A multi-ancestry genome-wide association study and evaluation of polygenic scores of LDL-C levels

Abstract

The genetic determinants of low-density lipoprotein cholesterol (LDL-C) levels in blood have been predominantly explored in European populations and remain poorly understood in Middle Eastern populations. We investigated the genetic architecture of LDL-C variation in Qatar by conducting a genome-wide association study (GWAS) on serum LDL-C levels using whole genome sequencing data of 13,701 individuals (discovery; n = 5,939, replication; n = 7,762) from the population-based Qatar Biobank (QBB) cohort. We replicated 168 previously reported loci from the largest LDL-C GWAS by the Global Lipids Genetics Consortium (GLGC), with high correlation in allele frequencies (R² = 0.77) and moderate correlation in effect sizes (Beta; R² = 0.53). We also performed a multi-ancestry meta-analysis with the GLGC study using MR-MEGA (Meta-Regression of Multi-Ethnic Genetic Association) and identified one novel LDL-C-associated locus; rs10939663 (SLC2A9; genomic control-corrected P = 1.25 × 10^-8). Lastly, we developed Qatari-specific polygenic score (PGS) panels and tested their performance against PGS derived from other ancestries. The multi-ancestry-derived PGS (PGS000888) performed best at predicting LDL-C levels, whilst the Qatari-derived PGS showed comparable performance. Overall, we report a novel gene associated with LDL-C levels, which may be explored further to decipher its potential role in the etiopathogenesis of cardiovascular diseases. Our findings also highlight the importance of population-based genetics in developing PGS for clinical utilization.

Keywords: LDL; atherosclerosis; cholesterol; dyslipidemias; genomics.

Fig. 1

Study design and analysis workflow. LDL-C GWAS was performed on the Qatar Biobank (QBB) discovery (n = 5,939) and replication (n = 7,762) cohorts separately using inverse-normalized LDL-C residuals. GWAS summary statistics from the QBB cohorts were then combined in a single-ancestry, fixed-effect meta-analysis using METAL. The results were further meta-analyzed using (MR-MEGA-v0.2) with 5 single-ancestry cohort summary statistics data (African; AFR; n = 99.4k, East-Asian: EAS; n = 146.5k, European; EUR; n = 1.32 m, Hispanic; HIS; n = 48.1k and South Asian; SAS; n = 41.0k) from the GLGC LDL-C study (11). GWAS results from the discovery cohort were then used to develop multiple PGS panels using different tools (refer to methods section). 27 QBB-derived PGS panels (QBB_PGS) were then tested on the QBB replication cohort. Previously derived PGS from other ancestries (African (AFR (11), European (EUR (11, 15, 16, 34)), East Asian (EAS (11)), South Asian (SAS (11)) and Multi-ancestry (MULTI (11)) were also tested in the QBB replication cohort.

Fig. 2

Manhattan and Quantile-Quantile plots from QBB GWAS discovery cohort. A: The Manhattan plot represents genetic variants (dots) plotted on the x-axis in accordance with chromosome position against their corresponding −log₁₀(P). The horizontally marked red line indicates the genome-wide significance threshold (P = 5.0 × 10⁻⁸). The horizontal blue line indicates the suggestive significance threshold (P = 1.0 × 10⁻⁵). B: The quantile-quantile plot shows the expected versus observed −Log₁₀ (P); λGC is the genomic inflation factor. Analysis was based on data from the discovery cohort (n = 5,939) and performed using SAIGE adjusted for age, age², the first 10 genetic principal components (PC1 to PC10) applying inverse-normalization of the LDL-C residuals and adjustment for relatedness.

Fig. 3

Comparison of allele frequency and effect size for replicated genetic variants. Scatter plots represent the (A) Allele frequency and (B) Effect size (Beta) comparisons between the QBB discovery cohort (n = 5,939) and data from the multi-ancestry GWAS study by Graham et al., (n = 1,654,960). R² is the coefficient of determination from correlation analysis. The red dotted line represents the line of best fit from linear regression.

Fig. 4

Regional Association plot for the novel locus identified from the multi-ancestry meta-analysis for LDL-C. A novel locus identified from the multi-ancestry meta-analysis of the QBB cohort (n = 13,701) combined with the GLGC multi-ancestry cohort (n = 1,654,960). Variants are plotted using the GRCh38 build with x-axis representing the position in the chromosome and the y-axis representing the –log10(P-values). rs10939663 located in SLC2A9. The linkage disequilibrium (LD) reference from the local population was generated from the QBB data. The color of the plotted dots represents the LD r² value with the lead variants depicted as diamond-shaped markers. Recombination rates are also plotted with the corresponding value on the right y-axis. The red line represents −log10 of the genome-wide significance threshold (P = 5.0 × 10⁻⁸).

Fig. 5

Performance assessment of polygenic score (PGS) panels derived from discovery cohort (n = 5,939) and tested on the replication cohort (n = 7,762). Interleaved scatter plot shows the adjusted R² values from the linear regression models for raw LDL-C values with PGS derived using different-P-value thresholds (x-axis) and clumping LD r² threshold values (0.2 or 0.8). The regression model included the PGS, age, gender, PC1-PC4, and cholesterol treatment as predictors. Data points represent mean adjusted R² with 95% Confidence Intervals (CI). Green diamond datapoints represent PGS derived based on thresholding only while PGS derived based on thresholding and clumping with r² < 0.2 are shown in blue and those with r² < 0.8 are shown in red.

Fig. 6

Performance metrics for the polygenic score panels. A: Bar plot shows the adjusted R² values from the performance of the 17 different PGS panels tested on the QBB replication cohort (n = 7,762). Performance was assessed using linear regression for raw LDL-C values using PGS, age, gender, PCs1-4, and cholesterol treatment as predictors in the model. B: Bar plot shows the adjusted R² from the LDL-C linear regression models for different PGS panels (n = 7); EUR_36SNP (n = 4,787) (34), EUR_12SNP (n = 3,020) (15) and EUR_1MNSNP (n = 389,158) (11) tested on European populations, AFR_295SNP tested on African ancestry (n = 6,863), MULTI_9009SNP tested on multi-ancestry populations (n = 461,918), EAS_66SNP tested on East Asian populations (n = 1,441) and SAS_13SNP tested on South Asian populations (n = 6,814) (11). Asterisks (∗) represent statistical significance for PGS from regression analysis (P < 0.05). QBB: Qatar Biobank, EUR: European, AFR: African, ASN: Asian, SAS: South Asian and EAS: East Asian populations.