Establishing multiple omics baselines for three Southeast Asian populations in the Singapore Integrative Omics Study

Abstract

The Singapore Integrative Omics Study provides valuable insights on establishing population reference measurement in 364 Chinese, Malay, and Indian individuals. These measurements include > 2.5 millions genetic variants, 21,649 transcripts expression, 282 lipid species quantification, and 284 clinical, lifestyle, and dietary variables. This concept paper introduces the depth of the data resource, and investigates the extent of ethnic variation at these omics and non-omics biomarkers. It is evident that there are specific biomarkers in each of these platforms to differentiate between the ethnicities, and intra-population analyses suggest that Chinese and Indians are the most biologically homogeneous and heterogeneous, respectively, of the three groups. Consistent patterns of correlations between lipid species also suggest the possibility of lipid tagging to simplify future lipidomics assays. The Singapore Integrative Omics Study is expected to allow the characterization of intra-omic and inter-omic correlations within and across all three ethnic groups through a systems biology approach.The Singapore Genome Variation projects characterized the genetics of Singapore's Chinese, Malay, and Indian populations. The Singapore Integrative Omics Study introduced here goes further in providing multi-omic measurements in individuals from these populations, including genetic, transcriptome, lipidome, and lifestyle data, and will facilitate the study of common diseases in Asian communities.

Fig. 1

PCAs of omics and clinical/lifestyle/diet data. Biplots are shown for five distinct PCAs using the respective first two axes of variations from each PCA. The five PCAs correspond to the analysis of: a 101,099 autosomal SNPs pseudo-randomly chosen to minimize linkage disequilibrium between the SNPs; b 21,649 gene transcript probesets; c 274 miRNAs; d 282 lipid species; e a set of 284 clinical, lifestyle, and dietary variables; and f only the 199 dietary variables. Each circle represents an individual from the iOmics and is assigned a color corresponding to the self-reported ethnicity of the subject, according to the color legend on the top right panel in a

Fig. 2

Distribution of the Wright F_ST value across three ethnic groups at the eight HLA loci. The distribution of the Wright F_ST value across three populations at the eight HLA loci. The alleles shown in the plot are the top three F_ST alleles at each HLA loci. The triangular shape indicates the HLA alleles, where the F_ST values are driven by differences between Chinese and Indians. The diamond shape indicates the HLA alleles, where the F_ST values are driven by the differences between Chinese and Malays. The square shape indicates the HLA alleles, where the F_ST values are driven by the differences between Malays and Indians. Shapes with red color outline are representing drug-associated HLA alleles (Table 3)

Fig. 3

A combined boxplot and scatter plot of the top three significant transcript probesets across three populations. The combined plot showing distribution of transcript intensities of a 7912136:UTS2 gene, b 8041061:PLB1 gene, c 8102362:TIFA gene across three populations. P-values were calculated using ANOVA, adjusted for batch effect and gender, and corrected for Bonferroni. The upper whisker represents either the maximum value observed or is 1.5 times the interquartile range greater than the third quartile, whichever is smaller. The lower whisker represents either the minimum value observed or is 1.5 times lower than the first quartile, whichever is greater. The details of the significant transcript probesets across three populations can be found in Supplementary Data 3

Fig. 4

Correlation heatmap of the 282 lipids in each ethnic group. The correlation heatmap between 282 lipids in the a Chinese; b Malays; and c Indians. The correlation was calculated by using concentration of the lipids via Pearson’s correlation, r ². The lipids are first categorized into lipid category and followed by lipid classes. In each of the lipid class, the lipid species are ordered according to their carbon chain length and degree of unsaturation (number of double bonds). The intensity of the color reflects the magnitude of the correlation, in which the white color means r ² = 0 and red color means r ² = 1

Fig. 5

Distribution of the differences and correlation of the 16 significant differentiated clinical phenotypes across three populations. The distribution of the differences of the 16 clinical significant phenotypes across three populations, with a correlation heatmap of the phenotypes. The correlation was calculated using Pearson’s correlation, r², across 358 samples. The intensity of the color reflects the magnitude of the correlation, in which the white color means r² = 0 and red color means r² = 1. The details of the differences can be found in Supplementary Table 9