Taiwan Biobank: A rich biomedical research database of the Taiwanese population

Abstract

The Taiwan Biobank (TWB) is an ongoing prospective study of >150,000 individuals aged 20-70 in Taiwan. A comprehensive list of phenotypes was collected for each consented participant at recruitment and follow-up visits through structured interviews and physical measurements. Biomarkers and genetic data were generated from blood and urine samples. We present here an overview of TWB's genetic data quality, population structure, and familial relationship, which consists of predominantly Han Chinese ancestry, and highlight its important attributes and genetic findings thus far. A linkage to Taiwan's National Health Insurance database of >25 years and other registries is underway to enrich the phenotypic spectrum and enable deep and longitudinal genetic investigations. TWB provides one of the largest biobank resources for biomedical and public health research in East Asia that will contribute to our understanding of the genetic basis of health and disease in global populations through collaborative studies with other biobanks.

Graphical abstract

Figure 1

Design and demographics of the TWB (A) TWB enrolled men and women aged 20–70 across recruitment centers in Taiwan that account for population density, with every county/city having at least one recruitment site (indicated by the location marks; a darker color represents a higher population density. The current release comprises individuals aged 30–70, but future releases will include participants down to age 20). Phenotypes were collected at baseline through a structured interview, physical examination, and blood/urine tests for each participant, with repeated measurements taken 2–4 years after the first visit. As of August 2021, 37,508 individuals have finished the first round of follow-up among all 150,000 participants. Multi-omics data were generated for all or subsets of the participants, including array genotyping, WGS, HLA typing, DNA methylation, and blood metabolome (see also Tables S1–S3). A linkage to the NHIRD and other health-related registries is underway to provide phenotypic information in addition to self-report for the TWB. Enrollment is expected to continue until reaching the target of 200,000 volunteers. (B) The distribution of age, sex, and education level of study participants at baseline. More female than male participants were enrolled in TWB. Age of the participants was evenly distributed across every 10-year age bracket. People with a college degree accounted for the largest proportion in the current cohort.

Figure 2

Population structure and familial relationship within TWB (A) PCA on the QC’ed genotype data revealed a homogeneous population structure of primarily Han Chinese descent among the TWB participants, which can be further divided into three distinct subgroups representing different geographic and ancestral origins (Holo, Hakka, and Mainlanders). Mainlanders were roughly separated into Southern and Northern Chinese for visualization (details in Table S4, based on the GWAS sample in Chen et al.³). Participants with the same paternal and maternal place of ancestral origin were assigned into one single subgroup; those with mixed origins were assigned with A/B labels. Projection of TWB onto 1KG data showed tight clustering with the EAS superpopulation as well as the two Han Chinese populations (Figure S1). Shown here are results from batch 2 of 66,000 individuals. PC plots for batch 1, while not shown, were nearly identical to the batch 2 results. (B) Kinship estimation of TWB participants showed a non-trivial number of relatedness, including over 25,000 pairs who are third-degree relatives or closer across batch 1 and batch 2 (Table S5, based on the GWAS sample in Chen et al.³). Plotted here are pairs of individuals in batch 2 with a kinship coefficient >0.02 (y axis; within fourth-degree relatedness), estimated using KING, against the proportion of loci with 0 allele shared by descent (Z0; x axis). Dashed lines indicate the midpoints of cutoffs commonly used for determining the pairwise familial relationship based on kinship coefficients (duplicates/MZ twins: kinship > 0.353; parent-offspring: 0.177 < kinship < 0.353 and Z0 < 0.05; full siblings: 0.177 < kinship < 0.353 and Z0 > 0.05; second-degree relatives: 0.088 < kinship < 0.177; third-degree relatives: 0.044 < kinship < 0.088; fourth-degree relatives: kinship < 0.044). A substantial number of TWB participants were found to be fourth-degree relatives, among whom a random subset of 10,000 pairs were included for demonstration in the figure.