The variation landscape of CYP2D6 in a multi-ethnic Asian population

Abstract

Cytochrome P450 2D6 (CYP2D6) plays a crucial role in metabolizing approximately 20% of medications prescribed clinically. This enzyme is encoded by the CYP2D6 gene, known for its extensive polymorphism with over 170 catalogued haplotypes or star alleles, which can have a profound impact on drug efficacy and safety. Despite its importance, a gap exists in the global genomic databases, which are predominantly representative of European ancestries, thereby limiting comprehensive knowledge of CYP2D6 variation in ethnically diverse populations. In an effort to bridge this knowledge gap, we focused on elucidating the CYP2D6 variation landscape within a multi-ethnic Asian cohort, encompassing individuals of Chinese, Malay, and Indian descent. Our study comprised data analysis of 1850 whole genomes from the SG10K_Health dataset using an in-house consensus algorithm, which integrates the capabilities of Cyrius, Aldy, and StellarPGx. This analysis unveiled distinct population-specific star-allele distribution trends, highlighting the unique genetic makeup of the Singaporean population. Significantly, 46% of our cohort harbored actionable CYP2D6 variants-those with direct implications for drug dosing and treatment strategies. Furthermore, we identified 14 potential novel CYP2D6 star-alleles, of which 7 were observed in multiple individuals, suggesting their broader relevance. Overall, our study contributes novel data on CYP2D6 genetic variations specific to the Southeast Asian context. The findings are instrumental for the advancement of pharmacogenomics and personalized medicine, not only in Southeast Asia but also in other regions with comparable genetic diversity.

Figure 1

Overview of data analysis workflow and study design. (a) Data analysis pipeline. The diagram illustrates the workflow for analyzing WGS 30X CRAM files. Three tools —Cyrius, Aldy, and StellarPGx— are utilized for CYP2D6 genotyping. A consensus algorithm determines the final diplotypes based on agreement from at least two of the three tools. StellarPGx is employed to identify potential novel alleles. (b) SG10K_Health cohort. The cohort for the study comprises individuals from the three major ethnic groups in Singapore, namely Chinese, Malay, and Indian. Colored bars represent the count and proportion of consensus diplotypes within the total cohort of 1,850 participants. (c) Comparison of call rates. The bar graph compares the successful calling rates of Cyrius, Aldy, and StellarPGx in generating diplotype calls. (d) UpSet plot of CYP2D6 diplotype calls. The plot displays the overlap in diplotype calls among the three callers. Numbers atop the bars show the size of each intersection set. The lower left side of the plot indicates the total successful calls per tool. The unique diplotypes identified by each tool are shown in the three leftmost bars (Aldy with 487, StellarPGx with 585 and Cyrius with 267 unique diplotype calls). The middle bars represent diplotypes identified by tool pairs, and the far-right bar shows 1040 diplotypes identified by all three tools.

Figure 2

Distribution of key CYP2D6 star alleles and phenotypes across different genetic ancestries. (a) Frequency distribution of the top 10 CYP2D6 star alleles in the SG10K_Health dataset, categorized by genetic ancestries. The x-axis depicts the star alleles, each annotated with their respective functional classification (normal, decreased, or no function). The y-axis shows the frequency distribution (%) of these star alleles within each genetic ancestry group. (b) Comparative analysis of CYP2D6 star allele frequencies between the SG10K_Health dataset and PharmGKB. The x-axis displays the frequency of star alleles observed in the SG10K_Health dataset. The y-axis indicates the frequency of star alleles as reported in PharmGKB for East Asian and Central/South Asian populations. The Pearson correlation coefficient (r) is represented for each comparison. (c) Frequency distribution of CYP2D6 phenotypes among different genetic ancestries. The x-axis lists all the possible CYP2D6 phenotypes, including normal metabolizers (NM), intermediate metabolizers (IM), indeterminate, ultrarapid metabolizers (UM), and poor metabolizers (PM). The y-axis shows the frequency distribution (%) of these phenotypes across each genetic ancestry.

Figure 3

Frequency distribution of CYP2D6 structural variations (SVs). (a) Frequency distribution of SV-containing CYP2D6 star alleles in the SG10K_Health dataset, categorized by genetic ancestries. The x-axis depicts the star alleles, each annotated with their respective functional classification (increased, decreased, or no function). The y-axis shows the frequency distribution (%) of these star alleles within each genetic ancestry group. (b) Comparison of the proportion of samples with SV and samples with no SV in the SG10K_Health dataset (N = 1,487). (c) Distribution of CYP2D6 phenotypes in the SG10K_Health dataset, categorized by SV presence. The left side illustrates the proportion in samples with SV (N = 827), and the right side shows the proportion in samples with no SV (N = 660).

Figure 4

Visual representation of potential novel alleles with *2 and *10 background alleles. The core variants of the background alleles are marked in blue, and the unique novel variants are marked in green.