Skip to main content
Download PDF

Similarity and diversity of genetic architecture for complex traits between East Asian and European populations

BMC Genomics volume 24, Article number: 314 (2023) Cite this article

Abstract

Background

Genome-wide association studies have detected a large number of single-nucleotide polymorphisms (SNPs) associated with complex traits in diverse ancestral groups. However, the trans-ethnic similarity and diversity of genetic architecture is not well understood currently.

Results

By leveraging summary statistics of 37 traits from East Asian (Nmax=254,373) or European (Nmax=693,529) populations, we first evaluated the trans-ethnic genetic correlation (ρg) and found substantial evidence of shared genetic overlap underlying these traits between the two populations, with \({\widehat{\rho }}_{g}\) ranging from 0.53 (se = 0.11) for adult-onset asthma to 0.98 (se = 0.17) for hemoglobin A1c. However, 88.9% of the genetic correlation estimates were significantly less than one, indicating potential heterogeneity in genetic effect across populations. We next identified common associated SNPs using the conjunction conditional false discovery rate method and observed 21.7% of trait-associated SNPs can be identified simultaneously in both populations. Among these shared associated SNPs, 20.8% showed heterogeneous influence on traits between the two ancestral populations. Moreover, we demonstrated that population-common associated SNPs often exhibited more consistent linkage disequilibrium and allele frequency pattern across ancestral groups compared to population-specific or null ones. We also revealed population-specific associated SNPs were much likely to undergo natural selection compared to population-common associated SNPs.

Conclusions

Our study provides an in-depth understanding of similarity and diversity regarding genetic architecture for complex traits across diverse populations, and can assist in trans-ethnic association analysis, genetic risk prediction, and causal variant fine mapping.

Peer Review reports

Background

Over the past few years, large-scale genome-wide association studies (GWASs) have convincingly detected a large number of genetic loci associated with a series of complex traits by casting single-nucleotide polymorphisms (SNPs) across the entire genome [1], generating novel biological knowledge and renovating diagnostic and treatment tools for diseases [2, 3]. However, existing GWASs have been heavily biased towards European (EUR) individuals, with 66.4% (= 7,377/11,113, until 2023-4-25) of studies conducted in EUR but only a few in other populations [4]. As medical genomics studies have become increasingly large and diverse, acquiring insights into similarity and diversity of trait-associated SNPs among distinct populations and consequently the transferability of disease genetic risk is imperative in clinical translation [5].

It has been revealed that many trait-associated SNPs identified in GWASs of EUR ancestry can be replicated in other ethnic groups [2, 6,7,8,9,10], in the sense that they show significant association with consistent direction of genetic effects in non-EUR individuals, indicating complex traits enjoy common genetic components across diverse continental populations. For example, the trans-ethnic genetic correlation is 0.79 (se = 0.04) for ulcerative colitis and 0.76 (se = 0.04) for Crohn’s disease [11], 0.46 (se = 0.06) for rheumatoid arthritis (RA) [12], 0.33 (se = 0.03) for major depressive disorder [13], and 0.39 (se = 0.15) for attention-deficit hyperactivity disorder between the EUR and East Asian (EAS) populations. However, these estimated trans-ethnic genetic correlation are in general significantly less than one in spite of being larger than zero [12, 14], which meanwhile suggests ancestral diversity.

Indeed, ancestry-relevant heterogeneity regarding varying allele frequency and linkage disequilibrium (LD) patterns is observed for some causal variants such that a significant association detected in one population is not readily identified in others [5, 15, 16]. Notable examples include a nonsense variant in TBC1D4, which confers muscle insulin resistance and increases risk for type 2 diabetes (T2D) and is common in Greenland but rare or absent in other populations [17], several common EAS-specific coding variants that influence blood lipids by exerting a protective effect against alcoholism [18, 19], and two loci associated with major depression that are more common in the Chinese population than EUR (i.e., 45% vs. 2% for rs12415800, and 28% vs. 6% for rs35936514) [20, 21]. As another example, multiple associated loci located within PCSK9, APOA, APOC, and ABCA1 all play key roles in lipid genetics in both the EUR and African (AFR) Americans, yet the precise alleles within each locus differ substantially between the two populations, supporting the perspective that lipid-associated SNPs are largely shared across ancestral groups, but the allelic structure within a locus may be shaped by population history and thus exhibits considerable heterogeneity [22, 23].

Prior work has investigated the replicability of GWAS discoveries for some particular traits, displaying the similarity and diversity of associated SNPs across ancestral groups [7, 12, 24, 25]. However, those studies primarily focused on limited traits or a small set of significant loci [7, 24]; it is unknown whether their conclusions can be generalized to other traits or to genome-wide SNPs given the polygenic nature of many complex phenotypes. In addition, some of previous studies focused mainly on the trans-ethnic genetic correlation [12, 25, 26], which however only quantifies an overall genetic similarity across the whole genome and cannot characterize detailed association pattern for individual SNPs. A comprehensive genome-wide assessment of trans-ethnic similarity and diversity of genetic components for traits is challenging because our knowledge of genetic architecture within each population is not fully understood.

To fill in this critical knowledge gap mentioned above, here we obtained GWAS summary statistics of 37 complex traits from the EAS and EUR populations to perform a complete comparison of genetic similarity and diversity across the two populations. We first evaluated the trans-ethnic genetic correlation to quantify the extent of common genetic basis to which these traits shared [12]. Then, we identified population-specific and population-common trait-associated SNPs [27, 28]. Afterwards, we conducted the marginal genetic correlation analysis and heterogeneity test for these associated SNPs [25]. Finally, we evaluated how the LD and minor allele frequency (MAF) patterns varied among various types of SNPs and studied whether the genetic differentiation between ancestry groups can be explained via natural selection [7, 24]. We also assessed the genetic influence of associated SNPs on traits across populations by calculating a genetic risk score (GRS) [24, 29].

Results

Overview of the used statistical methods

We here give a brief overview of some important statistical methods employed in our analyses and showed more methodological descriptions in the Materials and Methods Section. Technical details of all used methods could be found in respective original papers. We here analyzed a total of 37 complex traits (10 binary and 27 continuous) using summary statistics obtained from EAS-only or EUR-only GWASs (Table 1).

We first employed linkage disequilibrium score regression (LDSC) to estimate SNP-based heritability for every trait in each population [30]. Next, to evaluate genetic similarity and diversity of these traits across populations, we calculated the global trans-ethnic genetic correlation (ρg) via popcorn [12]. We identified common trait-associated SNPs via the conditional false discovery rate (cFDR) and conjunction conditional false discovery rate (ccFDR) methods in the two populations [27, 28].

Relying on cFDR and ccFDR, for each trait we divided all analyzed SNPs into four incompatible groups. For every trait we estimated the marginal genetic correlation (rm) of SNP effect sizes in each of the four groups using MAGIC [25]. For common trait-associated SNPs, we examined the heterogeneity in genetic effect on the trait across EAS and EUR populations via Cochran’s Q test.

Then, we obtained the two LD scores and MAF for every SNP in the four groups based on genotypes available from EAS or EUR individuals in the 1000 Genomes Project, and calculated their coefficient of variation of LD scores (LDCV) and coefficient of variation of MAF (MAFCV) across the populations to investigate whether there exist different patterns of LD and MAF for trait-associated SNPs compared to those null ones.

Finally, to further demonstrate the direction of genetic differentiation, for each trait we conducted a GRS analysis [29]. We also investigated whether the sample size difference could influence our findings with regards to genetic similarity and diversity of traits between the EAS and EUR populations.

Heritability and trans-ethnic genetic correlation

Estimated heritability

We first present the estimated SNP-based heritability (\({\widehat{h}}^{2}\)) of these traits. It was shown some traits (e.g., height and schizophrenia [SCZ]) were more heritable, with a large heritability estimate; but other (e.g., ischemic stroke [IS] and atopic dermatitis [AD]) exhibited low heritability (Table 1). Although the estimates of heritability were highly correlated between the EAS and EUR populations (Pearson’s correlation = 0.738, P = 1.83 × 10− 7), there still existed an obvious trans-ethnic distinction in heritability. For example, platelet count (PLT) showed the maximal difference in estimated heritability, with \({\widehat{h}}^{2}\)=1.1% (se = 1.4%) in EAS but 20.1% (se = 1.8%) in the EUR population; in contrast, the heritability of atrial fibrillation (AF) was much larger in the EAS population compared to that in the EUR population (9.5% (se = 2.7%) vs. 2.0% (se = 0.3%)). By carrying out an approximate normal test, we observed that 70.3% (= 26/37) of heritability estimates were significantly different across the two populations (FDR < 0.05) (Table 1), largely reflecting diversity in polygenic genetic architecture across traits and ancestral groups.

We did not find a substantial correlation between sample size and heritability in both populations (Pearson’s correlation = 0.112 with P = 0.508 in the EAS population, and Pearson’s correlation=-0.025 with P = 0.884 in the EUR population). However, as expected, we indeed discovered suggestive evidence that larger sample size can lead to more accurate estimate of heritability, with Pearson’s correlation=-0.266 in the EAS population and − 0.102 in the EUR population between sample size and standard error of heritability. We further used the coefficient of variation of sample sizes (NCV) to measure the difference of sample sizes, but did not observe a significant correlation between NCV and the coefficient of variation of heritability (Pearson’s correlation = 0.029, P = 0.867).

Table 1 Summary information of complex traits analyzed in the present study
Full size table

Estimated trans-ethnic genetic correlation

The trans-ethnic genetic correlation estimates of six traits (i.e., AF, AD, breast cancer [BRC], high-density lipoprotein cholesterol [HDL], age at menarche [AAM], and triglyceride [TG]) were larger than one and thus not included in the following descriptions. All the traits exhibited positive trans-ethnic genetic correlation between the EAS and EUR populations (Table 1), with \({\widehat{\rho }}_{g}\) ranging from 0.53 (se = 0.11) for adult-onset asthma (AOA) to 0.98 (se = 0.17) for hemoglobin A1c (HbA1c).

All the trans-ethnic genetic correlation estimates were significantly larger than zero (H0: ρg = 0), but 61.3% (= 19/31) were less than one (H0: ρg = 1) (FDR < 0.05), indicating potential heterogeneity in genetic effects across populations. The average of \({\widehat{\rho }}_{g}\) across traits was 0.79, and the average of \({\widehat{\rho }}_{g}\) for binary phenotypes was 0.74 (se = 0.15), which was slightly (although not significantly, P = 0.217, possibly due to limited number of binary phenotypes under analysis) smaller than that for continuous ones (the average of \({\widehat{\rho }}_{g}\)=0.81 (se = 0.09)).

Comparison for trans-ethnic genetic correlation, cross-trait trans-ethnic genetic correlation, and cross-trait genetic correlation

Furthermore, as an empirical comparison, we calculated cross-trait trans-ethnic genetic correlation and anticipated that the cross-trait correlation should be on average much smaller than that for the same traits in the two populations because of greatly distinct genetic foundations. As expected, it was found the same traits generally showed much greater genetic similarity compared to distinct traits although some of them, such as three lipids including low-density lipoprotein cholesterol [LDL], total cholesterol [TC], and TG, as well as three blood pressures including diastolic blood pressure [DBP], systolic blood pressure [SBP], and pulse pressure [PP], also exhibited relatively high cross-trait trans-ethnic genetic correlation (Fig. 1A and Table S1). For instance, the cross-trait trans-ethnic genetic correlation ranged from − 0.62 to 0.79, approximately 47.0% were negative, with an average of only 0.02 (Fig. 1B).

We also included the cross-trait genetic correlation in the same population as a reference. Totally, it was seen that the cross-trait genetic correlations in the same population were comparable to the cross-trait trans-ethnic genetic correlations across different populations but were much lower than the trans-ethnic genetic correlation of the same trait in diverse populations (Fig. 1B and Table S2).

Fig. 1
figure 1

(A) Estimated trans-ethnic genetic correlation of 37 traits between the EAS and EUR populations. Elements in the diagonal represent the trans-ethnic genetic correlation for the same traits in the two populations, while elements in the off-diagonal represent the trans-ethnic genetic correlation for two different traits in the two populations. (B) Comparison of the estimated genetic correlation for the trait in the same population, two different traits in the same population, and two different traits in the EAS and EUR populations

Full size image

We did not discover a substantial connection between NCV and \({\widehat{\rho }}_{g}\) (Pearson’s correlation=-0.148, P = 0.426), implying that the imbalanced sample sizes would not significantly affect the estimate of trans-ethnic genetic correlation. However, we observed that the difference of sample sizes in a pair of traits could affect the significance of \({\widehat{\rho }}_{g}\). More specifically, \({\widehat{\rho }}_{g}\) trended to be statistically different from one if larger imbalance of sample sizes was present (NCV = 0.43 (se = 0.25) for the 17 non-significant traits vs. 0.69 (se = 0.27) for the 14 significant ones, P = 0.010 in terms of the approximate normal test).

Fig. 2
figure 2

(A) Relationship between n1 and f11 + f10; (B) Relationship between n2 and f11 + f01. Here, f10 and f01 are the number of identified associated SNPs for the trait in the EAS or EUR population, respectively, and f11 is the number of shared trait-associated SNPs; n1 and n2 stand for the effective sample size for traits in the EAS or EUR population; r stands for the Pearson’s correlation, with P the corresponding P value; k on the x-axis means 1,000 for identified trait-associated SNPs and 10k on the y-axis means 10,000 for sample size of traits

Full size image

Associated SNPs for each trait in the EAS or EUR population

To discover trait-associated SNPs shared across the EAS and EUR populations, we carried out the cFDR analysis using a set of distinct SNPs [27, 28], with the number of significant associations displayed in Table 2. In general, more associated SNPs (cFDR < 0.05) were discovered for most of the analyzed traits (94.6%=35/37, except for telomere length [TL] and HbA1c) in the EUR population (denoted by f01 + f11 in Table 2) compared to the EAS population (denoted by f10 + f11 in Table 2). This is a direct consequence of higher power due to larger sample sizes of EUR traits (Fig. 2A-B) and implies that additional trait-associated loci would be detected if increasing samples especially in the EAS population. Only a few of SNPs were detected for some traits such as AD and IS, but much more associated SNPs were identified for others (e.g., BMI and height), partly indicating the polygenic nature of these traits [3, 31].

Table 2 Number of associated SNPs (cFDR/ccFDR < 0.05) for traits in the EAS and EUR populations
Full size table

Shared associated SNPs of traits between the EAS and EUR populations

Furthermore, we identified many significant SNPs shared across the EAS and EUR populations (ccFDR < 0.05, denoted by f11 in Table 2). On average, 21.7% of associated SNPs were discovered simultaneously for the same traits in both populations. The LRT implemented under the four-group model framework also offered considerably strong evidence supporting common genetic foundation underlying each trait between the two population. The proportion of shared associated SNPs varied substantially across these traits, ranging from only 3.0% for BRC and 3.3% for DBP to 49.4% for IS and 51.8% for TL (Table 2).

On average, approximately 67.1% of trait-associated SNPs in the EUR population were also detected to be significant in the EAS population, but only 26.8% of trait-associated SNPs in the EAS population were replicated to be significant in the EUR population. Among these shared significant SNPs, 44.1% and 77.0% showed genome-wide significance (P < 5 × 10− 8) in the EAS or EUR population (Table S3). We observed that the number of population-common associated loci (i.e., f11) was negatively correlated with NCV (Pearson’s correlation=-0.095, with a marginally significant P value of 0.057), indicating that smaller difference of sample sizes in a pair of traits (e.g., increasing the sample size of traits in the EAS population) might lead to more discoveries of shared SNPs.

Characteristics of shared associated SNPs

Similarity and heterogeneity of associated SNPs across populations

It needs to highlight that we can divide all analyzed SNPs into four groups based on the associations identified above: (i) null SNPs; (ii) EAS-specific associated SNPs; (iii) EUR-specific associated SNPs; (iv) population-common associated SNPs. For these SNPs, it is shown that population-common associated SNPs often exhibited a maximal positive correlation in effect sizes compared to population-specific associated SNPs and null ones (Fig. 3A and Table S4). For example, the marginal genetic correlation of effect sizes for shared trait-associated SNPs was 0.92 (se = 0.04) for TL, followed by 0.81 (se = 0.02) for white blood cell count (WBC) and 0.80 (se = 0.03) for NEUT, with an average estimate of 0.61 (se = 0.04) across these traits, which was much higher than that for EAS-specific (\({\widehat{r}}_{m}\)=0.29 (se = 0.05)) or EUR-specific (\({\widehat{r}}_{m}\)=0.33 (se = 0.05)) trait-associated SNPs or null SNPs (\({\widehat{r}}_{m}\)=0.09 (se = 0.02)).

Fig. 3
figure 3

(A) Estimated marginal genetic correlation of effect sizes for SNPs in the four incompatible groups; (B) Relationship between proportions of genetic effect heterogeneity of shared associated SNPs and the cross-population marginal genetic correlations of SNP effect sizes; × means the median

Full size image

We found that the effect size slopes of population-common associated SNPs of eight traits (i.e., AF, AD, BRC, TL, SBP, PP, age at natural non-surgical menopause [ANM], and basophil count [BASO]) were not significantly different from one (FDR > 0.05) (Supplement File and Figure S1), suggesting that effect sizes of shared associated SNPs are considerably consistent in magnitude for these traits; whereas great deviations of the estimated slopes from one were observed for the remaining 29 traits (FDR < 0.05) (e.g., HbA1c, T2D, estimated glomerular filtration rate (eGFR), SCZ and AAM), indicating substantial trans-ethnic diversity of SNP effect sizes for these traits. In addition, we did not detect substantial evidence supporting the influence of imbalanced sample sizes (i.e., NCV) on the estimated slope (P = 0.053).

We next performed a heterogeneity test using Cochran’s Q test and again focused only on population-common associated SNPs because the population-specific associated SNPs can be naturally considered to be heterogeneous (i.e., theoretically, their SNP effect sizes were non-zero in one population but zero in another population). On average, approximately 20.8% of the common trait-associated SNPs, ranging from 7.1% for TL to 52.9% for AAM, showed heterogeneous genetic effect on traits between the two ancestral populations after Bonferroni’s correction for multiple comparisons (Table S5). We here used Bonferroni’s method to take the LD among local SNPs into account as it was much more stringent compared to FDR. The high heterogeneity in SNP effect sizes for AAM was in agreement with a prior finding that AAM-associated SNPs often exhibited distinct effect sizes across populations [32]. It can be expected that greater proportion of shared trait-associated SNPs having heterogeneous effects would lead to weaker trans-ethic marginal genetic correlation (Pearson’s correlation=-0.48, P = 2.98 × 10− 3) (Fig. 3B). However, we found little evidence supporting the influence of the sample size difference in a pair of traits (i.e., NCV) on the proportion of population-common SNPs with heterogeneous effects (P = 0.771).

Difference in LD, MAF, and Fst for trait-associated SNPs across populations

First, we observed that, except for three lipid traits (i.e., TL, HDL, and LDL), all the remaining traits showed substantial different variations in LD for SNPs in various groups between the EAS and EUR populations (FDR < 0.05) (Fig. 4A and Figure S2). The average coefficient of variation of LD scores (LDCV) for population-common SNPs was smaller compared to that for null SNPs (0.22 (se = 0.02) vs. 0.30 (se = 0.02), P = 1.04 × 10− 17), for these population-specific loci in EAS (0.22 (se = 0.02) vs. 0.28 (se = 0.04), P = 1.82 × 10− 9) or EUR (0.22 (se = 0.02) vs. 0.28 (se = 0.03), P = 1.52 × 10− 12).

Fig. 4
figure 4

(A) Distribution for the average LDCV across all analyzed traits in the four groups of SNPs; (B) Distribution for the average MAFCV across all analyzed traits in the four groups of SNPs; (C) Distribution for the average Fst across all analyzed traits in the four groups of SNPs; × means the median

Full size image

On the other hand, we observed that, except for three traits (i.e., HDL, TG, and ANM), all the remaining traits showed substantially different MAFCV for SNPs in distinct groups between the EAS and EUR populations (FDR < 0.05) (Fig. 4B and Figure S3). On average, the MAFCV for population-common loci were much smaller compared to that for null SNPs (0.33 (se = 0.07 vs. 0.46 (se = 0.10), P = 1.54 × 10− 12), for these population-specific loci in EAS (0.33 (se = 0.07) vs. 0.43 (se = 0.11), P = 7.30 × 10− 8) or EUR (0.33 (se = 0.07) vs. 0.40 (se = 0.07), P = 2.22 × 10− 8).

We found that SNPs suffered from natural selection for more than half of traits (62.2%=23/37) (FDR < 0.05) (Figure S4). For all analyzed traits, we observed that population-common loci tended to have a smaller mean Fst compared with population-specific associated SNPs in EAS (0.056 (se = 0.01) vs. 0.062 (se = 0.02), P = 0.032) or EUR (0.056 (se = 0.01) vs. 0.063 (se = 0.006), P = 6.77 × 10− 4), and they also showed a lower mean Fst relative to null ones (0.056 (se = 0.01) vs. 0.060 (se = 0.001), with a marginally significant P = 0.086) (Fig. 4C). Furthermore, we did not observe significant correlation between Fst and LDCV (P = 0.781) as well as between Fst and MAFCV (P = 0.602) (Figure S5), indicating the between-population diversity of LD and MAF patterns is not possibly confounded by the differentiation in Fst.

Overall and partial GRS of trait-associated SNPs

First, we considered the GRS calculated from all associated SNPs (i.e., f10 + f11 for EAS or f01 + f11 for EUR) as an overall measurement of genetic effect on a given trait in each population. Among these, most of the traits had a substantially different GRS between the two populations (Figure S6). For example, six traits (i.e., RA, T2D, childhood-onset asthma [COA], AOA, BRC and prostate cancer [PCA]) showed a higher overall genetic effect on EAS, while some (i.e., SCZ, AF, AD and IS) displayed a larger overall genetic effect on EUR.

We further generated the GRS computed with only population-common associated SNPs (i.e., f11) or only population-specific associated SNPs (i.e., f10 for EAS or f01 for EUR), respectively. We viewed these new GRSs to quantify a measurement of partial genetic effect because only part of associated SNPs is employed. Interestingly, 16 traits (i.e., RA, T2D, AD, PCA, DBP, SBP, LDL, TC, TG, eGFR, ANM, mean corpuscular hemoglobin [MCV], hematocrit [HCT], monocyte count [MONO], eosinophil count [EO] and WBC) had consistent partial genetic effect compared to the overall one. Specifically, 15 traits (e.g., AF, COA, AOA, BRC, BMI, height, PP, HDL, HbA1c, RBC, mean corpuscular hemoglobin [MCH], mean corpuscular hemoglobin concentration [MCHC], hemoglobin concentration [HGB], NEUT, and BASO) showed consistent partial genetic effect in terms of GRS calculated with EAS- or EUR-specific associated SNPs, whereas only 4 traits (i.e., SCZ, AAM, PLT, and lymphocyte count [LYMPH]) exhibited consistent partial genetic effect in terms of GRS calculated with shared associated SNPs. Particularly, IS and TL exhibited a completely opposite partial genetic effect in terms of the EAS/EUR-specific associated SNPs or population-common associated SNPs compared with the overall genetic impact measured with all trait-associated SNPs.

Discussion

Summary of results in the present study

The present study has analyzed a total of 37 complex traits and sought to compare shared and distinct genetic components underlying them between the EAS and EUR populations. We discovered there existed pervasive consistence in heritability and trans-ethnic genetic correlation for these traits. Additionally, it needs to highlight that the trans-ethnic genetic correlation of continuous traits was on average higher compared to those binary traits, which may be due to the loss of information when converting some continuous phenotypes into binary ones (e.g., using HbA1c to define T2D) or in part reflects the discrepancy of classification and diagnosis of diseases in distinct populations.

Using cFDR [27, 28] we detected a great number of population-common association signals as well as many population-specific associated SNPs. A further exploration demonstrated these shared trait-associated SNPs generally showed the maximal positive correlation in effect sizes compared to population-specific trait-associated SNPs and null ones [7, 25]. Interestingly, we observed that even the shared association signals also exhibited a considerable degree of heterogeneity in genetic influence on traits across the EAS and EUR populations.

Furthermore, we revealed that population-specific associated SNPs were often more possible to suffer from natural selection compared with population-common associations, whereas population-common associated SNPs often displayed more consistent patterns in LD and MAF across continental populations.

Especially, the GRS analysis showed that population-common and population-specific associated SNPs have potentially different genetic influence on phenotypic variation and that the genetic differentiation from associated SNPs may at least partly explain the observed phenotypic variation across diverse ancestral groups. For example, the average GRS of SCZ in the EUR population was on average higher than that in the EAS population, partly contributing to the observation that SCZ was more prevalent in individuals of EUR ancestry than those of EAS ancestry [33]. The mean GRS of T2D was higher in the EAS population than that in the EUR population, supported by the observed incidence difference between the populations [34, 35]. It was reported that the absolute risk of T2D tended to be higher among Asians compared with Caucasians for any given level of body mass index [BMI] and waist-hip ratio [34]. As another example, for PCA higher mean GRS was observed in EAS compared to EUR, in line with a previous study which indicated that more than half of SNPs showed larger effect sizes in EAS than EUR [36].

However, we also observed patterns that seemed to be opposite against prior findings. For example, it was shown the mean GRS of AD in EUR was higher than that in EAS, in contrast to previous observation that Asians and Pacific Islanders were seven-fold more likely than whites to be diagnosed with AD because stronger Th17/Th22 polarization and mutations in immune-related genes such as DEFB1 [37,38,39]. These inconsistent results can be expected because of the complicated nature of these traits, and can be explained by gene-gene/gene-environment interaction, ethnic difference and genetic factors that are largely underappreciated in our current study.

Comparison our discoveries to prior studies

Like the findings obtained from prior studies that however often focused only on a single trait [11, 40, 41], our work, which considered much more traits and diseases, further confirmed extensive genetic overlap and identified a large number of common associated genetic loci across different populations. From a biological perspective, there is no doubt about the widespread existence of population-shared risk variants because the risk variants targeted by GWASs are often common genetic loci that are believed to be of ancient origin and largely shared among different populations.

As revealed in our study, while some of associated SNPs vary substantially across populations, common associated SNPs in the EAS and EUR populations nevertheless often show much more similar effect sizes and effect directions; therefore, at least part of trait-associated SNP mapping results discovered in one population can be transferred to the other populations [42]. It needs to highlight that we may underestimate the degree of genetic sharing between various ancestry groups because the much smaller number of individuals in the EAS GWASs reduces power to detect homogeneity of effect compared to the EUR GWASs. These findings are largely consistent with some recent discoveries that most common causal SNPs were shared across the EAS and EUR populations, high-posterior SNPs identified by fine-mapping often showed highly correlated effects, and population-specific genetic regions likely harbored common trait-associated SNPs which however failed to be detected in the other GWASs due to differences in LD, MAF, and/or sample size [43].

On the other hand, despite highly shared genetic architecture, we indeed found evidence of heterogeneity at trait-associated variants, which meanwhile challenges in assessing the transferability of risk variants across different ethnic populations based on associations discovered in EUR GWASs [44, 45]. For example, although the trans-ethnic marginal genetic correlation for population-common associated SNPs of HbA1c between EAS and EUR was as high as 0.63, heterogeneity was detected at 37.7% shared associated loci. This diversity may point to the difference in clinical definitions and phenotype measurements [46]; and it can be in part explained by interaction between gene-gene and gene-environment [47].

The genetic difference may also underlie the well-known trans-ethnic dissimilarities in prevalence or characteristics of the traits [48,49,50,51,52]. In our analysis, these population-specific association signals largely indicate the significant trans-ethnic difference, which exist in distinct LD and allele frequency [7, 53,54,55]. In addition, many studies have revealed that, unlike in most European ancestry populations, the population genetic history of non-European ancestry groups has undergone selective pressure due to the effects of malaria and other infectious diseases on erythrocytes [40, 56]. Another possible factor for the genetic inconsistency of complex traits between ancestry groups may be due to sample size difference and thus different statistical power between EAS and EUR studies.

Important scientific implications of our findings

Our findings regarding the degree of common or diverse genetic components of the traits across ancestral groups have important implication in practice. For example, theoretically, genetic correlation offers the maximal boundary of trans-ethnic genetic prediction power [5, 12]; however, both overall trans-ethnic correlation and marginal trans-ethnic correlation imply low accuracy when implementing genetic prediction in one population of interest on the basis of associated loci discovered in other populations [25], indicating the need to carry out GWASs with more ancestral groups.

In addition, it is helpful for aggregating multiple study cohorts across ethnicities to conduct trans-ethnic GWAS analysis [57,58,59,60], developing trans-ethnic genetic risk prediction [25, 61], and fine-mapping causal genetic variants in minority populations [62, 63]. It also holds the key to benefit more ethnic groups from current medical genomics researches [64,65,66]. All of these offer promising avenues in post-GWAS era by integrating trans-ethnic information.

Potential limitations

The present study is not without limitation. First, as mentioned before, the studies in EAS are in generally underpowered due to much smaller sample size compared to that in EUR and hence our results may be affected by power issues. The small sample size may also lead to unstable effect estimation for these SNPs. Moreover, as shown before, we cannot completely rule out the possibility that the imbalanced sample sizes can also partly interpret the observed trans-ethnic genetic difference in traits. For example, we found that sample size was significantly positively correlated with the number of trait-associated loci in both populations, and that the estimated trans-ethnic genetic correlation would become less significantly different from one when the difference of sample sizes in a pair of traits reduced. Thus, the external validation of our results with larger sample size especially for EAS GWASs is warranted.

Second, besides the difference in sample sizes, other discrepancies in study designs such as phenotypic definition, statistical methods, and covariate adjustment can be also contributable to the observed trans-ethnic genetic similarity and diversity. Examining and quantifying the relative contributions are imperative for understanding genetic heterogeneity across populations. However, compared to the difference in sample sizes which are already reported in summary statistics, the design discrepancies in GWAS are difficult to handle with only summary statistics. It needs large-scale individual-level data of phenotypes and genotypes, and is thus challenged by privacy concerns when sharing data [67].

Third, individuals of EAS and EUR ancestries are to a great extent genetically similar, whereas more major genetic differences are expected to be found between AFR and non-AFR populations [68]. Therefore, it is not clear whether our findings can be generalized to comparison in other populations such as EUR and AFR. Unfortunately, the number of GWASs performed in individuals of AFR descent is still too limited to enable comparative studies.

Fourth, our analysis only considered common SNPs (MAF > 0.01) and ignored rare variants, which usually have a recent origin compared with common ones from ancient origin. Theoretically, rare risk variants might be more likely to be population-specific and could possibly carry a greater risk effect, which probably leads us to underestimate the genetic heterogeneity across populations.

Fifth, to our knowledge, this is the first time that the cFDR method has been employed in detecting trans-ethnic genetic overlap for a large range of complex traits. However, it implements association mapping at a fixed FDR level rather than the standard error measure such as family-wise error rate (FWER) or type I error rate which is more widely-used in a typical GWAS. FDR is more liberal compared to FWER; thus, we can discover a much larger number of trait-associated SNPs. Although the cFDR method has been well-established under the context of pleiotropy-informed association mapping in ancestry-matched populations [27, 28] and also demonstrated to be well-calibrated in our study (Figures S7-S8), its ability of controlling FWER is less understood. Consequently, we just considered cFDR as a powerful tool for discovering evidence of trans-ethnic genetic overlap in our application, and by no means attempted to replace the standard GWAS analysis strategy with cFDR nor the cFDR level (e.g., 0.05 used here) with the genome-wide significance level (e.g., 5 × 10− 8).

Conclusions

Our work provides an in-depth understanding of similarity and diversity regarding genetic architecture for complex traits across diverse populations, and can assist in trans-ethnic association analysis, genetic risk prediction, and causal variant fine mapping.

Materials and methods

Summary statistics of 37 complex traits

We obtained summary statistics (e.g., marginal effect size and standard error) of 37 complex traits (10 binary and 27 continuous) analyzed on EAS-only or EUR-only individuals (Table 1 and Tables S6-S7). These traits included lipids (e.g., TG), blood cell phenotypes (e.g., neutrophil [NEUT] and MONO), diseases (e.g., BRC, T2D, and prostate cancer [PCA]), and anthropometric phenotypes (e.g., BMI and height).

For each analyzed trait, we carried out stringent quality control in both populations by following previous work [12, 30, 69]: (i) filtered out SNPs without rs label; (ii) deleted non-biallelic SNPs and those with strand-ambiguous alleles; (iii) removed SNPs whose alleles did not match with those in the 1000 Genomes Project; (iv) excluded duplicated SNPs; (v) filtered out palindromic SNPs containing the same bases; (vi) kept only common SNPs (MAF > 0.01 in each population) which were also included within the 1000 Genomes Project and shared between the EAS and EUR GWASs.

Here, MAF was calculated with genotypes of EAS (N = 504) or EUR (N = 503) individuals in the 1000 Genomes Project if missing in the original summary statistics data; the threshold value of 0.01 for MAF was selected as it was widely used in summary statistics-based studies [12, 30, 69]. We further aligned the effect allele of all remaining SNPs for each trait between the two populations.

Estimation of heritability and cross-trait genetic correlation in the same population

We first employed LDSC to estimate SNP-based heritability for each trait separately in the two populations [30]. The LDSC software (version v1.0.1) was downloaded from https://github.com/bulik/ldsc and the analysis was conducted with default parameter settings. The LD score was calculated with genotypes of SNPs (MAF > 0.01 and the P value of Hardy Weinberg equilibrium test > 1.0 × 10− 5) with a 10 Mb window on EAS or EUR individuals in the 1000 Genomes Project. Then, the LD score of SNP was regressed on the square of Z-statistic of the analyzed trait, with the regression slope offering an unbiased estimate for heritability. Besides quality control procedures described above, we here further removed SNPs located within the major histocompatibility complex region (chr6: 28.5 ~ 33.5 Mb) because of its complicated structure which was often difficult to estimate accurately from an external reference panel [30, 70]. Relying only on summary statistics and LD scores, LDSC can be also applied to calculate the cross-trait genetic correlation in the same population [69].

Estimation of trans-ethnic genetic correlation across populations

To evaluate genetic similarity and diversity for these traits across populations, we calculate the global trans-ethnic genetic correlation (ρg) via popcorn [12]. Conceptually, ρg is defined as the correlation between SNP effect sizes of the trait in various ancestral groups to measure the extent to which the same SNP exhibits the same or similar impact on phenotypic variation [12, 25, 26]. Methodologically, popcorn was proposed from the Bayesian perspective by assuming effect sizes of SNPs following an infinitesimal model [71], and can be considered as a natural trans-ethnic extension of LDSC. The trans-ethnic LD score of each SNP was downloaded from https://github.com/brielin/popcorn, which was calculated with genotypes of EAS or EUR individuals in the 1000 Genomes Project between the focal SNP and all the flanking ones within a 10 Mb window. To obtain an unbiased estimate of trans-ethnic genetic correlation, we implement an unbounded estimation algorithm in popcorn, which likely leads to an estimate less than − 1 or greater than 1.

Identification of associated SNPs and shared genetic overlap across populations

Conditional false discovery rate and conjunction conditional false discovery rate

From the statistical perspective, we find that the discovery of trans-ethnic genetic overlap can be viewed as the similar principle of detecting pleiotropic associations for genetically correlated traits in one population. Over the past few years, many novel methods have been developed to identify pleiotropy [72,73,74]. Among those, cFDR is a novel method that utilizes pleiotropy to identify genetic overlap and can be considered as an innovative expansion of the classical FDR for a single trait in one population to the same trait in trans-ethnic cases [27, 28]. By integrating association results from multiple traits, cFDR can offer important sights into trans-ethnic genetic overlap and increase power to identify less significant association signals.

In our application context, the null hypothesis of FDR is no association between a SNP and the trait under investigation in one population. Following the definition and principle of FDR, cFDR can be logically described as the likelihood that a SNP is null for the trait in one population based on posterior probability, given that the observed P values for the trait are below a pre-determined threshold in both populations. Formally, with two sets of P values as input, cFDR is calculated as cFDR1|2=Prob(\({H}_{0}^{\left(1\right)}|\)P1 < p1, P2 < p2), where p1 and p2 are the observed P values of a particular SNP for the trait in the two populations, respectively; and \({H}_{0}^{\left(1\right)}\) denotes the null hypothesis indicating there does not exist an association between the SNP and the trait in the first population. cFDR is efficiently estimated with an empirical Bayesian manner that was applicable for computing the local FDR [75].

As the principal and conditional positions for the trait in cFDR described above are exchangeable between the populations, cFDR2|1 is defined in a similar way. Therefore, ccFDR for identifying shared SNPs is simply expressed as ccFDR1|2=max(cFDR1|2, cFDR2|1), which is defined as the probability that a particular SNP has a false positive association with the trait in the two populations given the observed P values. Finally, SNPs with ccFDR less than a given significance threshold can be prioritized to be population-common SNPs. Although the traditional meta-analysis is also often applied in trans-ethic association studies [76,77,78,79,80,81]; however, the association discovered by meta-analysis cannot certainly suggest trans-ethic genetic overlap because such association might be only present in one special population.

Because cFDR and ccFDR are constructed for relatively independent SNPs, to generate uncorrelated SNPs, following prior work [82] we applied the LD pruning (the width of SNP window = 50 and r2 = 0.1) in PLINK with genotypes of EAS or EUR individuals separately in the 1000 Genomes Project as the reference panel for LD calculation, and then combined the two sets of SNPs available from both populations (Table S6). In addition, based on findings observed in other studies [73, 74], genetic/phenotypic correlation between traits can result in inflated test statistics. Therefore, to minimize false discovery in our cFDR analysis, we further generated uncorrelated Z-statistics for every trait across populations by multiplying them by the inverse of a correlation matrix, which can be easily calculated in terms of the two statistics of null SNPs (e.g., SNPs with P > 10− 4) [72, 83]. This decorrelation strategy maximizes the transformed test statistics and the original ones [84]; therefore, it has the minimal influence on association identification. These uncorrelated Z-statistics were ultimately transformed into two-sided P values for the cFDR analysis based on normal approximation.

Our simulation studies already demonstrated that the cFDR method could maintain well-calibrated control of FDR at the given level and behaved better when identifying population-common trait-associated SNPs compared to the naïve minimum P-value method (Supplement notes and Figures S7-S8).

Various types of SNPs and four-group model

Relying on cFDR and ccFDR, for each trait we could divide all the analyzed SNPs into four incompatible groups: (i) not associated with the trait in neither population (i.e., null SNPs); (ii) only associated with the trait in the EAS population but not in the EUR population (i.e., EAS-specific associated SNPs); (iii) only associated with the trait in the EUR population but not in the EAS population (i.e., EUR-specific associated SNPs); (iv) associated with the trait in both the two populations (i.e., population-common associated SNPs).

To measure the degree of genetic components shared by the same trait across the populations, we further applied the four-group model [85] which examined SNPs in the four groups mentioned above. This model aims to estimate the proportions of SNPs in each group, and employs the LRT method to assess the statistical significance for overall trans-ethnic genetic overlap. Statistically, the four-group model assumes that P values of null SNPs (not associated with the trait in neither populations) follow the uniform distribution and P values of non-null SNPs (associated with the trait at least in one population) follow the Beta distribution.

Genetic correlation and heterogeneity of SNPs between the two populations

Marginal genetic correlation across populations

For every trait we calculated the marginal genetic correlation (rm) of SNP effect sizes in each of the four groups using a recently proposed method called MAGIC [25]. Compared to the traditional Pearson’s correlation, which often underestimates the marginal genetic correlation thus leads to the so-called correlation attenuation because of failing to take the estimation error of effect sizes into account [25, 86], MAGIC has the advantage of generating unbiased correlation estimation by accounting for the uncertainty under the framework of measurement error model [87].

Linear regression for SNP effect sizes

We also carried out a simple linear model without the intercept term for only population-common associated SNPs of each trait by regressing effect sizes of SNPs in one population on those in another population. The slope of the linear regression model provides an indicator for the relative magnitude of effect sizes for shared trait-associated SNPs between the two populations. For common associated SNPs, we examined the heterogeneity in genetic effect of SNPs on the trait across EAS and EUR populations via Cochran’s Q test in the R metafor package.

Characteristics of associated SNPs between EAS and EUR populations

Patterns for LD, MAF, and Wright’s fixation index

After detecting common associations, we wondered whether there exist different patterns of LD and MAF for trait-associated SNPs compared to those null ones. To this aim, we first obtained the two LD scores for every SNP in the four groups based on genotypes available from EAS or EUR individuals in the 1000 Genomes Project, and then calculated their LDCV across the populations. Here, coefficient of variation, rather than variance, was utilized because SNPs with higher LD scores tended to have greater variation between populations [7, 24]. In a similar way, we calculated MAFCV for each SNP between the two populations to evaluate how MAF varies between populations.

We further evaluated whether an identified trait-associated SNP had been under natural selection. If this was the case, then a substantial between-population differentiation would be observed in the allele frequency [24, 88]. To this aim, we applied the Wright’s fixation index (Fst) to evaluate such an allele frequency diversity across populations under natural selection [24, 88, 89], and calculated Fst of each SNP in the four groups for every trait with genotypes available from EAS and EUR individuals in the 1000 Genomes Project using PLINK.

To examine LDCV, MAFCV, and Fst in the four groups, we carried out the Kruskal test for each trait, with the resulting P values being FDR corrected to account for multiple testing using the Benjamini-Hochberg procedure. We also performed a paired test to compare the average of LDCV, MAFCV, and Fst across the traits while simply ignoring the uncertainty of the estimated average in each group.

Partial and overall genetic risk score analysis

To further demonstrate the direction of genetic differentiation, for each trait we conducted a GRS analysis [29]. The calculated GRS in part measures the stratification of the whole population based on estimates of individual’s genetic susceptibility. In our analysis the GRS of a given individual was simply computed as GRS=\({\sum }_{j}^{J}{G}_{j}{\widehat{\beta }}_{j}/J\), where J was the number of SNPs used and G represented the genotype (coded as 0, 1, or 2) available from the EAS or EUR individuals in the 1000 Genomes Project. For binary phenotypes, \(\widehat{\beta }\) was the original marginal SNP effect size; whereas, for the compatibility across populations, for continuous phenotypes we re-scaled \(\widehat{\beta }\) based on Z-statistic [90]; that is, \({\widehat{\beta }}_{new}=\text{Z}/\sqrt{2\times \text{M}\text{A}\text{F}\times \left(1-\text{M}\text{A}\text{F}\right)\times \text{s}\text{a}\text{m}\text{p}\text{l}\text{e} \text{s}\text{i}\text{z}\text{e}+{Z}^{2}}\).

For each pair of traits under analysis, three types of GRS were generated using three distinct sets of SNPs (Table 2), including population-specific loci (J = f10 or f01), population-common loci (J = f11) and population-associated loci (J = f10 + f11 or f01 + f11). For the convenience of description, we referred to the first two GRSs as partial GRS, while the third one as overall GRS.

Exploring the influence of sample size difference

We finally attempted to investigate whether the sample size difference could influence our findings with regards to genetic similarity and diversity of traits between the EAS and EUR populations. First, after obtaining the heritability estimate and its standard error for a given trait in the two populations (denoted by \({\widehat{h}}_{1}^{2}\) and \({\widehat{h}}_{2}^{2}\),, se1 and se2, respectively), we studied the influence of sample sizes of traits on the estimated heritability and its standard error.

Second, to assess the impact of sample size on the trans-ethnic diversity of estimated heritability, we employed NCV to measure the difference of sample sizes, and examined its relation with the coefficient of variation of heritability, the estimated slope of effect sizes for population-common SNPs and the proportion of population-common SNPs with heterogeneous effects.

Third, we implemented an approximation normal test to examine the significance of the difference in the estimated heritability between the two populations by calculating u=\(({\widehat{h}}_{1}^{2}-{\widehat{h}}_{2}^{2})/{{(se}_{1}^{2}{+se}_{2}^{2})}^{-1/2}\). We obtained the P value of u using the standard normal distribution as the null distribution, with the issue of multiple testing corrected via by the Benjamini-Hochberg method. By doing this, we explicitly took the difference of sample sizes in traits into account by modeling its standard error when comparing heritability.

Data Availability

All data generated or analyzed during this study are included in this published article and its supplementary information file.

Abbreviations

ccFDR:

Conjunction conditional false discovery rate

cFDR:

Conditional false discovery rate

EAS:

East Asian

EUR:

European

F st :

Wright’s fixation index

GRS:

Genetic risk score

GWAS:

Genome-wide association study

LD:

Linkage disequilibrium

LDCV:

Coefficient of variation of LD scores

LDSC:

Linkage disequilibrium score regression

MAF:

Minor allele frequency

MAFCV:

Coefficient of variation of MAF

SNP:

Single-nucleotide polymorphism

References

  1. Tam V, Patel N, Turcotte M, Bosse Y, Pare G, Meyre D. Benefits and limitations of genome-wide association studies. Nat Rev Genet. 2019;20(8):467–84.

    Article  CAS  PubMed  Google Scholar 

  2. Hindorff LA, Sethupathy P, Junkins HA, Ramos EM, Mehta JP, Collins FS, Manolio TA. Potential etiologic and functional implications of genome-wide association loci for human diseases and traits. Proc Natl Acad Sci U S A. 2009;106(23):9362–7.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  3. Visscher PM, Wray NR, Zhang Q, Sklar P, McCarthy MI, Brown MA, Yang J. 10 years of GWAS Discovery: Biology, function, and translation. Am J Hum Genet. 2017;101(1):5–22.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  4. Buniello A, MacArthur JAL, Cerezo M, Harris LW, Hayhurst J, Malangone C, McMahon A, Morales J, Mountjoy E, Sollis E, et al. The NHGRI-EBI GWAS catalog of published genome-wide association studies, targeted arrays and summary statistics 2019. Nucleic Acids Res. 2019;47(D1):D1005–12.

    Article  CAS  PubMed  Google Scholar 

  5. Martin AR, Gignoux CR, Walters RK, Wojcik GL, Neale BM, Gravel S, Daly MJ, Bustamante CD, Kenny EE. Human demographic history impacts genetic risk prediction across diverse populations. Am J Hum Genet. 2017;100(4):635–49.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  6. Carlson CS, Matise TC, North KE, Haiman CA, Fesinmeyer MD, Buyske S, Schumacher FR, Peters U, Franceschini N, Ritchie MD, et al. Generalization and dilution of Association results from european GWAS in populations of Non-European Ancestry: the PAGE study. PLoS Biol. 2013;11(9):e1001661.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  7. Marigorta UM, Navarro A. High trans-ethnic replicability of GWAS results implies common causal variants. PLoS Genet. 2013;9(6):e1003566.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  8. Kraft P, Zeggini E, Ioannidis JPA. Replication in genome-wide Association Studies. Stat Sci. 2009;24(4):561–73.

    Article  PubMed  PubMed Central  Google Scholar 

  9. Li YH, Lan L, Wang YL, Yang C, Tang WG, Cui GQ, Luo SQ, Cheng YB, Liu YC, Liu JY, et al. Extremely cold and hot temperatures increase the risk of diabetes mortality in metropolitan areas of two chinese cities. Environ Res. 2014;134:91–7.

    Article  CAS  PubMed  Google Scholar 

  10. Waters KM, Stram DO, Hassanein MT, Le Marchand L, Wilkens LR, Maskarinec G, Monroe KR, Kolonel LN, Altshuler D, Henderson BE, et al. Consistent association of type 2 diabetes risk variants found in Europeans in diverse racial and ethnic groups. PLoS Genet. 2010;6(8):e1001078.

    Article  PubMed  PubMed Central  Google Scholar 

  11. Liu JZ, van Sommeren S, Huang HL, Ng SC, Alberts R, Takahashi A, Ripke S, Lee JC, Jostins L, Shah T, et al. Association analyses identify 38 susceptibility loci for inflammatory bowel disease and highlight shared genetic risk across populations. Nat Genet. 2015;47(9):979–86.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  12. Brown BC, Ye CJ, Price AL, Zaitlen N, Network AGE. Transethnic genetic-correlation estimates from Summary Statistics. Am J Hum Genet. 2016;99(1):76–88.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  13. Bigdeli TB, Ripke S, Peterson RE, Trzaskowski M, Bacanu SA, Abdellaoui A, Andlauer TFM, Beekman ATF, Berger K, Blackwood DHR, et al. Genetic effects influencing risk for major depressive disorder in China and Europe. Translational Psychiatry. 2017;7:e1074. e1074.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  14. Veturi Y, de los Campos G, Yi NJ, Huang W, Vazquez AI, Kuhnel B. Modeling heterogeneity in the Genetic Architecture of ethnically diverse groups using Random Effect Interaction Models. Genetics. 2019;211(4):1395–407.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  15. Sirugo G, Williams SM, Tishkoff SA. The missing diversity in Human Genetic Studies. Cell. 2019;177(1):26–31.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  16. Popejoy AB, Fullerton SM. Genomics is failing on diversity. Nature. 2016;538(7624):161–4.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  17. Moltke I, Grarup N, Jorgensen ME, Bjerregaard P, Treebak JT, Fumagalli M, Korneliussen TS, Andersen MA, Nielsen TS, Krarup NT, et al. A common Greenlandic TBC1D4 variant confers muscle insulin resistance and type 2 diabetes. Nature. 2014;512(7513):190–3.

    Article  CAS  PubMed  Google Scholar 

  18. Tang CS, Zhang H, Cheung CYY, Xu M, Ho JCY, Zhou W, Cherny SS, Zhang Y, Holmen O, Au KW, et al. Exome-wide association analysis reveals novel coding sequence variants associated with lipid traits in chinese. Nat Commun. 2015;6:10206.

    Article  CAS  PubMed  Google Scholar 

  19. Clarke TK, Adams MJ, Davies G, Howard DM, Hall LS, Padmanabhan S, Murray AD, Smith BH, Campbell A, Hayward C, et al. Genome-wide association study of alcohol consumption and genetic overlap with other health-related traits in UK Biobank (N = 112117). Mol Psychiatry. 2017;22(10):1376–84.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  20. Wray NR, Ripke S, Mattheisen M, Trzaskowski M, Byrne EM, Abdellaoui A, Adams MJ, Agerbo E, Air TM, Andlauer TMF, et al. Genome-wide association analyses identify 44 risk variants and refine the genetic architecture of major depression. Nat Genet. 2018;50(5):668–81.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  21. Cai N, Bigdeli TB, Kretzschmar W, Li YH, Liang JQ, Song L, Hu JC, Li QB, Jin W, Hu ZF, et al. Sparse whole-genome sequencing identifies two loci for major depressive disorder. Nature. 2015;523(7562):588–91.

    Article  CAS  Google Scholar 

  22. Wu Y, Waite LL, Jackson AU, Sheu WHH, Absher SBD, Absher D, Arnett DK, Boerwinkle E, Bonnycastle LL, Carty CL, et al. Trans-ethnic fine-mapping of lipid loci identifies Population-Specific signals and allelic heterogeneity that increases the trait Variance explained. PLoS Genet. 2013;9(3):e1003379.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  23. Coram MA, Duan Q, Hoffmann TJ, Thornton T, Knowles JW, Johnson NA, Ochs-Balcom HM, Donlon TA, Martin LW, Eaton CB, et al. Genome-wide characterization of Shared and distinct genetic components that influence blood lipid levels in ethnically diverse human populations. Am J Hum Genet. 2013;92(6):904–16.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  24. Guo J, Wu Y, Zhu ZH, Zheng ZL, Trzaskowski M, Zeng J, Robinson MR, Visscher PM, Yang J. Global genetic differentiation of complex traits shaped by natural selection in humans. Nat Commun. 2018;9:1865.

    Article  PubMed  PubMed Central  Google Scholar 

  25. Lu HJ, Wang T, Zhang JH, Zhang SO, Huang SP, Zeng P. Evaluating marginal genetic correlation of associated loci for complex diseases and traits between european and east asian populations. Hum Genet. 2021;140(9):1285–97.

    Article  CAS  PubMed  Google Scholar 

  26. Galinsky KJ, Reshef YA, Finucane HK, Loh PR, Zaitlen N, Patterson NJ, Brown BC, Price AL. Estimating cross-population genetic correlations of causal effect sizes. Genet Epidemiol. 2019;43(2):180–8.

    Article  PubMed  Google Scholar 

  27. Andreassen OA, Djurovic S, Thompson WK, Schork AJ, Kendler KS, O’Donovan MC, Rujescu D, Werge T, van de Bunt M, Morris AP, et al. Improved detection of Common Variants Associated with Schizophrenia by leveraging Pleiotropy with Cardiovascular-Disease Risk factors. Am J Hum Genet. 2013;92(2):197–209.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  28. Smeland OB, Frei O, Shadrin A, O’Connell K, Fan CC, Bahrami S, Holland D, Djurovic S, Thompson WK, Dale AM, et al. Discovery of shared genomic loci using the conditional false discovery rate approach. Hum Genet. 2020;139(1):85–94.

    Article  CAS  PubMed  Google Scholar 

  29. Choi SW, Mak TSH, O’Reilly PF. Tutorial: a guide to performing polygenic risk score analyses. Nat Protoc. 2020;15(9):2759–72.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  30. Bulik-Sullivan BK, Loh PR, Finucane HK, Ripke S, Yang J, Patterson N, Daly MJ, Price AL, Neale BM, Grp SW. LD score regression distinguishes confounding from polygenicity in genome-wide association studies. Nat Genet. 2015;47(3):291–5.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  31. Boyle EA, Li YI, Pritchard JK. An expanded view of Complex Traits: from polygenic to Omnigenic. Cell. 2017;169(7):1177–86.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  32. Pathak PK, Tripathi N, Subramanian SV. Secular Trends in Menarcheal Age in India-Evidence from the Indian Human Development Survey. PLoS ONE. 2014;9(11):e111027.

    Article  PubMed  PubMed Central  Google Scholar 

  33. Ikeda M, Takahashi A, Kamatani Y, Okahisa Y, Kunugi H, Mori N, Sasaki T, Ohmori T, Okamoto Y, Kawasaki H, et al. A genome-wide association study identifies two novel susceptibility loci and trans population polygenicity associated with bipolar disorder. Mol Psychiatry. 2018;23(3):639–47.

    Article  CAS  PubMed  Google Scholar 

  34. Huxley R, James WPT, Barzi F, Patel JV, Lear SA, Suriyawongpaisal P, Janus E, Caterson I, Zimmet P, Prabhakaran D, et al. Ethnic comparisons of the cross-sectional relationships between measures of body size with diabetes and hypertension. Obes Rev. 2008;9:53–61.

    Article  PubMed  Google Scholar 

  35. Spracklen CN, Horikoshi M, Kim YJ, Lin K, Bragg F, Moon S, Suzuki K, Tam CHT, Tabara Y, Kwak SH, et al. Identification of type 2 diabetes loci in 433,540 east asian individuals. Nature. 2020;582(7811):240–5.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  36. Virlogeux V, Graff RE, Hoffmann TJ, Witte JS. Replication and heritability of prostate Cancer Risk Variants: Impact of Population-Specific factors. Cancer Epidemiol Biomark Prev. 2015;24(6):938–43.

    Article  Google Scholar 

  37. Kaufman BP, Guttman-Yassky E, Alexis AF. Atopic dermatitis in diverse racial and ethnic groups-variations in epidemiology, genetics, clinical presentation and treatment. Exp Dermatol. 2018;27(4):340–57.

    Article  PubMed  Google Scholar 

  38. Janumpally SR, Feldman SR, Gupta AK, Fleischer AB. In the United States, blacks and Asian/Pacific Islanders are more likely than whites to seek medical care for atopic dermatitis. Arch Dermatol. 2002;138(5):634–7.

    Article  PubMed  Google Scholar 

  39. Kim E, Lee JE, Namkung JH, Kim PS, Kim S, Shin ES, Cho EY, Yang JM. Single nucleotide polymorphisms and the haplotype in the DEFB1 gene are associated with atopic dermatitis in a korean population. J Dermatol Sci. 2009;54(1):25–30.

    Article  CAS  PubMed  Google Scholar 

  40. Wheeler E, Leong A, Liu CT, Hivert MF, Strawbridge RJ, Podmore C, Li M, Yao J, Sim X, Hong JY, et al. Impact of common genetic determinants of Hemoglobin A1c on type 2 diabetes risk and diagnosis in ancestrally diverse populations: a transethnic genome-wide meta-analysis. PLoS Med. 2017;14(9):e1002383.

    Article  PubMed  PubMed Central  Google Scholar 

  41. Dastani Z, Johnson T, Kronenberg F, Nelson CP, Assimes TL, Marz W, Richards JB, Consortium C, Consortium A. The shared allelic architecture of adiponectin levels and coronary artery disease. Atherosclerosis. 2013;229(1):145–8.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  42. Stranger BE, Montgomery SB, Dimas AS, Parts L, Stegle O, Ingle CE, Sekowska M, Smith GD, Evans D, Gutierrez-Arcelus M, et al. Patterns of Cis Regulatory Variation in Diverse Human populations. PLoS Genet. 2012;8(4):272–84.

    Article  Google Scholar 

  43. Shi H, Burch KS, Johnson R, Freund MK, Kichaev G, Mancuso N, Manuel AM, Dong N, Pasaniuc B. Localizing Components of Shared Transethnic Genetic Architecture of Complex Traits from GWAS Summary Data. Am J Hum Genet. 2020;106(6):805–17.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  44. Marigorta UM, Lao O, Casals F, Calafell F, Morcillo-Suarez C, Faria R, Bosch E, Serra F, Bertranpetit J, Dopazo H, et al. Recent human evolution has shaped geographical differences in susceptibility to disease. BMC Genomics. 2011;12:55.

    Article  PubMed  PubMed Central  Google Scholar 

  45. Fu JY, Festen EAM, Wijmenga C. Multi-ethnic studies in complex traits. Hum Mol Genet. 2011;20:R206–13.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  46. Evangelou E, Fellay J, Colombo S, Martinez-Picado J, Obel N, Goldstein DB, Telenti A, Ioannidis JPA. Impact of phenotype definition on genome-wide Association signals: empirical evaluation in human immunodeficiency virus type 1 infection. Am J Epidemiol. 2011;173(11):1336–42.

    Article  PubMed  PubMed Central  Google Scholar 

  47. Baye TM, Abebe T, Wilke RA. Genotype-environment interactions and their translational implications. Personalized Med. 2011;8(1):59–70.

    Article  Google Scholar 

  48. HermanGiddens ME, Slora EJ, Wasserman RC, Bourdony CJ, Bhapkar MV, Koch GG, Hasemeier CM. Secondary sexual characteristics and menses in young girls seen in office practice: a study from the pediatric research in office settings network. Pediatrics. 1997;99(4):505–12.

    Article  CAS  PubMed  Google Scholar 

  49. Anderson SE, Dallal GE, Must A. Relative weight and race influence average age at menarche: results from two nationally representative surveys of US girls studied 25 years apart. Pediatrics. 2003;111(4):844–50.

    Article  PubMed  Google Scholar 

  50. Chumlea WC, Schubert CM, Roche AF, Kulin HE, Lee PA, Himes JH, Sun SS. Age at menarche and racial comparisons in US girls. Pediatrics. 2003;111(1):110–3.

    Article  PubMed  Google Scholar 

  51. Sarnowski C, Cousminer DL, Franceschini N, Raffield LM, Jia G, Fernández-Rhodes L, Grant SFA, Hakonarson H, Lange LA, Long J, et al. Large trans-ethnic meta-analysis identifies AKR1C4 as a novel gene associated with age at menarche. Hum Reprod. 2021;36(7):1999–2010.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  52. Dvornyk V, Waqar-ul-Haq. Genetics of age at menarche: a systematic review. Hum Reprod Update. 2012;18(2):198–210.

    Article  CAS  PubMed  Google Scholar 

  53. Beavan S, Prentice A, Dibba B, Yan L, Cooper C, Ralston SH. Polymorphism of the collagen type Ialpha1 gene and ethnic differences in hip-fracture rates. N Engl J Med. 1998;339(5):351–2.

    Article  CAS  PubMed  Google Scholar 

  54. Dvornyk V, Liu XH, Shen H, Lei SF, Zhao LJ, Huang QR, Qin YJ, Jiang DK, Long JR, Zhang YY, et al. Differentiation of Caucasians and Chinese at bone mass candidate genes: implication for ethnic difference of bone mass. Ann Hum Genet. 2003;67:216–27.

    Article  CAS  PubMed  Google Scholar 

  55. Lei SF, Deng FY, Liu XH, Huang QR, Qin YJ, Zhou Q, Jiang DK, Li YM, Mo XY, Liu MY, et al. Polymorphisms of four bone mineral density candidate genes in chinese populations and comparison with other populations of different ethnicity. J Bone Miner Metab. 2003;21(1):34–42.

    Article  CAS  PubMed  Google Scholar 

  56. Luzzatto L, Usanga FA, Reddy S. Glucose-6-phosphate dehydrogenase deficient red cells: resistance to infection by malarial parasites. Science. 1969;164(3881):839–42.

    Article  CAS  PubMed  Google Scholar 

  57. Wojcik GL, Graff M, Nishimura KK, Tao R, Haessler J, Gignoux CR, Highland HM, Patel YM, Sorokin EP, Avery CL, et al. Genetic analyses of diverse populations improves discovery for complex traits. Nature. 2019;570(7762):514–8.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  58. Guenther CA, Tasic B, Luo L, Bedell MA, Kingsley DM. A molecular basis for classic blond hair color in Europeans. Nat Genet. 2014;46(7):748–52.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  59. Zaitlen N, Pasaniuc B, Gur T, Ziv E, Halperin E. Leveraging genetic variability across populations for the identification of causal variants. Am J Hum Genet. 2010;86(1):23–33.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  60. Morris AP. Transethnic meta-analysis of genomewide association studies. Genet Epidemiol. 2011;35(8):809–22.

    Article  PubMed  PubMed Central  Google Scholar 

  61. Coram MA, Fang HY, Candille SI, Assimes TL, Tang H. Leveraging multi-ethnic evidence for Risk Assessment of quantitative traits in minority populations. Am J Hum Genet. 2017;101(2):218–26.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  62. Spracklen CN, Chen P, Kim YJ, Wang X, Cai H, Li SX, Long JR, Wu Y, Wang YX, Takeuchi F, et al. Association analyses of east asian individuals and trans-ancestry analyses with european individuals reveal new loci associated with cholesterol and triglyceride levels. Hum Mol Genet. 2017;26(9):1770–84.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  63. Coram MA, Candille SI, Duan Q, Chan KHK, Li Y, Kooperberg C, Reiner AP, Tang H. Leveraging multi-ethnic evidence for Mapping Complex Traits in Minority populations: an empirical Bayes Approach. Am J Hum Genet. 2015;96(5):740–52.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  64. Rosenberg NA, Huang L, Jewett EM, Szpiech ZA, Jankovic I, Boehnke M. Genome-wide association studies in diverse populations. Nat Rev Genet. 2010;11(5):356–66.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  65. Bustamante CD, Burchard EG, De La Vega FM. Genomics for the world. Nature. 2011;475(7355):163–5.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  66. Teo YY, Small KS, Kwiatkowski DP. Methodological challenges of genome-wide association analysis in Africa. Nat Rev Genet. 2010;11(2):149–60.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  67. Stein CM. Challenges of Genetic Data sharing in African Studies. Trends Genet. 2020;36(12):895–6.

    Article  CAS  PubMed  Google Scholar 

  68. Li H, Durbin R. Inference of human population history from individual whole-genome sequences. Nature. 2011;475(7357):493–U484.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  69. Bulik-Sullivan B, Finucane HK, Anttila V, Gusev A, Day FR, Loh PR, Duncan L, Perry JRB, Patterson N, Robinson EB, et al. An atlas of genetic correlations across human diseases and traits. Nat Genet. 2015;47(11):1236–41.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  70. Finucane HK, Bulik-Sullivan B, Gusev A, Trynka G, Reshef Y, Loh PR, Anttila V, Xu H, Zang CZ, Farh K, et al. Partitioning heritability by functional annotation using genome-wide association summary statistics. Nat Genet. 2015;47(11):1228–35.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  71. Falconer D, Mackay TJULG. Introduction to quantitative genetics. Essex. 1996.

  72. Zeng P, Hao XJ, Zhou X. Pleiotropic mapping and annotation selection in genome-wide association studies with penalized gaussian mixture models. Bioinformatics. 2018;34(16):2797–807.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  73. Ray D, Chatterjee N. A powerful method for pleiotropic analysis under composite null hypothesis identifies novel shared loci between type 2 diabetes and prostate Cancer. PLoS Genet. 2020;16(12):e1009218.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  74. Wang T, Lu H, Zeng P. Identifying pleiotropic genes for complex phenotypes with summary statistics from a perspective of composite null hypothesis testing. Brief Bioinform 2021.

  75. Efron B. Size, power and false discovery rates. Ann Stat. 2007;35(4):1351–77.

    Article  Google Scholar 

  76. Wang X, Chua HX, Chen P, Ong RTH, Sim XL, Zhang WH, Takeuchi F, Liu XY, Khor CC, Tay WT, et al. Comparing methods for performing trans-ethnic meta-analysis of genome-wide association studies. Hum Mol Genet. 2013;22(11):2303–11.

    Article  CAS  PubMed  Google Scholar 

  77. Keller MF, Reiner AP, Okada Y, van Rooij FJA, Johnson AD, Chen MH, Smith AV, Morris AP, Tanaka T, Ferrucci L, et al. Trans-ethnic meta-analysis of white blood cell phenotypes. Hum Mol Genet. 2014;23(25):6944–60.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  78. Iyengar SK, Sedor JR, Freedman BI, Kao WHL, Kretzler M, Keller BJ, Abboud HE, Adler SG, Best LG, Bowden DW, et al. Genome-wide Association and Trans-ethnic Meta-analysis for Advanced Diabetic kidney disease: Family Investigation of Nephropathy and Diabetes (FIND). PLoS Genet. 2015;11(8):e1005352.

    Article  PubMed  PubMed Central  Google Scholar 

  79. Tang CSM, Gui HS, Kapoor A, Kim JH, Luzon-Toro B, Pelet A, Burzynski G, Lantieri F, So MT, Berrios C, et al. Trans-ethnic meta-analysis of genome-wide association studies for Hirschsprung disease. Hum Mol Genet. 2016;25(23):5265–75.

    CAS  PubMed  PubMed Central  Google Scholar 

  80. Magi R, Horikoshi M, Sofer T, Mahajan A, Kitajima H, Franceschini N, McCarthy MI, Morris AP, Consortium C-K, Consortium TD-G. Trans-ethnic meta-regression of genome-wide association studies accounting for ancestry increases power for discovery and improves fine-mapping resolution. Hum Mol Genet. 2017;26(18):3639–50.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  81. van Rooij FJA, Qayyum R, Smith AV, Zhou Y, Trompet S, Tanaka T, Keller MF, Chang LC, Schmidt H, Yang ML, et al. Genome-wide trans-ethnic Meta-analysis identifies seven genetic loci influencing erythrocyte traits and a role for RBPMS in Erythropoiesis. Am J Hum Genet. 2017;100(1):51–63.

    Article  PubMed  Google Scholar 

  82. Li CY, Yang TM, Ou RW, Wei QQ, Shang HF. Genome-wide genetic links between amyotrophic lateral sclerosis and autoimmune diseases. BMC Med. 2021;19(1):27.

    Article  PubMed  PubMed Central  Google Scholar 

  83. Lin DY, Sullivan PF. Meta-analysis of genome-wide Association Studies with overlapping subjects. Am J Hum Genet. 2009;85(6):862–72.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  84. LeBlanc M, Zuber V, Thompson WK, Andreassen OA, Frigessi A, Andreassen BK, Consortium PG. A correction for sample overlap in genome-wide association studies in a polygenic pleiotropy-informed framework. BMC Genomics. 2018;19:494.

    Article  PubMed  PubMed Central  Google Scholar 

  85. Chung DJ, Yang C, Li C, Gelernter J, Zhao HY. GPA: A Statistical Approach to prioritizing GWAS results by integrating pleiotropy and annotation. PLoS Genet. 2014;10(11):e1004787.

    Article  PubMed  PubMed Central  Google Scholar 

  86. Charles EP. The correction for attenuation due to measurement error: clarifying concepts and creating confidence sets. Psychol Methods. 2005;10(2):206–26.

    Article  PubMed  Google Scholar 

  87. Buonaccorsi JP. Measurement Error: Models, Methods and Applications: Measurement Error: Models, Methods and Applications; 2010.

  88. Lewontin RC, Krakauer J. Distribution of gene frequency as a test of the theory of the selective neutrality of polymorphisms. Genetics. 1973;74(1):175–95.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  89. Holsinger KE, Weir BS. Genetics in geographically structured populations: defining, estimating and interpreting F(ST). Nat Rev Genet. 2009;10(9):639–50.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  90. Zhu ZH, Zhang FT, Hu H, Bakshi A, Robinson MR, Powell JE, Montgomery GW, Goddard ME, Wray NR, Visscher PM, et al. Integration of summary data from GWAS and eQTL studies predicts complex trait gene targets. Nat Genet. 2016;48(5):481–7.

    Article  CAS  PubMed  Google Scholar 

Download references

Acknowledgements

We thank all the GWAS consortia for making summary statistics publicly available for us and are grateful to all the investigators and participants contributed to those studies. The data analyses in the present study were carried out with the high-performance computing cluster that was supported by the special central finance project of local universities for Xuzhou Medical University. We thank two anonymous reviewers for their constructive comments which substantially improved our manuscript.

Funding

The research of Ping Zeng was supported in part by the National Natural Science Foundation of China (82173630 and 81402765), the Youth Foundation of Humanity and Social Science funded by Ministry of Education of China (18YJC910002), the Natural Science Foundation of Jiangsu Province of China (BK20181472), the China Postdoctoral Science Foundation (2018M630607 and 2019T120465), the QingLan Research Project of Jiangsu Province for Young and Middle-aged Academic Leaders, the Six-Talent Peaks Project in Jiangsu Province of China (WSN-087), the Training Project for Youth Teams of Science and Technology Innovation at Xuzhou Medical University (TD202008). The research of Ting Wang was supported in part by the Social Development Project of Xuzhou City (KC20062).

Author information

Author notes

  1. Jinhui Zhang, Shuo Zhang and Jiahao Qiao are co-first authors.

Authors and Affiliations

  1. Department of Biostatistics, School of Public Health, Xuzhou Medical University, Xuzhou, Jiangsu, 221004, China

    Jinhui Zhang, Shuo Zhang, Jiahao Qiao, Ting Wang & Ping Zeng

  2. Center for Medical Statistics and Data Analysis, Xuzhou Medical University, Xuzhou, Jiangsu, 221004, China

    Ping Zeng

  3. Key Laboratory of Human Genetics and Environmental Medicine, Xuzhou Medical University, Xuzhou, Jiangsu, 221004, China

    Ping Zeng

  4. Key Laboratory of Environment and Health, Xuzhou Medical University, Xuzhou, Jiangsu, 221004, China

    Ping Zeng

  5. Engineering Research Innovation Center of Biological Data Mining and Healthcare Transformation, Xuzhou Medical University, Xuzhou, Jiangsu, 221004, China

    Ping Zeng

Authors
  1. Jinhui Zhang
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Shuo Zhang
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Jiahao Qiao
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Ting Wang
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Ping Zeng
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

PZ conceived the idea for the study. PZ and JZ obtained and cleared the datasets; TW, PZ, JQ, SZ and JZ performed the data analyses. TW, PZ, JQ, SZ, and JZ interpreted the results of the data analyses. TW, PZ and JZ wrote the manuscript with the help from other authors.

Corresponding authors

Correspondence to Ting Wang or Ping Zeng.

Ethics declarations

Ethics approval and consent to participate

All methods were carried out in accordance with relevant guidelines and regulations (declaration of Helsinki).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary Material 1

Supplementary Material 2

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Zhang, J., Zhang, S., Qiao, J. et al. Similarity and diversity of genetic architecture for complex traits between East Asian and European populations. BMC Genomics 24, 314 (2023). https://doi.org/10.1186/s12864-023-09434-x

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12864-023-09434-x

Keywords

BMC Genomics

ISSN: 1471-2164

Contact us