Development of genome-wide polygenic risk scores for lipid traits and clinical applications for dyslipidemia, subclinical atherosclerosis, and diabetes cardiovascular complications among East Asians

Abstract

Background The clinical utility of personal genomic information in identifying individuals at increased risks for dyslipidemia and cardiovascular diseases remains unclear. Methods We used data from Biobank Japan (n = 70,657–128,305) and developed novel East Asian-specific genome-wide polygenic risk scores (PRSs) for four lipid traits. We validated (n = 4271) and subsequently tested associations of these scores with 3-year lipid changes in adolescents (n = 620), carotid intima-media thickness (cIMT) in adult women (n = 781), dyslipidemia (n = 7723), and coronary heart disease (CHD) (n = 2374 cases and 6246 controls) in type 2 diabetes (T2D) patients. Results Our PRSs aggregating 84–549 genetic variants (0.251 < correlation coefficients (r) < 0.272) had comparably stronger association with lipid variations than the typical PRSs derived based on the genome-wide significant variants (0.089 < r < 0.240). Our PRSs were robustly associated with their corresponding lipid levels (7.5 × 10− 103 < P < 1.3 × 10− 75) and 3-year lipid changes (1.4 × 10− 6 < P < 0.0130) which started to emerge in childhood and adolescence. With the adjustments for principal components (PCs), sex, age, and body mass index, there was an elevation of 5.3% in TC (β ± SE = 0.052 ± 0.002), 11.7% in TG (β ± SE = 0.111 ± 0.006), 5.8% in HDL-C (β ± SE = 0.057 ± 0.003), and 8.4% in LDL-C (β ± SE = 0.081 ± 0.004) per one standard deviation increase in the corresponding PRS. However, their predictive power was attenuated in T2D patients (0.183 < r < 0.231). When we included each PRS (for TC, TG, and LDL-C) in addition to the clinical factors and PCs, the AUC for dyslipidemia was significantly increased by 0.032–0.057 in the general population (7.5 × 10− 3 < P < 0.0400) and 0.029–0.069 in T2D patients (2.1 × 10− 10 < P < 0.0428). Moreover, the quintile of TC-related PRS was moderately associated with cIMT in adult women (β ± SE = 0.011 ± 0.005, Ptrend = 0.0182). Independent of conventional risk factors, the quintile of PRSs for TC [OR (95% CI) = 1.07 (1.03–1.11)], TG [OR (95% CI) = 1.05 (1.01–1.09)], and LDL-C [OR (95% CI) = 1.05 (1.01–1.09)] were significantly associated with increased risk of CHD in T2D patients (4.8 × 10− 4 < P < 0.0197). Further adjustment for baseline lipid drug use notably attenuated the CHD association. Conclusions The PRSs derived and validated here highlight the potential for early genomic screening and personalized risk assessment for cardiovascular disease.

ExPRSweb - An Online Repository with Polygenic Risk Scores for Common Health-related Exposures

“…The former two can be attributed to their underlying equations and related measurements, while the linear relationship between SBP and DBP is wellestablished [47][48][49] . Several of the other observed correlations are also well-documented often reflecting related disease etiologies [50][51][52] . Similar but more attenuated patterns were seen for the ExPRSs whose correlations ranged from -0.78 to 0.72 (Figure 3B).…”

“…The often lower pairwise correlations (e.g., r[PRSTC, PRSLDL] = 0.72 and r[PRSeGFR, PRScreatine] = -0.78) were expected because ExPRSs capture only a fraction of the exposure's variance (see diagonal of Figure 3C). The consistent patterns suggested that several ExPRSs can replicate correlations of measured exposures relatively well and thus might be suitable surrogates for exposures, especially for studies where measurements might not be feasible or likely be biased 50,53,54…”

“…As genetic proxies at the individual level, ExPRSs have been used in many applications, e.g., risk prediction and stratification [16][17][18] , predicting exposures 19 , instruments for Mendelian randomization analyses, or phenome-wide association studies (PheWASs) [20][21][22][23] . Including ExPRSs to prediction models could improve disease diagnosis, screening, therapeutic interventions, and precision medicine approaches.…”

“…Michigan Genomics Initiative (MGI) MGI CohortAdult participants aged between18 and 101 years at enrollment were recruited through the Michigan Medicine health system between 2012 and 2020. Participants have consented to allow research on both their biospecimens and EHR data, as well as linking their EHR data to national data sources such as medical and pharmaceutical claims data.…”

A randomized clinical trial of genetic testing and personalized risk counselling in patients with type 2 diabetes receiving integrated care -The genetic testing and patient empowerment (GEM) trial

“…In this proof-of-concept study, we only assessed genetic risk of complications based on 3 loci discovered in our own population with validation in a meta-analysis. Although the use of genome-wide polygenic risk scores [35] might be preferable, this is not possible due to the current limited knowledge of the genetic basis of diabetic complications. With identification of additional genetic variants, future studies could utilize polygenic risk scores to capture a greater proportion of the heritability of risk of diabetic complications.…”