<i>FTO</i> Haplotypes involving index SNPs and their association with BMI and type 2 diabetes.

It is assumed that a cluster of base-substitutions (involving rs1421085, rs3751812, rs9941349, rs56137030, rs17817964, and rs12149832) initially occurred, followed by another cluster of base-substitutions (involving rs9939609) within the target interval at FTO. Then, presumably the underlined alleles were produced by either recombination or recurrent base-substitutions that occurred independently at each SNP (see Figure 2 and Table S5). Ancestral types of alleles are shown in bold letters.
In the estimates of haplotype frequency, they are labelled with “Low”when the frequency is <0.004 (except for H1–2 in the European-ancestry population).
There is recombination between rs7206790 and other assayed SNPs. For rs7206790, haplotypes with recombination are further split by two alleles; the major and minor alleles are shown in the left and right sides of the slash for the corresponding haplotypes.
rs3751812 was genotyped in T2D-association samples alone (but not in BMI-association samples) because rs3751812 and rs9939609 could not be differentiated due to the absence of H2 haplotype class in populations of East Asian and European ancestry. BMI/T2D association in each population was calculated by using haplotype-based linear/logistic regression with PLINK. The effect sizes of three Asian populations (Japanese, Vietnamese and Sri Lankan) were combined by a fixed-effect meta-analysis, while SriLankan data were not used for H1-3 and H3-4 because of the low frequency.
*Besides the SNPs tagging individual LD clusters that were drawn from T2D association results (see Figure 1), rs1421085 and rs56137030 were characterized in the present study, because they were reported to be promising candidates for causal variants at FTO in African American populations [7].
**Test of heterogeneity in effect size between the three populations meta-analyzed in the present study.