The present study has been approved by the ethics committee of Xi’an Jiaotong University Health Science Center and Southern Medical University (Ethical Approval Number: 2019-1039). This research was conducted in accordance with the ethical principle for medical research involving human subjects recommended by the World Medical Association Declaration of Helsinki. Sample collection and the following sequencing experiments were performed in strict compliance with the ethical regulations. Peripheral venous blood samples were collected from 101 unrelated volunteers who lived in the Gannan Tibetan Autonomous Prefecture in the Gansu province, China for at least three generations. The blood samples were dried on the FTA cards, and then stored at room temperature. Each volunteer signed an informed consent form before the sample collection.

For each sample, five pieces of 1.0 mm${}^2$ bloodstain were used to extract the genomic DNA using the QIAamp® DNA Investigator kit (QIAGEN, Hilden, Germany) according to the manufacturer’s instructions [25]. The concentration of genomic DNA was then measured by quantitative fluorescence using the Qubit™ dsDNA HS Assay kit (Thermo Fisher Scientific Inc, Waltham, MA, USA) and the Qubit 4.0 Fluorometer (Thermo Fisher Scientific Inc, Waltham, MA, USA) following the manufacturer’s recommendations [26].

The Precision ID DL8 Kit and the Precision ID Ancestry Panel (Thermo Fisher Scientific Inc, Waltham, MA, USA) were used to prepare DNA library on the automated Ion Chef™ system (Thermo Fisher Scientific Inc, Waltham, MA, USA). For each sample, 15 µL of DNA (1 ng in total) was pipetted into A1 to A8 wells of a Precision ID DL8 IonCode™ Barcode Adapters plate. And twenty-one cycles were used to amplify the target regions. The concentrations of sample libraries were measured by the Ion Library TaqMan® Quantitation Kit (Thermo Fisher Scientific Inc, Waltham, MA, USA) on the QuantStudio™ 5 real-time PCR system (Thermo Fisher Scientific Inc, Waltham, MA, USA) following the manufacturer’s instructions. Before template preparation, all DNA libraries were pooled in equimolar concentrations (33 pM).

Template preparation, enrichment of beads with template, and the sequence of template on beads were performed with the Ion Chef™ and Ion S5™ XL (Thermo Fisher Scientific Inc, Waltham, MA, USA) instruments according to the manufacturer’s instructions. After template preparation, high-throughput sequencing was preformed using four pieces of Ion™ 530 chips (Thermo Fisher Scientific Inc, Waltham, MA, USA). Thirty-two samples were loaded on each Ion™ 530 chip. Positive DNA 007 and deionized water were used as positive and negative controls in each run.

The analysis of primary sequence data was performed with the Torrent Suite Software (Thermo Fisher Scientific Inc, Waltham, MA, USA). All reads were aligned to hg19 reference sequences. Further sequence analyses were carried out using the HID-SNP Genotyper v4.2 plug-in and Converge™ Software (Thermo Fisher Scientific Inc, Waltham, MA, USA) with default analysis settings. Statistical analyses of read depths and balance ratios for the 165 AI-SNPs were calculated using the OriginPro 2021b software (version 9.8, OriginLab Corporation, Northampton, MA, USA).

DNA samples with an ‘A’ allele at locus rs7722456 were further confirmed by Sanger sequencing. Forward primer (5′-GCTGACTCTAGCCCTTTGGG-3′) and reverse primer (5′-GGTGGGTCTTGTTTGGCATT-3′) were synthesized and then used to amplify the locus rs7722456. The PCR products were separated by the method of agarose gel electrophoresis, and then purified using the DiaSpin DNA Gel Extraction kit (Sangon Biotech, Guangzhou, China) following manufacturer’s instructions. Sanger sequencing was conducted at rs7722456 locus by Sangon Biotech (Sangon Biotech, Guangzhou, China).

In the present study, 164 AI-SNP data sets of 26 populations were acquired from the 1000 Genomes Project Phase III database (locus rs10954737 was excluded from the population genetic studies as its genotype data were not available in the 1000 Genomes Project) [27]. The 165 AIM-SNP data sets for 22 populations were obtained from HGDP-CEPH database [28]. SNP genotype data for nine populations (Sichuan Tibetan, Qinghai Tibetan and Liangshan Yi [29], Hainan Han, Hainan Li and Gelao [23], Wuzhong Hui [30], Chinese Kazak [31], and Basque [32]) were obtained from previously published literature. Detailed information, such as population names and their abbreviations, population sizes of all reference populations, is listed in Supplementary Table 1.

The linkage disequilibrium (LD) test, allelic frequencies and forensic statistical parameters of the 165 AI-SNP loci in the Gannan Tibetan group were calculated using STRAF [33]. Tests of Hardy-Weinberg equilibrium (HWE) for 165 AI-SNP loci in Gannan Tibetan group were performed by the Arlequin software (version 3.5.1.2, Laurent Excoffier & Heidi Lischer, Berne, Switzerland) [34] . The heatmap for the minimal allelic frequencies (MAF) of 164 AI- SNP loci was created using the R software (version 4.0.5, R Foundation for Statistical Computing, Vienna, Austria). Pairwise fixation index (F${}_{\text{𝑆𝑇}}$) values, which were used to measure the genetic distances of pairwise populations, were calculated between Gannan Tibetan group and reference populations using the Arlequin software (version 3.5.1.2, Excoffier & Lischer, Berne, Switzerland) [34]. Moreover, Nei’s genetic distances (D${}_{\text{𝐴}}$) between the Gannan Tibetan group and other reference populations were calculated using the DISPAN program (Nei & Tajima & Tateno, Kanagawa, Japan) [35]. A neighbor-joining (NJ) phylogenetic tree was constructed on the basis of pairwise D${}_{\text{𝐴}}$ values by the MEGA software (version 7, Sudhir Kumar & Glen Stecher & Koichiro Tamura, PA, USA) [36], and then visualized with the ggtree package (version 3.4.2, Guangchuang Yu, Guangdong, China) [37] from the R software (version 4.0.5, R Foundation for Statistical Computing, Vienna, Austria). Phylogenetic trees constructed using the maximum likelihood (ML) method were generated using the TreeMix software (version 1.13, Joseph K. Pickrell & Jonathan K. Pritchard, IL, USA), which assumed that 0–10 migration events happened among all the populations [38]. Principal component analyses (PCA) were also carried out to evaluate the genetic relationships between the Gannan Tibetan group and reference groups. PCAs at individual scale were performed based on the genotype data of 164 AI-SNP loci of Gannan Tibetan group and reference populations using the R software and the SmartPCA package implemented in the EIGENSOFT software (version 6.1.4, Alkes L Price & Nick J Patterson, Boston, MA, USA), respectively [39]. Population genetics and ancestry component analyses of the Gannan Tibetan group were performed using the model-based ADMIXTURE (version 1.3, David H Alexander & John Novembre & Kenneth Lange, Los Angeles, CA, USA) and STRUCTURE softwares (version 2.3.4, Pritchard & Stephens & Donnelly, Oxford, UK) [40, 41] with the setting K values from 2 to 6. The CLUMPAK tool (http://clumpak.tau.ac.il/) was used to analyze the results of ADMIXTURE (version 1.3, David H Alexander & John Novembre & Kenneth Lange, Los Angeles, CA, USA).

In this study, all 165 AI-SNP genotypes in 101 Gannan Tibetan individuals and positive control DNA were successfully generated using the Ion S5™ XL system. Four Ion™ 530 chips generated a total of 1.5~1.87 gigabases (Gb). Ion sphere particle (ISP) loading rate of each Ion™ 530 chip was more than 78%. As recommended by the manufacturer’s instructions, total reads should be greater than 1 Gb; the ISP loading rate should be greater than 50%; and the total usable reads should be greater than 30%. These sequencing metrics all met the recommended values of the manufacturer’s instructions. Read depths of 165 AI-SNP loci in 101 Gannan Tibetan individuals are shown in Supplementary Fig. 1. The minimum depth of coverage was 24$\times$, which was noted at rs9845457 locus in sample GNZ13, while the maximum read depth was 8287$\times$, which was seen at rs7657799 locus in GNZ54. The mean read depths for all 101 individuals ranged from 231 $\pm$ 87$\times$ (mean $\pm$ standard deviation) to 2932 $\pm$ 1359$\times$. The balance ratios of all amplicons from the forward direction in the 101 samples are shown in Fig. 1, and the box plots revealed that the balance ratios of almost all amplicons were around 0.5.

Fig. 1.

Box plots of the balance ratios from forward direction of all SNP loci in 101 Gannan Tibetan individuals.

Sixty-three individuals (accounting for 62.4% of total tested Gannan Tibetan individuals) were genotyped ‘AT’ at rs7722456. Genotypes of rs7722456 at these samples were further confirmed by Sanger sequencing. Results of Sanger sequencing showed that the genotypes in these 63 individuals were ‘TT’ at rs7722456 (Supplementary Fig. 2).

Before the population genetic investigations, the HWE tests of the 165 AI-SNPs were carried out in the Gannan Tibetan group, and the results are shown in Supplementary Table 2. The rs1462906, rs1871534, rs2814778, rs3916235, rs4880436, rs4891825, rs6754311, rs7326934 and rs7722456 of 165 AI-SNPs were excluded from the HWE tests because of their homozygous genotypes. The results of HWE tests showed that no significant deviations from HWE for the remaining 156 AI-SNPs were observed in the Gannan Tibetan group after sequential Bonferroni corrections (p = 0.05/156 = 0.0003).

The LD tests for pairwise SNP loci were performed based on the genotype data, and the p-values are shown in Supplementary Table 3. Similarly, rs1462906, rs1871534, rs2814778, rs3916235, rs4880436, rs4891825, rs6754311, rs7326934 and rs7722456 were also excluded from the LD tests because of their homozygous genotypes. After sequential Bonferroni corrections (p = 0.05/24150 = 2.07 $\times$ 10${}^{-6}$), all pairwise SNP loci reached to the state of linkage equilibrium in the studied Gannan Tibetan group except for the pair of rs1229984 and rs3811801 (p = 8.72 $\times$ 10${}^{-19}$).

Among the 165 AI-SNPs, nine loci (rs1462906, rs1871534, rs2814778, rs3916235, rs4880436, rs4891825, rs6754311, rs7326934 and rs7722456) were homozygous. Forensic statistical parameters including expected heterozygosity (H${}_{\text{exp}}$), matching probability (MP), power of discrimination (PD), polymorphism information content (PIC), observed heterozygosity (H${}_{\text{obs}}$), probability of exclusion (PE) and typical paternity index (TPI) were calculated for the 165 AI-SNPs in the Gannan Tibetan group, and the results are shown in Supplementary Table 2. Nine homozygous SNP loci showed the minimum values of H${}_{\text{exp}}$, H${}_{\text{obs}}$, PD, PE, PIC and TPI in the Gannan Tibetan group. Among the 165 AI-SNP loci, 71 AI-SNP loci showed H${}_{\text{exp}}$ values greater than 0.4 while 37 SNP loci showed H${}_{\text{exp}}$ less than 0.2. There were 79 SNP loci with PIC values greater than 0.3, whereas 25 loci showed relatively small PIC values (PIC 0.1). Except for nine homozygous SNP loci, rs3943253 locus displayed the largest PD value (0.6662), and the minimum values of PD were noted at rs12913832, rs17642714, and rs2042762 loci (PD = 0.0196). The largest PE value was 0.2957, which was observed at rs37369 locus, while the minimum PE value was 0.0001 at rs12913832, rs17642714, and rs2042762 loci. The combined PD (CPD) and combined PE (CPE) values for the 164 AI-SNP loci in the Gannan Tibetan group were 1-7.269 $\times$ 10${}^{-47}$ and 0.999999941, respectively (rs3811801 was excluded from the calculations due to LD).

Admixture predictions and population likelihood analyses for 101 Gannan Tibetans were first conducted using the plugin in Torrent Suite, and all individuals were predicted as East Asian individuals. When trying to predict the ancestral information of East Asian subgroups using this panel, the prediction results are not particularly accurate (Supplementary Fig. 3). Genetic ancestry analyses for the Gannan Tibetan group were evaluated based on the 164 AI-SNP loci using both the ADMIXTURE (version 1.3, David H Alexander & John Novembre & Kenneth Lange, CA, USA) and STRUCTURE softwares (version 2.3.4, Pritchard & Stephens & Donnelly, UK), and the results are shown in Fig. 2. The Bayesian clustering analyses (K values are set to 2~6) and corresponding CV error values are shown in Fig. 2A,B, respectively. The smallest CV error value was found when K = 5 (Fig. 2B, CV error = 0.50892). Therefore, we adopted the K = 5 as the optimal K value, and all individuals were assigned to five different clusters, which could differentiate among African (a cyan-based ancestry cluster), American (a mixture of purple and green), South Asian (a brick-red-based ancestry cluster), East Asian (an orange-based ancestry cluster) and European populations (a purple-based ancestry cluster), as shown in Fig. 2A.

Fig. 2.

Analyses of genetic ancestry for the Gannan Tibetan group (GNZ) on the basis of 164 AI-SNP loci. (A) Bayesian clustering analyses with predefined K values from 3 to 6. (B) Line chart of CV error values at K values from 2 to 8. (C) African, East Asian and European ancestral components in the Gannan Tibetan group.

The Bayesian structure analyses indicated that the Gannan Tibetan group shared similar genetic ancestry structure with East Asian populations when K values were set to 3~5. Cluster analysis also showed that the Gannan Tibetans clustered tightly with the East Asian individuals (Fig. 2C). At K = 6, subcluster with pink-based ancestry structure was identified in Tibeto-Burman groups such as the Gannan Tibetan group, the Sichuan Tibetan group and the Liangshan Yi group. The Gannan Tibetan group displayed a mixture of pink and orange ancestry structure, which were different from Han populations in flat land. The studied Gannan Tibetan group shared similar ancestry structure with groups residing in its adjacent areas, such as the Sichuan Tibetan group and the Liangshan Yi group.