(P29-015-24) Using Customized Polygenic Risk Scores to Improve the Study of Nutrient Metabolism Gene Effects in Nutrient-Gene Association Models

Objectives: This study aims to enhance the precision of association studies by distinguishing the effects of individual nutrient metabolism genes from the broader genetic background on various health measures.

Methods: The approach involves generating polygenic risk scores (PRS) from genome-wide association study (GWAS) summary statistics to be utilized in regression models to assess nutrient-gene associations with health. The health impact of a single gene of interest is often embedded within pre-existing, genome-wide PRS, making separation challenging; to address this hurdle, we created a customized PRS and excluded the genetic region of the gene of interest. We tested this approach on metal transporter gene SLC39A8 and single nucleotide polymorphism (SNP) rs13107325 due to its known associations with increased body mass index (BMI) and reduced circulating high-density lipoprotein (HDL). We used the UK Biobank population cohort (Caucasian only, no kinship) and PRSice-2 software to compute individual PRS as an additional (experimental) covariate. Quantile (median) regressions included rs13107325 genotype (encoded as 0, 1, or 2 under an additive model) as the primary exposure, BMI or HDL as the primary outcome, and the experimental PRS for BMI or HDL (omitting a ±1 million base pair region around SLC39A8), dietary manganese intake, age, sex, deprivation index (as a measure of socioeconomic status), education (qualifications), and household income as covariates. Control regressions lack the experimental PRS as a variable.

Results: The results revealed that inclusion of the experimental PRS increased statistical significance and narrowed variability for BMI (p-value=5.97E-10 to 4.46E-12, beta=0.249±0.0402 to 0.274±0.0395, N=107,966) and HDL (p-value=2.99E-16 to 1.08E-16, beta=-0.0307±0.00375 to -0.0285±0.00343, N=95,318).

Conclusions: This tailored PRS demonstrates improved statistical power in regression models, thereby better elucidating the influence of single nutrient metabolism genes and their respective SNPs on health measures. This methodology can be universally applied to any target nutrient metabolism gene and enhance the robustness of genetic associations.

Funding Sources: American Heart Association and President’s Faculty Fellows Research Award