Integration of exome databases

The gnomAD version 2.1.1 exome data are available at https://gnomad.broadinstitute.org/downloads; and we downloaded data in the VCF format. The data included 17,209,972 variants from 125,748 samples of eight populations. The VCF file contained the rs-number, AN_POP for Allele Count, and AC_POP for Alt Allele Count for a population POP, where POP is one of the following: afr (African/African American), amr (Latino/Admixed American), asj (Ashkenazi Jewish), eas (East Asian), fin (Finnish), nfe (Non-Finish European), sas (South Asian), and oth (Other). We calculated the number of reference alleles as AN_POP–AC_POP for each variant per population.

Additionally, we downloaded GME Variome autosomal data from http://igm.ucsd.edu/gme/download.shtml. This project provides genotype frequencies of 669,953 variants on the autosomal chromosome for seven regions in the Greater Middle East (Northeast Africa, Northwest Africa, Arabian Peninsula, Israel, Syrian Desert, Turkish Peninsula, and Central Asia). The Arabian Peninsula included the following regions: Kuwait, Oman, Qatar, Kingdom of Saudi Arabia, the United Arab Emirates, and Yemen [9].

Since both databases are based on the human reference genome GRCh37/hg19, we merged the data when the chromosome number, nucleotide position, reference allele, and alternative allele were the same in both databases for each SNP. We included the alleles if the allele frequencies in each population were >0.01 to exclude rare variants. Because gnomAD data are based on populations, and GME Variome are based on their definitions; thus, we use “populations/regions” as a unit for this analysis. We disregarded the differences in allele frequencies among other populations/regions to identify important SNPs that are unique to the Arabian Peninsula population. We summed up the numbers of alleles over the populations/regions except for the Arabian Peninsula to compare allele frequencies between the Arabian Peninsula population and the other populations/regions (which will be referred to as "other regions" hereafter). In this study, we ignored the differences among populations in “other regions.”

Identification of T2D-related SNPs

To obtain T2D-related SNPs, we searched the ClinVar database (https://www.ncbi.nlm.nih.gov/clinvar/) with the keyword ‘”type 2 diabetes”[dis] OR “diabetes mellitus type 2”[dis]’. We extracted SNPs that were common among gnomAD exome, GME Variome, and the result of keyword search at the ClinVar database based on the chromosome, position, reference allele, and alternative allele. ClinVar provides clinical significance, which we used to evaluate the clinical significance of SNPs.

Statistical analyses

To compare the allele frequencies and genotype frequencies between the Arabian Peninsula and other regions, we divided the data into the Arabian Peninsula and the other regions.

The observed number of each allele in the Arabian Peninsula was calculated as follows:

For the other regions, the calculation was as follows: where i is the population except for the population in the Arabian Peninsula.

Then, the expected allele counts in the Arabian Peninsula were calculated as follows:

Additionally, we calculated the expected values for the genotype count using the same method. These expected allele counts represented the allele counts when the allele frequencies of each population were equal.

We performed Fisher’s exact test for each SNP to identify statistically significant SNPs that were not independent for the population, allele counts, or genotype counts. The Benjamini–Hochberg method was used with a q-value of 0.05 for the correction of multiple tests. For SNPs that were statistically significant, we further examined them using a residual analysis to reveal the differences between the observed counts and expected counts.

Integration of gnomAD exome and GME Variome

Before the GME Variome data became available, genotype data for the Middle Eastern populations were scarce. In 2016, an exome project on the Greater Middle Eastern region was finally published [9], which allowed for analyses of worldwide genotype data including Middle Eastern populations by integrating gnomAD exome and GME Variome data.

We downloaded 17,209,972 and 669,953 variants from gnomAD exome and GME Variome, respectively. We merged the genotype information for gnomAD exome and GME Variome and obtained 475,239 variants for 15 populations/regions. After eliminating the rare variants, we obtained 195,718 variants (Fig 1a). Then, we summed up the numbers of alleles for the Arabian Peninsula and other regions and calculated the expected value for each variant. The allele frequency spectra of overlapped variants showed similar tendencies among the Arabian Peninsula and other regions (S1a Fig) and also among each population/region (S1b Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Number of variants.

(a) The intersection between the gnomAD exome and GME Variome data. 475,239 variants commonly exist in two databases. After the eliminating rare variants, 195,718 variants remained. (b) Intersection among the gnomAD exome, GME Variome, and T2D-related variants.

https://doi.org/10.1371/journal.pone.0249226.g001

Significant T2D-related SNPs in the Arabian Peninsula

In addition to the genotype data, we integrated clinical information via ClinVar to reveal genetic features related to T2D. We downloaded 176 T2D-related variants from ClinVar as of January 17, 2021. Only 18 SNPs were common among the three databases (Table 1 and Fig 1b). To identify SNPs that were specifically biased in the Arabian Peninsula subpopulations, we performed Fisher’s exact test on these 18 SNPs. Even though the variants were mainly derived from populations outside of the Middle East, we found that six SNPs were statistically significant (Table 2). Two of the six SNPs were clinically significant, as described below. Among the two statistically and clinically significant SNPs, residual analysis revealed that a variant, rs13266634, increased the number of the risk allele (Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Standardized residuals of alleles.

White: non-risk allele and diagonal stripes: risk allele. (a) SLC30A8 rs13266634. (b) KCNJ11 rs5219.

https://doi.org/10.1371/journal.pone.0249226.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. List of T2D-related SNPs.

https://doi.org/10.1371/journal.pone.0249226.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Results of Fisher’s exact test for T2D-related SNPs.

https://doi.org/10.1371/journal.pone.0249226.t002

Interestingly, for rs13266634, the homozygous risk genotype increased whereas the heterozygous genotype decreased in the Arabian Peninsula (Fig 3). The remaining SNP was rs5219. The number of this risk allele decreased in the Arabian Peninsula (Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Standardized residuals of genotypes.

White: homozygote non-risk alleles, dotted: heterozygote, and diagonal stripes: homozygote of risk alleles. (a) SLC30A8 rs13266634. (b) KCNJ11 rs5219.

https://doi.org/10.1371/journal.pone.0249226.g003

The numbers of alleles and numbers of genotypes were tested. In the rs-number column, * represents the clinically significant SNP, # represents statistically significant SNP according to Fisher’s exact test the allele count, and § represents statistically significant SNP by Fisher’s exact test using the genotype count. The Benjamini–Hochberg method was used with a q-value of 0.05 for the correction of multiple tests. Observed and expected numbers of alleles and genotypes are shown. REF: reference allele, ALT: alternative allele, REF/REF: homozygote of reference allele, REF/ALT: heterozygote, ALT/ALT: homozygote of alternative allele, Fisher: p-value of Fisher’s exact test.