The breadth and impact of the Global Lipids Genetics Consortium.

Circulating blood lipids, including LDL cholesterol, HDL cholesterol, and triglyceride, have been exemplary traits for genetic studies, given their heritability estimates of up to 50% [ 1 ] and availability for analysis because of their routine measurements as clinical biomarkers. A set of articles published in 2008 assessed the human genome for common DNA variation associated with levels of these lipid traits [ 2 – 4 ]. Together, these genome-wide association studies (GWASs) were hailed as a ‘treasure trove for lipoprotein biology’ in a Nature Genetics News and Views article as they collectively uncovered seven genomic loci newly associated with lipid levels [ 5 ]. These early genetic associations observed in thousands of individuals demonstrated the potential for uncovering new biology by performing unbiased scans of genetic variation across the genome. The investigators recognized that their studies were still limited by sample size in identifying genetic associations across the genome, resulting in the formation of the Global Lipids Genetics Consortium (GLGC). The GLGC is a worldwide collaboration of investigators dedicated to understanding the genetic etiology of quantitative blood lipid traits (Fig. 1 ). The GLGC has established itself as a central resource for lipid-related genetic research, having uncovered over 900 genomic loci to date significantly associated with lipid traits and making the summary statistics from each meta-analysis publicly available ( http://www.lipidgenetics.org/ ). By sharing these results, both members of the GLGC and the wider research community have leveraged this information (Fig. 2 ) for more in-depth analyses of lipid and lipoprotein biology and applied these resources toward research on related traits and diseases, such as cardiovascular and cardiometabolic disease, neurodegenerative and neuropsychiatric disease, and cancer. In this review, we will describe GLGC efforts (Table 1 ), detailing their contributions to our understanding of lipid traits and their broader impact on the research community. no caption available Countries with samples that have been included in published Global Lipids Genetics Consortium meta-analyses. Since 2009, the Global Lipids Genetics Consortium community has grown to include hundreds of groups, including individual labs, research centers, hospital biobanks, and population biobanks that span 48 countries. The map was generated using rworldmap R package (version 1.3-4). Citations across Global Lipids Genetics Consortium publications. Citation metrics were retrieved from Google Scholar on 31 October 2024. Summary of published summary statistics from the Global Lipids Genetics Consortium ADM AFR/AFR, admixed African/African; CSA, Central and South Asian; EAS, East Asian; EUR, European; EWAS, exome-wide association study; GWAS, genome-wide association study (i.e. common variant association study); HIS, Hispanic or admixed American; HRC, Haplotype Reference Consortium; MAF, minor allele frequency. Forty-one loci reached genome-wide significance in the East Asian population, including three novel loci. The values shown in the main table are the results from the meta-analysis of individuals from East Asian and European populations. The sample sizes correspond to the number of samples used in the analyses for total cholesterol.

In the upcoming iteration of GLGC analyses, both sample size and diversity will increase dramatically, alongside a higher proportion of GLGC contributors using improved imputation panels, such as the TOPMed reference panel, to enhance variant imputation accuracy across global populations [ 65 , 66 ]. However, a new opportunity to understand the genetics of lipid traits will come from extending from the traditional GWAS framework to a well powered genome-wide interaction study (GWIS) framework. This shift will not only allow us to continue assessing genetic effects on lipid traits but also elucidate how these effects are modified by exposures such as age and BMI. Previous GWISs have identified genomic loci whose effects on lipid traits are modified by physical activity, smoking, alcohol consumption, and age [ 58 ▪▪ , 67 , 68 ▪▪ ]. By identifying loci influenced by key risk factors for cardiovascular and cardiometabolic disease, the upcoming GLGC interaction results are expected to provide deeper insights into lipid and lipoprotein biology and have more clinically relevant implications. There will also be a continued and deliberate effort to recruit new diverse research groups into the GLGC community, with a particular focus on collaborating with scientists and datasets representing South Asian, South American, Middle Eastern, and African populations. These population groups have been historically underrepresented in previous GLGC research efforts, so we are making concerted efforts to grow the GLGC community for a true, global, collaboration.

GWASs are a core method used by the GLGC to identify DNA variation – referred to as single-nucleotide polymorphisms (SNPs) – associated with blood lipid levels. Prior to the first GLGC GWAS, 19 loci had been identified as being associated with lipid traits [ 2 , 3 , 6 – 8 ]. Through the collaborative nature of the GLGC and methodological improvements described in the current section, this number has increased nearly 50-fold, with now over 923 associated loci (Fig. 3 ). Increases in sample size and diversity across Global Lipids Genetics Consortium publications that released association summary statistics. (a) Stacked bar graphs show the sample size for each population group across each publication. The dashed lines and annotated values reflect the maximum discovery sample size for each publication. (b) Stacked bar graphs show the proportion of each population group used for each publication. The Liu et al. (2017) and Lu et al. (2017) publications and summary statistics were released in tandem. Kanoni et al. (2022) is not shown because the same individuals were analyzed by Graham et al. (2021). Ramdas et al. (2022) is not shown because summary statistics were not released with this publication. The first GLGC publication was a GWAS meta-analysis of ∼20 000 individuals that identified 30 lipid-associated loci [ 9 ]. Many loci contained known lipoprotein metabolism genes causal for dyslipoproteinemia and Mendelian lipid disorders increasing confidence in the findings, and 11 loci had never been previously linked to lipid and lipoprotein metabolism using human genetic data. This initial GLGC publication emphasized how larger sample sizes increase statistical power, allowing for the identification of novel associations. The second GLGC GWAS meta-analysis expanded the consortium's efforts by including total cholesterol as a lipid outcome and analyzing ∼100 000 individuals. Ninety-five lipid-associated loci were identified, 59 of which were novel [ 10 ]. A key advancement in this analysis was the inclusion of non-European-like populations to determine whether genome-wide significant SNPs identified in European-like populations could be replicated in East Asian-like, South Asian-like, and African-like population groups [ 10 ]. The majority of loci had consistent directions of effect without evidence of heterogeneity but had limited power to detect significant variants specific to non-European-like groups. The next GLGC publication demonstrated the benefits of larger, globally diverse populations for discovery. In over 188 000 individuals from European-like populations and almost 8000 individuals from non-European-like populations, 157 loci were significantly associated with lipid levels, including 62 novel loci, 30 of which had never been previously implicated in lipid and lipoprotein biology [ 11 ]. Of the 157 loci identified, 58 had associations across multiple lipid traits. For example, 4 loci were associated with total cholesterol, LDL cholesterol, HDL cholesterol, and triglyceride levels, while 36 loci were associated with total cholesterol and LDL cholesterol levels [ 11 ]. Recent studies have similarly observed loci with both shared and unique associations across dyslipoproteinemia phenotypes [ 12 ▪▪ ]. This emphasizes the complexity of lipid metabolism and highlights the therapeutic potential of targeting loci that influence multiple atherogenic lipid traits. Subsequent GLGC publications shifted focus from common noncoding variants to low-frequency coding variants using an exome-based genotyping array in diverse global populations [ 13 ], and assessed their impact on coronary artery disease (CAD) risk in East Asians [ 14 ]. In parallel, GLGC collaborators assessed the association of low-frequency variants and rare variant burden with lipid traits in European-like populations [ 15 ], and identified genetic variants associated with lipid traits in almost 300 000 individuals from a diverse United States cohort [ 16 ]. Together, these efforts revealed that there are lipid-associated rare and low-frequency coding variants that are population-specific, emphasizing the importance of studying multiple population groups. The most recent GLGC GWAS meta-analysis significantly increased sample diversity. Most GWAS data for lipid traits and other noncommunicable diseases have been generally derived from European-like populations [ 17 ▪ , 18 ]; few studies had included ancestrally diverse populations in their main analyses [ 19 , 20 ▪▪ ]. Due to differences in allele frequency and linkage disequilibrium among global ancestral groups, genetic variation contributing towards lipid levels can be missed when studying genetic variation in a single ancestral population. In the largest lipid-related GWAS to date, Graham et al. [ 21 ▪▪ ] conducted multipopulation meta-analyses in ∼1.65 million individuals, including 350 000 individuals from East Asian-like, admixed African or African-like, Hispanic-like, and South Asian-like populations. This effort, which included non-HDL cholesterol – a key biomarker for atherosclerotic cardiovascular disease – identified 923 loci associated with at least one of five lipid traits across population groups, 237 of which were novel. Although larger and more diverse sample sizes were major contributors to the substantial improvement in locus discovery, this analysis also benefited from the 1000Genomes and Haplotype Reference Consortium (HRC) imputation reference panels, in contrast to previous analyses that used the HapMap panel. Newer reference panels feature denser genotype data from more genetically diverse individuals, allowing for improved variant imputation accuracy, particularly across diverse populations.

Summary statistics from each GLGC meta-analysis, including effect estimates and P values for the millions of assayed and imputed SNPs, are publicly available for download from the GLGC website ( http://www.lipidgenetics.org/#data-downloads-title ) or the GWAS Catalog website (Table 1 ). This accessibility has supported further research on lipids and related phenotypes, particularly in the development of polygenic scores (PGSs), a genetic metric reflecting the weighted sum of alleles and their associated effects on a particular phenotype. The potential of PGSs in a clinical setting for risk prediction, diagnosis, prognosis, and disease management has driven extensive efforts to methodologically improve PGS development and calculation [ 32 , 33 ▪▪ , 34 ▪ ]. PGSs are constructed using association summary statistics for SNP selection and effect size weighting [ 35 ], making publicly available summary statistics essential for PGS development. The GLGC lipid summary statistics have been widely utilized to develop lipid PGSs [ 36 – 39 ] and polygenic risk scores for related conditions, such as CAD [ 40 ▪▪ ]. An important limitation of PGSs is their limited portability across population groups and that humans cannot always be categorized into a single population group; PGSs typically show lower predictive power in non-European-like populations due to SNP effect sizes predominantly derived from European-like populations [ 41 ]. The work by Graham et al. [ 21 ▪▪ ] marks a significant advancement in PGS development, as they found that incorporating summary statistics from multiple ancestral population groups – facilitated by global collaboration – resulted in PGSs being more predictive across population groups compared with using single-population summary statistics for LDL cholesterol. For example, a European-based PGS for LDL cholesterol applied to admixed African Americans from the MVP cohort had an adjusted R 2 of 0.04, which increased to 0.10 with a PGS derived from admixed African American statistics, and peaked at 0.16 when using a PGS that combined summary statistics from multiple populations (see Supplemental Table 17 from Graham et al. ). Although the transferability of this multipopulation improvement to other complex traits requires validation, recent methods to enhance PGS performance increasingly leverage diverse summary statistics for improved score accuracy [ 42 ▪ , 43 ▪ ]. Some of these methods have even utilized GLGC summary statistics to test and validate their novel approaches [ 44 ▪ ]. This underscores the importance of the GLGC not only releasing summary statistics but also ensuring that analyzed samples are representative of global populations.

Although the benefits of the GLGC and its research efforts related to lipid biology are well recognized, the utility of GLGC summary statistics extends to nonlipid phenotypes. The impact of GLGC on cardiovascular disease research is particularly evident in risk prediction and disease mechanism studies. Contemporary polygenic risk scores integrating both CAD and related risk factor data, including blood lipid levels, have shown marked improvements in prediction accuracy, especially when leveraging multiancestry data. For instance, GPS Mult , a polygenic risk score that incorporates data from over 269 000 CAD cases and 1.18 million controls across five population groups and 10 disease-related risk factors, can identify 20% of UK Biobank participants with three-fold increased CAD risk compared with individuals with a GPS Mult in the middle quantile of the population. This is a substantial improvement over earlier scores that only used CAD summary statistics and identified only 8.3% of individuals at this risk level [ 40 ▪▪ ]. Beyond risk prediction, GLGC summary statistics help advance our understanding of cardiovascular disease mechanisms, as demonstrated in a study of abdominal aortic aneurysm (AAA) where analysis of newly identified AAA-associated variants against GLGC results revealed the role of lipids and lipoproteins in AAA pathogenesis, with evidence to support PCSK9 inhibition as a possible therapeutic strategy [ 45 ]. The utility of GLGC summary statistics extends further through Mendelian randomization, a genetic approach to assess putative causal relationships of an exposure on an outcome by leveraging genetic variation associated with the exposure of interest [ 46 ▪▪ ]. In the commonly used two-sample Mendelian randomization framework, summary statistics for both the exposure and outcome are required. Using the most recent summary statistics generated by Graham et al. , Mendelian randomization studies have provided evidence of a causal relationship for LDL cholesterol and triglycerides with amyotrophic lateral sclerosis (ALS) [ 47 ▪ , 48 ▪ ]. There is also evidence suggesting causal relationships between LDL cholesterol and esophageal cancer [ 49 ], gallstone disease [ 50 ▪▪ ], interstitial lung disease [ 51 ], bone mineral density [ 52 ▪ ], osteoporosis [ 53 ▪ ], and small vessel stroke [ 54 ▪▪ ]. Many of these findings carry important implications for repurposing lipid-lowering medications for nonlipid conditions, particularly for those that found lower LDL cholesterol levels were linked to decreased disease risk. Causal associations have also been demonstrated between triglyceride levels and increased risk for endometriosis [ 55 ▪ ] as well as increased risk for acute pancreatitis [ 56 ▪ ]. GLGC summary statistics have also been used in methodological Mendelian randomization papers to benchmark methods, such as IVW, MR-PRESSO, Egger, MR-RAPS, and MRCUE [ 57 ▪▪ ], and show how the framework to detect horizontal pleiotropy in Mendelian randomization studies can be used to detect gene–environment interactions [ 58 ▪▪ ]. Beyond Mendelian randomization, other groups have leveraged GLGC summary statistics to demonstrate shared individual genetic loci between lipid traits and major depressive disorder [ 59 ▪▪ ], posttraumatic stress disorder [ 60 ], and breast cancer [ 61 ]. This highlights the potential for shared biological mechanisms and drug repurposing. Leveraging shared genetic loci, other research groups have also been able to further define mechanisms for type 2 diabetes through common biological mechanisms [ 62 ▪▪ , 63 ]. By leveraging the lead lipid-associated SNPs reported by Graham et al. and lead SNPs associated with dozens of other cardiometabolic traits, eight pathophysiological processes were defined for type 2 diabetes, four of which – metabolic syndrome, obesity, lipodystrophy, and liver and lipid metabolism – were characterized by abnormal lipid levels [ 62 ▪▪ ]. In a broader use case, the Impact of Genomic Variation on Function (IGVF) consortium used the summary statistics from Graham et al. in conjunction with summary statistics from other large-scale genetic analyses and functional datasets to create cell-specific maps of how genetic variants can alter gene expression, protein activity, and impact interaction networks. These maps and related information will be assembled into an open-access resource for the research community, facilitating the advancement of research on the genomic impact on biology and disease across different cell types and populations [ 64 ]. Collectively, these examples illustrate the broad, interdisciplinary impact of GLGC resources beyond lipid biology.

The GLGC is the leading source of association summary statistics for lipid traits and is a key reference for genetic variation influencing lipid levels. By identifying over 900 lipid-associated loci and improving fine-mapping and PGS efforts, we have dramatically contributed to advancing our understanding of lipid biology. Like other global consortia focused on unraveling the genetic basis of various traits and diseases [ 69 ▪▪ , 70 , 71 ▪▪ , 72 , 73 ], the success of the GLGC is due to the worldwide collaborative efforts of researchers who are invested in understanding the genetic cause of quantitative lipid traits.

Each GLGC publication described the biological relevance of newly identified loci through approaches such as fine-mapping and/or functional validation using mouse or cellular models. Fine-mapping with colocalization assessments revealed that many of the novel loci were hepatic expression quantitative trait loci (eQTLs) or were in strong linkage disequilibrium with known eQTLs [ 9 – 11 ], whereas mouse models were key in validating and elucidating the roles of novel loci in lipoprotein metabolism and regulation [ 10 , 22 ]. Importantly, many loci identified in the early European-centric GLGC studies have been fine-mapped in other ancestral groups, demonstrating the robustness of these early signals across diverse populations [ 23 – 25 ]. Graham et al. described a substantial improvement in fine-mapping efforts by leveraging multipopulation meta-analyzed results. They showed that multipopulation fine-mapping reduces the credible set of causal variants and does so more quickly than analyses that consider only single population groups. As done for other traits [ 26 ▪ ], integrating GWAS summary statistics generated from multiple population groups can refine the list of likely causal variants because of differences in linkage disequilibrium patterns and variant effect sizes across ancestral groups. This multipopulation consideration has been a central focus in the recent improvement of methods for fine-mapping [ 27 , 28 ]. Two additional GLGC studies thoroughly explored the biological significance of lipid-associated loci identified by Graham et al. Using the data that was meta-analyzed by Graham et al. , one study leveraged concordant functional evidence, including expression data for dozens of tissues, areas of open chromatin, transcription factor-binding sites, and epigenetic marks to identify lipid regulatory mechanisms for noncoding SNPs, prioritizing candidates such as CREBRF and RRBP1 [ 29 ▪▪ ]. The other study used multiple gene-prediction tools to prioritize 466 genes with potential causal impacts on lipid biology and identified 21 loci on the X chromosome that were newly associated with lipid traits, and that 2–5% of autosomal loci have sex-specific effects on lipids, highlighting the importance of possible sex differences on the expression of lipid-associated alleles [ 30 ▪▪ ]. Beyond direct GLGC efforts, independent research groups have leveraged GLGC results to uncover novel insights into lipid biology. For example, Votava et al. recently used loci described by Graham et al. in a coexpression network analysis to identify network modules enriched with cholesterol biosynthesis genes. They then selected genes with unknown functions for functional validation in mice and successfully identified ALDOC as a regulator for de novo cholesterol biosynthesis [ 31 ▪▪ ].

We thank all past and present members of the GLGC community for their continued support and effort in contributing to each iteration of GLGC analyses. We also thank Ida Surakka for providing the sample size values in Table 1 for the Kanoni et al. (2022) publication. Funding for the GLGC (G.M.P. and P.N.) is provided by the National Institutes of Health (NIH) (R01HL127564). J.S.D. is supported by the National Heart, Lung, and Blood Institute of the NIH (K99HL175031). J.S.D reports spousal employment at Biogen Inc., unrelated to the present work. P.N. reports research grants from Allelica, Amgen, Apple, Boston Scientific, Genentech/Roche, and Novartis, personal fees from Allelica, Apple, AstraZeneca, Blackstone Life Sciences, Creative Education Concepts, CRISPR Therapeutics, Eli Lilly & Co, Esperion Therapeutics, Foresite Capital, Foresite Labs, Genentech/Roche, GV, HeartFlow, Magnet Biomedicine, Merck, Novartis, TenSixteen Bio, and Tourmaline Bio, equity in Bolt, Candela, Mercury, MyOme, Parameter Health, Preciseli, and TenSixteen Bio, and spousal employment at Vertex Pharmaceuticals, all unrelated to the present work.