in the Mongolian and Ewenki ethnic groups.

Allelic frequencies for the 47 InDel loci in the two ethnic minorities were calculated and the results were presented in Table 1. In the Mongolian group, the insertion allele frequencies ranged from 0.2803 at rs67939200 to 0.7548 at rs3834231, while in the Ewenki group they ranged from 0.2750 at rs67264216 and rs1127697 to 0.8050 at rs3834231. Overall, 89.36% and 82.98% of insertion allele frequencies were within the range of 0.3 to 0.7 in the Mongolian and Ewenki groups, respectively. PD values ranged from 0.5299 at rs3834231 to 0.6494 at rs66739142 in the Mongolian group, and from 0.4778 at rs3834231 to 0.6526 at rs5787309 in the Ewenki group. Hob values ranged from 0.3694 at rs67700747 to 0.5478 at rs5787309 in the Mongolian group, and from 0.2900 at rs3834231 to 0.6300 at rs151335218 in the Ewenki group. PE values ranged from 0.0964 at rs67700747 to 0.2328 at rs5787309 in the Mongolian group, and from 0.0595 at rs3834231 to 0.3284 at rs151335218 in the Ewenki group. Subsequently, boxplots were constructed to show the distribution of forensic parameters and to make comparison in the two ethnic groups. As depicted in Fig. 1, red and green boxes show the general data dispersion of forensic parameters in the two groups. The left side of the figure shows that most parameters (GD, Hobs, PM) were concentrated between 0.4 and 0.5, but not PDs and PEs. The median value of GD, Hobs, PD, PE and PM exceeded 0.4, 0.4, 0.6, 0.15 and 0.4, respectively, in both the Mongolian and Ewenki groups. Larger variations in the GD, Hob, PD, PE and PM values were observed in the Ewenki group. Forensic statistical parameters were then calculated to investigate the efficiencies of the 47 InDel loci for applications in individual identification and parentage testing. The CDP and CPE values were calculated to be 0.999999999999999999874 and 0.99981 respectively in the Mongolian group, and 0.999999999999999999677 and 0.99975 respectively in the Ewenki group. These results strongly support the 47 InDels as suitable tool for personal identification in the two ethnic groups. Supplementary Tables 2 and 3 summarize the detailed information regarding forensic parameters for the 47 InDel loci.

Fig. 1.

Box plot of the forensic statistical parameters (GD, Hobs, PD, PE and PM) in the two ethnic minorities. Vertical lines in the boxplot denote the range, middle lines denote the median, while the bottom and top of each box corresponds to the first and third quartiles, respectively. Individual data are represented by dark red dots.

Gene interactions are prone to occur among different populations living in the same biogeographic region. To examine how historical events shaped the genetic diversities of the two ethnic minorities in IMAR, genotype data for the 47 InDel loci in the overall 39 populations were assembled to determine population structures and genetic affinities. Firstly, a heat map of the insertion allele frequency matrix for the 47 InDel variations was generated to reflect population genetic diversities (Fig. 2). Insertion allele frequencies were represented by different colors, varying from light blue (0) to red (1). On the left and top of the figure, population clusters and locus clusters were generated based on the allele frequency distributions of the 47 InDel loci. In general, populations with similar allele frequency distributions shared the same sub-branch on the left clustering tree, with five clusters easily distinguished. The exceptions were CLM and PUR populations from America positioned in the European cluster. The Mongolian and Ewenki groups clustered with East Asian populations and located on the same sub-branch of the tree. InDel variants that exhibited similar insertion allelic frequencies in different populations were assembled at the top of the heat map and were in the same sub-branch as the other clustering tree.

Fig. 2.

Heat map of insertion allelic frequencies for the 47 InDel variations in the two ethnic groups and 37 reference populations. The color gradient for the insertion allele frequency distribution ranges from light blue to red. At the top and left sides of the heat map, clustering trees are generated based on the corresponding insertion allele frequency distributions in different populations and at different loci, respectively.

The heat map revealed relatively obvious discrepancies in allele frequencies between different populations. As proposed by Shriver [43, 44], the genetic distance between populations for any single molecular marker can be estimated from the $\delta$ metric, known as the allele frequency differential value. We therefore speculated that some of the InDel variations could dissect the population structures to some extent. We also calculated the I${}_n$ value for each InDel variation in different populations from three continents (Africa, Europe, East Asia), four continents (Africa, Europe, East Asia and South Asia) and five continents (Africa, Europe, East Asia, South Asia and America) with Snipper-based PCA. Table 1 showed I${}_n$ values for the 47 InDels ranking in descending order. The seven most informative InDel loci with I${}_n$ values $>$0.1 were extracted and their insertion allelic frequency distributions in the five populations from three continents were shown in Fig. 3. The ESN from Africa, FIN from Europe and CHB from East Asia were considered here to be representative of their continent’s populations. The seven most informative InDel variations were the rs72085595, rs34287950, rs71852971 rs66477007, rs79225518, rs5897566 and rs538690481 loci. As shown in Fig. 3, differences in insertion allele frequencies were observed between the continental populations, with the African populations showing significantly different allelic frequency distributions compared to non-African populations. In contrast, similar insertion allelic frequency distributions were observed amongst populations from the same continent (CHB, Mongolian and Ewenki).

Fig. 3.

The seven most informative InDel variations (I${}_{\mathrm{n}}$ $\mathrm{>}$0.1) for the differentiation of African, European and East Asian populations are shown. Insertion allele frequencies for the InDels in the Mongolian and Ewenki groups and for the reference populations (ESN, African; FIN, European; CHB, East Asian) are shown using different colors in the bar plot. The insertion allele frequencies for the InDel variations are shown below the figure.

Next, PCA at the individual level was performed to reveal the sample distributions from Mongolian, Ewenki and reference populations. As shown in Fig. 4A, individuals from the two ethnic groups and the reference populations from five continents (Africa, Europe, East Asia, South Asia and America) were represented by dots labeled with different colors. PCA showed that individuals were scattered in the middle of the plot, with the first two PCs accounting for 12.5% of the total variance. Population distinctions at the five continent level were not clearly obvious in this analysis. We next drew the PCA plot at the four continent population level (Africa, Europe, East Asia and South Asia) to reveal the distribution of recruited individuals, with the two topmost PCs making up 12.71% of the total variance (Fig. 4B). Previous studies reported that each PC captured only a portion of the total variability, but here the first PC captured the largest portion followed by the second PC. The combined PCs define a sample’s eigenvector, and usually the 2D PC1–PC2 plots contain the most information in the simplest space [45, 46]. Based on the PCA plots generated from clustering patterns of different continental populations, we concluded the genetic structures of populations from five continents and four continents could not be ideally dissected by the 47 InDel variations. However, the tendency indicated that populations from Africa, Europe and East Asia could be distinguished by the set of 47 InDel loci. Hence, we next generated the clustering pattern of individuals from the two ethnic study groups and the reference populations from Africa, Europe and East Asia. This allowed us to evaluate the potential of the 47 InDels in correctly assigning individuals to their corresponding biogeographic clusters (Fig. 4C). Compared to the results shown in Fig. 4B, more distinct boundaries were seen in Fig. 4C between the African, European and East Asian populations. Populations from the same continents assembled together to form relatively distinguishable clusters. The two ethnic minorities from IMAR assembled with the East Asian populations, thus indicating their close genetic affinities. An additional PCA plot (Fig. 4D) was generated to confirm the efficiency of the seven most informative InDel variations (I${}_{n\mathit{}\mathrm{3}}$ $>$0.1) with regard to ancestry inference amongst populations from Africa, Europe and East Asia. The seven InDel variations produced an analogous clustering pattern to the PCA plot generated by the 47 InDels, with several tightly assembled clusters seen within the African populations.

Fig. 4.

PCA plots based on the two ethnic groups and the corresponding reference populations. (A) PCA of the two ethnic groups and reference populations from five continents (Africa, Europe, America, South Asia and East Asia). (B) PCA of the two ethnic groups and reference populations from four continents (Africa, Europe, South Asia and East Asia). (C) PCA of the two ethnic groups and reference populations from three continents (Africa, Europe and East Asia). (D) PCA of the two ethnic groups and reference populations from three continents (Africa, Europe and East Asia) based upon the seven most informative InDels with I${}_{n\mathit{}\mathrm{3}}$ $>$0.1.

Based on the maximum-likelihood algorithm, STRUCTURE software [47] was used to further characterize the substructure patterns of the two ethnic minorities and to infer the proportion of ancestral components using the reference continental populations. Using the online Harvest program (http://taylor0.biology.ucla.edu/structureHarvester/) [38], the optimal K was determined to be 3 (Supplementary Fig. 2). At the best-fit model of K = 3 (Fig. 5A), highly specific genetic components were observed in populations from the same continent. Accordingly, the genetic ancestries of African, European and East Asian populations are represented by green, pink and blue-green lines, respectively. A large fraction of East Asian ancestral genetic component was detected in the Mongolian and Ewenki groups. We also investigated ancestry component discrepancies for the overall populations when the genetic ancestry was pre-assumed to be 2–5. No extra discrepancies in ancestry component compositions were detected among the studied and reference populations with the increased K. Thus, together with the population genetic patterns inferred from K = 3, the present results corroborated the unambiguous differentiation of African, European and East Asian individuals by STRUCTURE genetic ancestry prediction analyses using these 47 InDels. The triangle plot (Fig. 5B) also revealed that individuals from the same continent assembled into independent clusters and reflected the pattern of genetic similarity of the studied and reference populations. Furthermore, the bar plot was used to represent the estimated ancestry composition of the populations by STRUCTURE (Fig. 5C), with the results showing that Mongolian and Ewenki groups possess a large fraction of East Asian ancestry. The estimated East Asian ancestry components for the Mongolian and Ewenki groups were 88.6% and 88.2%, respectively. However, the average East Asian ancestry component of the Chinese populations investigated in this study was 92.8%, indicating slightly less East Asian ancestry in the Mongolians and Ewenkis compared to other Chinese populations.

Fig. 5.

STRUCTURE analyses of the two ethnic groups and the corresponding reference populations. (A) Bar plot of the estimated genetic ancestral components at the individual level in Mongolian, Ewenki and reference populations from Africa, Europe and East Asia. Individuals are represented by a vertical line divided into K colored segments, with the length of each segment being proportional to the estimated membership. (B) Triangle plot of Mongolian and Ewenki groups and the reference populations from three continents. (C) The bar plot represents the corresponding ancestry composition of reference populations from Africa, Europe and East Asia. The African ancestry is labeled in green, the European ancestry in pink, and the East Asian ancestry in blue-green.

F${}_{S\mathit{}T}$ was used to estimate the genetic distances amongst the 39 different populations. Data matrixes of pairwise F${}_{S\mathit{}T}$ values and D${}_A$ distances between different populations were visualized in two different heat maps. The color gradient for pairwise F${}_{S\mathit{}T}$ went from white to sea green and then red, while the color gradient for D${}_A$ distances went from white to yellow and then pink. Blocks in red or pink represent larger pairwise F${}_{S\mathit{}T}$ or D${}_A$ distances, respectively. From the distribution of pairwise F${}_{S\mathit{}T}$ values (Fig. 6A), we concluded that African populations were genetically more distant than most non-African populations, showing in particular larger pairwise F${}_{S\mathit{}T}$ values with East Asian populations than with non-East Asian populations. Smaller pairwise F${}_{S\mathit{}T}$ values were observed among different populations from the same continent, as seen by the representation of white and light-yellow blocks. The Mongolian and Ewenki ethnic groups shared close genetic relationships with most East Asian populations, but showed relatively less genetic affinities with KHV and HNL populations. The distribution of D${}_A$ distances among populations (Fig. 6B) mirrored the analogous population genetic relatedness observed with the distribution of pairwise F${}_{S\mathit{}T}$ values. The dataset for F${}_{S\mathit{}T}$ values and D${}_A$ distances is summarized in Supplementary Tables 4 and 5, respectively.

Fig. 6.

Heat maps of pairwise F${}_{\mathrm{S}\mathit{}\mathrm{T}}$ values and D${}_{\mathrm{A}}$ distances among the overall populations. (A) Data matrix of F${}_{S\mathit{}T}$ values among the overall 39 populations. The color gradient varies from white to red, corresponding to a F${}_{S\mathit{}T}$ from low to high. (B) D${}_A$ distances among the overall populations. The color gradient varies from light blue to pink, corresponding to the D${}_A$ distance from low to high.

Phylogenetic reconstruction of the overall 39 populations was also generated based on D${}_A$ distances to further assess the phylogeny relationships amongst these populations. As shown in Fig. 7, five primary clusters could easily be distinguished from the phylogenetic tree. These were the African cluster (ASW, ACB, LWK, MSL, GWD, ESN and YRI) labeled in green, the European cluster (FIN, IBS, TSI, CEU and GBR) in blue, the East Asian cluster (CDX, KHV, CHS, JPT, CHB, HNL, HAH, CDH, GDH, CDH, HNH, SXH, HLJH, Yi, ZGL, Tibetan, Ewenki and Mongolian) in red, the American cluster (MXL and PEL) in brown, and the South Asian cluster (BEB, STU, GIH, ITU and PJL) in purple. Two additional populations, PUR and CLM, were not positioned at the American cluster (MXL and PEL) and this could be explained by the mixed ancestry of these populations. Furthermore, the Mongolian and Ewenki groups were found to share a sub-branch in the East Asian cluster. Coupled with the genetic ancestral component revealed by STRUCTURE analysis (at K = 3), the results of phylogenetic reconstruction reconfirmed the genetic relatedness of these populations.

Fig. 7.

Phylogenetic reconstruction of the Mongolian group, Ewenki group and the 37 reference populations. Five primary clusters can be easily distinguished from the phylogenetic tree, i.e., the African cluster (ASW, ACB, LWK, MSL, GWD, ESN and YRI) labeled in green, the European cluster (FIN, IBS, TSI, CEU and GBR) labeled in blue, the East Asian cluster (CDX, KHV, CHS, JPT, CHB, HNL, HAH, CDH, GDH, CDH, HNH, SXH, HLJH, Yi, ZGL, Tibetan, Ewenki and Mongolian) in red, the American cluster (MXL and PEL) in brown and the South Asian cluster (BEB, STU, GIH, ITU and PJL) in purple. The other two American populations, PUR and CLM, are scattered among the European and South Asian populations.

To better understand the genetic relatedness of East Asian populations, locus-by-locus p values of intra-population genetic differences between Mongolian and Ewenki groups and the 16 reference populations were evaluated based on genotype data for the 47 InDels. The significance threshold level for p was adjusted to 0.0011 (p = 0.05/47 = 0.0011) according to the Bonferroni correction. As shown in Supplementary Table 6, significant differences were observed between Ewenki and CHS at two loci, with SDH, CDH and Yi at three loci, with CHB, GDH and ZGL at four loci, with CDX and Tibetan at five loci, with KHV at six loci, and with HAH at eight loci. None of the InDel loci showed significant differences between the Ewenki and HNH or SXH populations using the adjusted threshold of locus-by-locus p value. In contrast, significant differences were seen between Ewenki and HNL at ten loci. No significant differences were observed between the Mongolian group and the HNH, HLJH and SDH reference populations (Supplementary Table 7). The Tibetan and HNL groups were significantly different to the Mongolian group at eight and 17 loci, respectively. With regard to single locus diversity, the top three loci showing the greatest diversity between the two ethnic groups and the reference populations were rs140683187, rs10558392, rs3834231, rs79225518 and rs67939200 at nine, nine, seven, six and six loci, respectively, in the Ewenki group, and rs67939200, rs35267904 and rs140683187 at ten, nine, and eight loci, respectively, in the Mongolian group.