A versatile, fast and unbiased method for estimation of gene-by-environment interaction effects on biobank-scale datasets

Di Scipio, Matteo; Khan, Mohammad; Mao, Shihong; Chong, Michael; Judge, Conor; Pathan, Nazia; Perrot, Nicolas; Nelson, Walter; Lali, Ricky; Di, Shuang; Morton, Robert; Petch, Jeremy; Paré, Guillaume

doi:10.1038/s41467-023-40913-7

Cited by 9 publications

(11 citation statements)

References 53 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…First, we distinguish three different types of GxE interaction: Imperfect genetic correlation, Varying genetic variance, and Proportional amplification (Figure 1a; also see Supplementary Figure 6). Second, most variance components methods for detecting genome-wide GxE [13][14][15]17,18 cannot detect genome-wide GxE unless SNP-heritability varies across E bins (Scenario 2). An exception is GxEMM 16 , which detects other types of GxE by explicitly modeling genetic and environmental variance that varies with the E variable; however, GxEMM is less computationally tractable and generally less powerful than our framework (Supplementary Figure 2).…”

Section: Discussionmentioning

confidence: 99%

“…Previous studies have detected GxE at a limited number of specific loci [5][6][7] (including studies that associated genotype to phenotypic variance without knowing the underlying E variable [8][9][10][11][12] ). Previous studies have also proposed variance components methods for detecting genome-wide contributions of GxE to disease heritability [13][14][15][16][17][18] , but these methods have not been applied at biobank scale across a broad range of disease/traits. Thus, the overall contribution of GxE to disease/trait architectures is currently unknown.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Distinct explanations underlie gene-environment interactions in the UK Biobank

Durvasula,

Price

2023

Preprint

View full text Add to dashboard Cite

The role of gene-environment (GxE) interaction in disease and complex trait architectures is widely hypothesized, but currently unknown. Here, we apply three statistical approaches to quantify and distinguish three different types of GxE interaction for a given disease/trait and E variable. First, we detect locus-specific GxE interaction by testing for genetic correlation (rg) < 1 across E bins. Second, we detect genome-wide effects of the E variable on genetic variance by leveraging polygenic risk scores (PRS) to test for significant PRSxE in a regression of phenotypes on PRS, E, and PRSxE, together with differences in SNP-heritability across E bins. Third, we detect genome-wide proportional amplification of genetic and environmental effects as a function of the E variable by testing for significant PRSxE with no differences in SNP-heritability across E bins. Simulations show that these approaches achieve high sensitivity and specificity in distinguishing these three GxE scenarios. We applied our framework to 33 UK Biobank diseases/traits (average N=325K) and 10 E variables spanning lifestyle, diet, and other environmental exposures. First, we identified 19 trait-E pairs with rg significantly < 1 (FDR<5%) (average rg=0.95); for example, white blood cell count had rg=0.95 (s.e. 0.01) between smokers and non-smokers. Second, we identified 28 trait-E pairs with significant PRSxE and significant SNP-heritability differences across E bins; for example, type 2 diabetes had a significant PRSxE for alcohol consumption (P=1e-13) with 4.2x larger SNP-heritability in the largest versus smallest quintiles of alcohol consumption (P<1e-16). Third, we identified 15 trait-E pairs with significant PRSxE with no SNP-heritability differences across E bins; for example, triglyceride levels had a significant PRSxE effect for composite diet score (P=4e-5) with no SNP-heritability differences. Analyses using biological sex as the E variable produced additional significant findings in each of the three scenarios. Overall, we infer a substantial contribution of GxE and GxSex effects to disease and complex trait variance.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Distinct explanations underlie gene-environment interactions in the UK Biobank

Durvasula,

Price

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…To analyze binary outcomes, we transformed the observed-scale heritability to liability-scale heritability using Roberston transformation 18 . It has been pointed out that when the GE interaction variance is large, the normality assumption of the phenotype liability may be violated, resulting in biased results of Roberston transformation 8,38 . However, our simulation results showed that when the GE interaction variance proportion was small Roberston transformation still yielded reasonably accurate result (S Figure 1, S Table 1), consistent with previous studies 8 .…”

Section: Methodsmentioning

confidence: 99%

LDER-GE estimates phenotypic variance component of gene-environment interactions in human complex traits accurately with GE interaction summary statistics and full LD information

Dong,

Jiang,

et al. 2023

Preprint

View full text Add to dashboard Cite

Gene-environment (GE) interactions are essential in understanding human complex traits. Identifying these interactions is necessary for deciphering the biological basis of such traits. In this study, we introduce a statistical method Linkage-Disequilibrium Eigenvalue Regression for Gene-Environment interactions (LDER-GE). LDER-GE improves the accuracy of estimating the phenotypic variance component explained by genome-wide GE interactions using large-scale biobank association summary statistics. LDER-GE leverages the complete Linkage Disequilibrium (LD) matrix, as opposed to only the diagonal squared LD matrix utilized by LDSC (Linkage Disequilibrium Score)-based methods. Our extensive simulation studies demonstrate that LDER-GE performs better than LDSC-based approaches by enhancing statistical efficiency by approximately 23%. This improvement is equivalent to a sample size increase of around 51%. Additionally, LDER-GE effectively controls type-I error rate and produces unbiased results. We conducted an analysis using UK Biobank data, comprising 307,259 unrelated European-Ancestry subjects and 966,766 variants, across 151 environmental covariate-phenotype (E-Y) pairs. LDER-GE identified 35 significant E-Y pairs while LDSC-based method only identified 25 significant E-Y pairs with 23 overlapped with LDER-GE. Furthermore, we employed LDER-GE to estimate the aggregated variance component attributed to multiple GE interactions, leading to an increase in the explained phenotypic variance with GE interactions compared to considering main genetic effects only. Our results suggest the importance of impacts of GE interactions on human complex traits.

show abstract

“…Broad-scale sequencing of wild populations Many computational methods available (e.g., GREML Yang et al, 2015;Di Scipio et al, 2023), which can parse genetic, environmental and their interaction using SNP matrices.…”

Section: Reciprocal Common Gardensmentioning

confidence: 99%

The drivers of intraspecific trait variation and their implications for future tree productivity and survival

Blumstein

2024

American J of Botany

View full text Add to dashboard Cite

Forests are facing unprecedented levels of stress from pest and disease outbreaks, disturbance, fragmentation, development, and a changing climate. These selective agents act to alter forest composition from regional to cellular levels. Thus, a central challenge for understanding how forests will be impacted by future change is how to integrate across scales of biology. Phenotype, or an observable trait, is the product of an individual's genes (G) and the environment in which an organism lives (E). To date, researchers have detailed how environment drives variation in tree phenotypes over long time periods (e.g., long‐term ecological research sites [LTERs]) and across large spatial scales (e.g., flux network). In parallel, researchers have discovered the genes and pathways that govern phenotypes, finding high degrees of genetic control and signatures of local adaptation in many plant traits. However, the research in these two areas remain largely independent of each other, hindering our ability to generate accurate predictions of plant response to environment, an increasingly urgent need given threats to forest systems. I present the importance of both genes and environment in determining tree responses to climate stress. I highlight why the difference between G versus E in driving variation is critical for our understanding of climate responses, then propose means of accelerating research that examines G and E simultaneously by leveraging existing long‐term, large‐scale phenotypic data sets from ecological networks and adding newly affordable sequence (‐omics) data to both drill down to find the genes and alleles influencing phenotypes and scale up to find how patterns of demography and local adaptation may influence future response to change.

show abstract

A versatile, fast and unbiased method for estimation of gene-by-environment interaction effects on biobank-scale datasets

Cited by 9 publications

References 53 publications

Distinct explanations underlie gene-environment interactions in the UK Biobank

Distinct explanations underlie gene-environment interactions in the UK Biobank

LDER-GE estimates phenotypic variance component of gene-environment interactions in human complex traits accurately with GE interaction summary statistics and full LD information

The drivers of intraspecific trait variation and their implications for future tree productivity and survival

Contact Info

Product

Resources

About