Genome-wide analysis of epistasis in body mass index using multiple human populations

Complex genetic architecture of the chicken Growth1 QTL region

The genetic complexity of polygenic traits represents a captivating and intricate facet of biological inheritance. Unlike Mendelian traits controlled by a single gene, polygenic traits are influenced by multiple genetic loci, each exerting a modest effect on the trait. This cumulative impact of numerous genes, interactions among them, environmental factors, and epigenetic modifications results in a multifaceted architecture of genetic contributions to complex traits. Given the well-characterized genome, diverse traits, and range of genetic resources, chicken (Gallus gallus) was employed as a model organism to dissect the intricate genetic makeup of a previously identified major Quantitative Trait Loci (QTL) for body weight on chromosome 1. A multigenerational advanced intercross line (AIL) of 3215 chickens whose genomes had been sequenced to an average of 0.4x was analyzed using genome-wide association study (GWAS) and variance-heterogeneity GWAS (vGWAS) to identify markers associated with 8-week body weight. Additionally, epistatic interactions were studied using the natural and orthogonal interaction (NOIA) model. Six genetic modules, two from GWAS and four from vGWAS, were strongly associated with the studied trait. We found evidence of both additive- and non-additive interactions between these modules and constructed a putative local epistasis network for the region. Our screens for functional alleles revealed a missense variant in the gene ribonuclease H2 subunit B (RNASEH2B), which has previously been associated with growth-related traits in chickens and Darwin's finches. In addition, one of the most strongly associated SNPs identified is located in a non-coding region upstream of the long non-coding RNA, ENSGALG00000053256, previously suggested as a candidate gene for regulating chicken body weight. By studying large numbers of individuals from a family material using approaches to capture both additive and non-additive effects, this study advances our understanding of genetic complexities in a highly polygenic trait and has practical implications for poultry breeding and agriculture.

SORCS2 activity in pancreatic α-cells safeguards insulin granule formation and release from glucose-stressed β-cells

Sorting receptor SORCS2 is a stress-response factor protecting neurons from acute insults, such as during epilepsy. SORCS2 is also expressed in the pancreas, yet its action in this tissue remains unknown. Combining metabolic studies in SORCS2-deficient mice with ex vivo functional analyses and single-cell transcriptomics of pancreatic tissues, we identified a role for SORCS2 in protective stress response in pancreatic islets, essential to sustain insulin release. We show that SORCS2 is predominantly expressed in islet alpha cells. Loss of expression coincides with inability of these cells to produce osteopontin, a secreted factor that facilitates insulin release from stressed beta cells. In line with diminished osteopontin levels, beta cells in SORCS2-deficient islets show gene expression patterns indicative of aggravated cell stress, and exhibit defects in insulin granule maturation and a blunted glucose response. These findings corroborate a function for SORCS2 in protective stress response that extends to metabolism.

Leveraging the genetic correlation between traits improves the detection of epistasis in genome-wide association studies

Epistasis, commonly defined as the interaction between genetic loci, is known to play an important role in the phenotypic variation of complex traits. As a result, many statistical methods have been developed to identify genetic variants that are involved in epistasis, and nearly all of these approaches carry out this task by focusing on analyzing one trait at a time. Previous studies have shown that jointly modeling multiple phenotypes can often dramatically increase statistical power for association mapping. In this study, we present the "multivariate MArginal ePIstasis Test" (mvMAPIT)-a multioutcome generalization of a recently proposed epistatic detection method which seeks to detect marginal epistasis or the combined pairwise interaction effects between a given variant and all other variants. By searching for marginal epistatic effects, one can identify genetic variants that are involved in epistasis without the need to identify the exact partners with which the variants interact-thus, potentially alleviating much of the statistical and computational burden associated with conventional explicit search-based methods. Our proposed mvMAPIT builds upon this strategy by taking advantage of correlation structure between traits to improve the identification of variants involved in epistasis. We formulate mvMAPIT as a multivariate linear mixed model and develop a multitrait variance component estimation algorithm for efficient parameter inference and P-value computation. Together with reasonable model approximations, our proposed approach is scalable to moderately sized genome-wide association studies. With simulations, we illustrate the benefits of mvMAPIT over univariate (or single-trait) epistatic mapping strategies. We also apply mvMAPIT framework to protein sequence data from two broadly neutralizing anti-influenza antibodies and approximately 2,000 heterogeneous stock of mice from the Wellcome Trust Centre for Human Genetics. The mvMAPIT R package can be downloaded at https://github.com/lcrawlab/mvMAPIT.

Indigenous people from Amazon show genetic signatures of pathogen-driven selection

Ecological conditions in the Amazon rainforests are historically favorable for the transmission of numerous tropical diseases, especially vector-borne diseases. The high diversity of pathogens likely contributes to the strong selective pressures for human survival and reproduction in this region. However, the genetic basis of human adaptation to this complex ecosystem remains unclear. This study investigates the possible footprints of genetic adaptation to the Amazon rainforest environment by analyzing the genomic data of 19 native populations. The results based on genomic and functional analysis showed an intense signal of natural selection in a set of genes related to Trypanosoma cruzi infection, which is the pathogen responsible for Chagas disease, a neglected tropical parasitic disease native to the Americas that is currently spreading worldwide.

Concurrent outcomes from multiple approaches of epistasis analysis for human body mass index associated loci provide insights into obesity biology

Genome wide association studies (GWAS) have focused on elucidating the genetic architecture of complex traits by assessing single variant effects in additive genetic models, albeit explaining a fraction of the trait heritability. Epistasis has recently emerged as one of the intrinsic mechanisms that could explain part of this missing heritability. We conducted epistasis analysis for genome-wide body mass index (BMI) associated SNPs in Alzheimer's Disease Neuroimaging Initiative (ADNI) and followed up top significant interacting SNPs for replication in the UK Biobank imputed genotype dataset. We report two pairwise epistatic interactions, between rs2177596 (RHBDD1) and rs17759796 (MAPK1), rs1121980 (FTO) and rs6567160 (MC4R), obtained from a consensus of nine different epistatic approaches. Gene interaction maps and tissue expression profiles constructed for these interacting loci highlights co-expression, co-localisation, physical interaction, genetic interaction, and shared pathways emphasising the neuronal influence in obesity and implicating concerted expression of associated genes in liver, pancreas, and adipose tissues insinuating to metabolic abnormalities characterized by obesity. Detecting epistasis could thus be a promising approach to understand the effect of simultaneously interacting multiple genetic loci in disease aetiology, beyond single locus effects.

Detection of epistasis between ACTN3 and SNAP-25 with an insight towards gymnastic aptitude identification

In this study, we performed an analysis of the impact of performance enhancing polymorphisms (PEPs) on gymnastic aptitude while considering epistatic effects. Seven PEPs (rs1815739, rs8192678, rs4253778, rs6265, rs5443, rs1076560, rs362584) were considered in a case (gymnasts)-control (sedentary individuals) setting. The study sample comprised of two athletes' sets: 27 elite (aged 24.8 ± 2.1 years) and 46 sub-elite (aged 19.7 ± 2.4 years) sportsmen as well as a control group of 245 sedentary individuals (aged 22.5 ± 2.1 years). The DNA was derived from saliva and PEP alleles were determined by PCR, RT-PCR. Following Multifactor Dimensionality Reduction, logistic regression models were built. The synergistic effect for rs1815739 x rs362584 reached 5.43%. The rs1815739 x rs362584 epistatic regression model exhibited a good fit to the data (Chi-squared = 33.758, p ≈ 0) achieving a significant improvement in sportsmen identification over naïve guessing. The area under the receiver operating characteristic curve was 0.715 (Z-score = 38.917, p ≈ 0). In contrast, the additive ACTN3 -SNAP-25 logistic regression model has been verified as non-significant. We demonstrate that a gene involved in the differentiation of muscle architecture-ACTN3 and a gene, which plays an important role in the nervous system-SNAP-25 interact. From the perspective originally established by the Berlin Academy of Science in 1751, the matter of communication between the brain and muscles via nerves adopts molecular manifestations. Further in-vitro investigations are required to explain the molecular details of the rs1815739 -rs362584 interaction.

Genome-Wide Interaction and Pathway Association Studies for Body Mass Index

Objective: We investigated gene interactions (epistasis) for body mass index (BMI) in a European-American adult female cohort via genome-wide interaction analyses (GWIA) and pathway association analyses. Methods: Genome-wide pairwise interaction analyses were carried out for BMI in 493 extremely obese cases (BMI > 35 kg/m2) and 537 never-overweight controls (BMI < 25 kg/m2). To further validate the results, specific SNPs were selected based on the GWIA results for haplotype-based association studies. Pathway-based association analyses were performed using a modified Gene Set Enrichment Algorithm (GSEA) (GenGen program) to further explore BMI-related pathways using our genome wide association study (GWAS) data set, GIANT, ENGAGE, and DIAGRAM Consortia. Results: The EXOC4-1q23.1 interaction was associated with BMI, with the most significant epistasis between rs7800006 and rs10797020 (P = 2.63 × 10-11). In the pathway-based association analysis, Tob1 pathway showed the most significant association with BMI (empirical P < 0.001, FDR = 0.044, FWER = 0.040). These findings were further validated in different populations. Conclusion: Genome-wide pairwise SNP-SNP interaction and pathway analyses suggest that EXOC4 and TOB1-related pathways may contribute to the development of obesity.

Clinical Implications of Single Nucleotide Polymorphisms in Diagnosis of Asthma and its Subtypes

For the past three decades, a large number of genetic studies have been performed to examine genetic variants associated with asthma and its subtypes in hopes of gaining better understanding of the mechanisms underlying disease pathology and to identify genetic biomarkers predictive of disease outcomes. Various methods have been used to achieve these objectives, including linkage analysis, candidate gene polymorphism analysis, and genome-wide association studies (GWAS); however, the degree to which genetic variants contribute to asthma pathogenesis has proven to be much less significant than originally expected. Subsequent application of GWAS to well-defined phenotypes, such as occupational asthma and non-steroidal anti-inflammatory drugexacerbated respiratory diseases, has overcome some of these limitations, although with only partial success. Recently, a combinatorial analysis of single nucleotide polymorphisms (SNPs) identified by GWAS has been used to develop sets of genetic markers able to more accurately stratify asthma subtypes. In this review, we discuss the implications of the identified SNPs in diagnosis of asthma and its subtypes and the progress being made in combinatorial analysis of genetic variants.

Genome-wide search for higher order epistasis as modifiers of treatment effects on bone mineral density in childhood cancer survivors

Single-nucleotide polymorphisms (SNPs) contributing to interactions between regulatory elements that modulate gene transcription may explain some of the uncharacterized variation for complex traits. We explored this hypothesis among 856 adult survivors of pediatric cancer exposed to curative treatments that adversely affect bone mineral density (BMD). To restrict our search to interactions among SNPs in regulatory elements, our analysis considered 75523 SNPs mapped to putative promoter or enhancer regions. In anticipation that power to detect higher order epistasis would be low using an exhaustive search and a Bonferroni-corrected threshold for genome-wide significance (e.g., P < 5.6 × 10^-14), a novel non-exhaustive statistical algorithm was implemented to detect chromosome-wide three-way regulatory interactions. We used a permutation-based evaluation statistic to identify candidate SNP interactions with stronger associations with BMD than expected. Of the six regulatory 3-SNP interactions identified as candidate interactions (P < 3.5 × 10^-11) among cancer survivors exposed to treatments, five were replicated in an independent cohort of survivors (N = 1428) as modifiers of treatment effects on BMD (P < 0.05). Analyses with publicly available bioinformatics data revealed that SNPs contributing to replicated interactions were enriched for gene expressions (P = 3.6 × 10^-4) and enhancer states (P < 0.05) in cells relevant for bone biology. For each replicated interaction, implicated SNPs were within or directly adjacent to 100-kb windows of genomic regions that plausibly physically interact in lymphoblastoid cells. Our study demonstrates the utility of a hypothesis-driven approach in revealing epistasis associated with complex traits.

Influence of SNP*SNP interaction on BMI in European American adolescents: findings from the National Longitudinal Study of Adolescent Health

BACKGROUND

Adolescent obesity is predictive of future weight gain, obesity and adult onset severe obesity (body mass index [BMI] ≥40 kg m(-2) ). Despite successful efforts to identify Single Nucleotide Polymorphisms (SNPs) influencing BMI, <5% of the 40-80% heritability of the phenotype has been explained. Identification of gene-gene (G-G) interactions between known variants can help explain this hidden heritability as well as identify potential biological mechanisms affecting weight gain during this critical developmental period.

OBJECTIVE

We have recently shown distinct genetic effects on BMI across the life course, and thus it is important to examine the evidence for epistasis in adolescence.

METHODS

In adolescent participants of European descent from wave II of the National Longitudinal Study of Adolescent Health (Add Health, n = 5072, ages 12-21, 52.5% female), we tested 34 established BMI-related SNPs for G-G interaction effects on BMI z-score. We used mixed-effects regression, assuming multiplicative interaction models adjusting for age, sex and geographic region, with random effects for family and school.

RESULTS

For 28 G-G interactions that were nominally significant (P < 0.05), we attempted to replicate our results in an adolescent sample from the Childhood European American Cohort from Philadelphia. In the replication study, one interaction (PRKD1-FTO) was significant after correction for multiple testing.

CONCLUSIONS

Our results are suggestive of epistatic effects on BMI during adolescence and point to potentially interactive effects between genes in biological pathways important in obesity.

Paradoxical Relationship Between Body Mass Index and Thyroid Hormone Levels: A Study Using Mendelian Randomization

CONTEXT

Free T3 (FT3) has been positively associated with body mass index (BMI) in cross-sectional studies in healthy individuals. This is difficult to reconcile with clinical findings in pathological thyroid dysfunction.

OBJECTIVE

We aimed to investigate whether childhood adiposity influences FT3 levels.

DESIGN

Mendelian randomization using genetic variants robustly associated with BMI.

SETTING

Avon Longitudinal Study of Parents and Children, a population-based birth cohort.

PARTICIPANTS

A total of 3014 children who had thyroid function measured at age 7, who also underwent dual x-ray absorptiometry scans at ages 9.9 and 15.5 years and have genetic data available.

MAIN OUTCOME MEASURES

FT3.

RESULTS

Observationally at age 7 years, BMI was positively associated with FT3: β-standardized (β-[std]) = 0.12 (95% confidence interval [CI]: 0.08, 0.16), P = 4.02 × 10(-10); whereas FT4 was negatively associated with BMI: β-(std) = -0.08 (95% CI: -0.12, -0.04), P = 3.00 × 10(-5). These differences persisted after adjustment for age, sex, and early life environment. Genetic analysis indicated 1 allele change in BMI allelic score was associated with a 0.04 (95% CI: 0.03, 0.04) SD increase in BMI (P = 6.41 × 10(-17)). At age 7, a genetically determined increase in BMI of 1.89 kg/m(2) was associated with a 0.22 pmol/L (95% CI: 0.07, 0.36) increase in FT3 (P = .004) but no substantial change in FT4 0.01 mmol/L, (95% CI: -0.37, 0.40), P = .96.

CONCLUSION

Our analysis shows that children with a genetically higher BMI had higher FT3 but not FT4 levels, indicating that higher BMI/fat mass has a causal role in increasing FT3 levels. This may explain the paradoxical associations observed in observational analyses. Given rising childhood obesity levels, this relationship merits closer scrutiny.

Analysis of gene-gene interactions among common variants in candidate cardiovascular genes in coronary artery disease

Objective

Only a small fraction of coronary artery disease (CAD) heritability has been explained by common variants identified to date. Interactions between genes of importance to cardiovascular regulation may account for some of the missing heritability of CAD. This study aimed to investigate the role of gene-gene interactions in common variants in candidate cardiovascular genes in CAD.

Approach and Results

2,101 patients with CAD from the British Heart Foundation Family Heart Study and 2,426 CAD-free controls were included in the discovery cohort. All subjects were genotyped with the Illumina HumanCVD BeadChip enriched for genes and pathways relevant to the cardiovascular system and disease. The primary analysis in the discovery cohort examined pairwise interactions among 913 common (minor allele frequency >0.1) independent single nucleotide polymorphisms (SNPs) with at least nominal association with CAD in single locus analysis. A secondary exploratory interaction analysis was performed among all 11,332 independent common SNPs surviving quality control criteria. Replication analyses were conducted in 2,967 patients and 3,075 controls from the Myocardial Infarction Genetics Consortium. None of the interactions amongst 913 SNPs analysed in the primary analysis was statistically significant after correction for multiple testing (required P<1.2x10^-7). Similarly, none of the pairwise gene-gene interactions in the secondary analysis reached statistical significance after correction for multiple testing (required P = 7.8x10^-10). None of 36 suggestive interactions from the primary analysis or 31 interactions from the secondary analysis was significant in the replication cohort. Our study had 80% power to detect odds ratios > 1.7 for common variants in the primary analysis.

Conclusions

Moderately large additive interactions between common SNPs in genes relevant to cardiovascular disease do not appear to play a major role in genetic predisposition to CAD. The role of genetic interactions amongst less common SNPs and with medium and small magnitude effects remain to be investigated.

The role of the retromer complex in aging-related neurodegeneration: a molecular and genomic review

The retromer coat complex is a vital component of the intracellular trafficking mechanism sorting cargo from the endosomes to the trans-Golgi network or to the cell surface. In recent years, genes encoding components of the retromer coat complex and members of the vacuolar protein sorting 10 (Vps10) family of receptors, which play pleiotropic functions in protein trafficking and intracellular/intercellular signaling in neuronal and non-neuronal cells and are primary cargos of the retromer complex, have been implicated as genetic risk factors for sporadic and autosomal dominant forms of several neurodegenerative diseases, including Alzheimer's disease, Parkinson's disease and frontotemporal lobar degeneration. In addition to their functions in protein trafficking, the members of the Vps10 receptor family (sortilin, SorL1, SorCS1, SorCS2, and SorCS3) modulate neurotrophic signaling pathways. Both sortilin and SorCS2 act as cell surface receptors to mediate acute responses to proneurotrophins. In addition, sortilin can modulate the intracellular response to brain-derived neurotrophic factor (BDNF) by direct control of BDNF levels and regulating anterograde trafficking of Trk receptors to the synapse. This review article summarizes the emerging data from this rapidly growing field of intracellular trafficking signaling in the pathogenesis of neurodegeneration.

Detecting epistasis in human complex traits

Genome-wide association studies (GWASs) have become the focus of the statistical analysis of complex traits in humans, successfully shedding light on several aspects of genetic architecture and biological aetiology. Single-nucleotide polymorphisms (SNPs) are usually modelled as having additive, cumulative and independent effects on the phenotype. Although evidently a useful approach, it is often argued that this is not a realistic biological model and that epistasis (that is, the statistical interaction between SNPs) should be included. The purpose of this Review is to summarize recent directions in methodology for detecting epistasis and to discuss evidence of the role of epistasis in human complex trait variation. We also discuss the relevance of epistasis in the context of GWASs and potential hazards in the interpretation of statistical interaction terms.

Abundant local interactions in the 4p16.1 region suggest functional mechanisms underlying SLC2A9 associations with human serum uric acid

Human serum uric acid concentration (SUA) is a complex trait. A recent meta-analysis of multiple genome-wide association studies (GWAS) identified 28 loci associated with SUA jointly explaining only 7.7% of the SUA variance, with 3.4% explained by two major loci (SLC2A9 and ABCG2). Here we examined whether gene-gene interactions had any roles in regulating SUA using two large GWAS cohorts included in the meta-analysis [the Atherosclerosis Risk in Communities study cohort (ARIC) and the Framingham Heart Study cohort (FHS)]. We found abundant genome-wide significant local interactions in ARIC in the 4p16.1 region located mostly in an intergenic area near SLC2A9 that were not driven by linkage disequilibrium and were replicated in FHS. Taking the forward selection approach, we constructed a model of five SNPs with marginal effects and three epistatic SNP pairs in ARIC-three marginal SNPs were located within SLC2A9 and the remaining SNPs were all located in the nearby intergenic area. The full model explained 1.5% more SUA variance than that explained by the lead SNP alone, but only 0.3% was contributed by the marginal and epistatic effects of the SNPs in the intergenic area. Functional analysis revealed strong evidence that the epistatically interacting SNPs in the intergenic area were unusually enriched at enhancers active in ENCODE hepatic (HepG2, P = 4.7E-05) and precursor red blood (K562, P = 5.0E-06) cells, putatively regulating transcription of WDR1 and SLC2A9. These results suggest that exploring epistatic interactions is valuable in uncovering the complex functional mechanisms underlying the 4p16.1 region.

Analysis of multiple association studies provides evidence of an expression QTL hub in gene-gene interaction network affecting HDL cholesterol levels

Epistasis has been suggested to underlie part of the missing heritability in genome-wide association studies. In this study, we first report an analysis of gene-gene interactions affecting HDL cholesterol (HDL-C) levels in a candidate gene study of 2,091 individuals with mixed dyslipidemia from a clinical trial. Two additional studies, the Atherosclerosis Risk in Communities study (ARIC; n = 9,713) and the Multi-Ethnic Study of Atherosclerosis (MESA; n = 2,685), were considered for replication. We identified a gene-gene interaction between rs1532085 and rs12980554 (P = 7.1×10⁻⁷) in their effect on HDL-C levels, which is significant after Bonferroni correction (P_c = 0.017) for the number of SNP pairs tested. The interaction successfully replicated in the ARIC study (P = 7.0×10⁻⁴; P_c = 0.02). Rs1532085, an expression QTL (eQTL) of LIPC, is one of the two SNPs involved in another, well-replicated gene-gene interaction underlying HDL-C levels. To further investigate the role of this eQTL SNP in gene-gene interactions affecting HDL-C, we tested in the ARIC study for interaction between this SNP and any other SNP genome-wide. We found the eQTL to be involved in a few suggestive interactions, one of which significantly replicated in MESA. Importantly, these gene-gene interactions, involving only rs1532085, explain an additional 1.4% variation of HDL-C, on top of the 0.65% explained by rs1532085 alone. LIPC plays a key role in the lipid metabolism pathway and it, and rs1532085 in particular, has been associated with HDL-C and other lipid levels. Collectively, we discovered several novel gene-gene interactions, all involving an eQTL of LIPC, thus suggesting a hub role of LIPC in the gene-gene interaction network that regulates HDL-C levels, which in turn raises the hypothesis that LIPC's contribution is largely via interactions with other lipid metabolism related genes.

Weighted Interaction SNP Hub (WISH) network method for building genetic networks for complex diseases and traits using whole genome genotype data

Background

High-throughput genotype (HTG) data has been used primarily in genome-wide association (GWA) studies; however, GWA results explain only a limited part of the complete genetic variation of traits. In systems genetics, network approaches have been shown to be able to identify pathways and their underlying causal genes to unravel the biological and genetic background of complex diseases and traits, e.g., the Weighted Gene Co-expression Network Analysis (WGCNA) method based on microarray gene expression data. The main objective of this study was to develop a scale-free weighted genetic interaction network method using whole genome HTG data in order to detect biologically relevant pathways and potential genetic biomarkers for complex diseases and traits.

Results

We developed the Weighted Interaction SNP Hub (WISH) network method that uses HTG data to detect genome-wide interactions between single nucleotide polymorphism (SNPs) and its relationship with complex traits. Data dimensionality reduction was achieved by selecting SNPs based on its: 1) degree of genome-wide significance and 2) degree of genetic variation in a population. Network construction was based on pairwise Pearson's correlation between SNP genotypes or the epistatic interaction effect between SNP pairs. To identify modules the Topological Overlap Measure (TOM) was calculated, reflecting the degree of overlap in shared neighbours between SNP pairs. Modules, clusters of highly interconnected SNPs, were defined using a tree-cutting algorithm on the SNP dendrogram created from the dissimilarity TOM (1-TOM). Modules were selected for functional annotation based on their association with the trait of interest, defined by the Genome-wide Module Association Test (GMAT). We successfully tested the established WISH network method using simulated and real SNP interaction data and GWA study results for carcass weight in a pig resource population; this resulted in detecting modules and key functional and biological pathways related to carcass weight.

Conclusions

We developed the WISH network method which is a novel 'systems genetics' approach to study genetic networks underlying complex trait variation. The WISH network method reduces data dimensionality and statistical complexity in associating genotypes with phenotypes in GWA studies and enables researchers to identify biologically relevant pathways and potential genetic biomarkers for any complex trait of interest.

How genome-wide SNP-SNP interactions relate to nasopharyngeal carcinoma susceptibility

This study is the first to use genome-wide association study (GWAS) data to evaluate the multidimensional genetic architecture underlying nasopharyngeal cancer. Since analysis of data from GWAS confirms a close and consistent association between elevated risk for nasopharyngeal carcinoma (NPC) and major histocompatibility complex class 1 genes, our goal here was to explore lesser effects of gene-gene interactions. We conducted an exhaustive genome-wide analysis of GWAS data of NPC, revealing two-locus interactions occurring between single nucleotide polymorphisms (SNPs), and identified a number of suggestive interaction loci which were missed by traditional GWAS analyses. Although none of the interaction pairs we identified passed the genome-wide Bonferroni-adjusted threshold for significance, using independent GWAS data from the same population (Stage 2), we selected 66 SNP pairs in 39 clusters with P<0.01. We identified that in several chromosome regions, multiple suggestive interactions group to form a block-like signal, effectively reducing the rate of false discovery. The strongest cluster of interactions involved the CREB5 gene and a SNP rs1607979 on chromosome 17q22 (P = 9.86×10(-11)) which also show trans-expression quantitative loci (eQTL) association in Chinese population. We then detected a complicated cis-interaction pattern around the NPC-associated HLA-B locus, which is immediately adjacent to copy-number variations implicated in male susceptibility for NPC. While it remains to be seen exactly how and to what degree SNP-SNP interactions such as these affect susceptibility for nasopharyngeal cancer, future research on these questions holds great promise for increasing our understanding of this disease's genetic etiology, and possibly also that of other gene-related cancers.

Properties of local interactions and their potential value in complementing genome-wide association studies

Local interactions between neighbouring SNPs are hypothesized to be able to capture variants missing from genome-wide association studies (GWAS) via haplotype effects but have not been thoroughly explored. We have used a new high-throughput analysis tool to probe this underexplored area through full pair-wise genome scans and conventional GWAS in diastolic and systolic blood pressure and six metabolic traits in the Northern Finland Birth Cohort 1966 (NFBC1966) and the Atherosclerosis Risk in Communities study cohort (ARIC). Genome-wide significant interactions were detected in ARIC for systolic blood pressure between PLEKHA7 (a known GWAS locus for blood pressure) and GPR180 (which plays a role in vascular remodelling), and also for triglycerides as local interactions within the 11q23.3 region (replicated significantly in NFBC1966), which notably harbours several loci (BUD13, ZNF259 and APOA5) contributing to triglyceride levels. Tests of the local interactions within the 11q23.3 region conditional on the top GWAS signal suggested the presence of two independent functional variants, each with supportive evidence for their roles in gene regulation. Local interactions captured 9 additional GWAS loci identified in this study (3 significantly replicated) and 73 from previous GWAS (24 in the eight traits and 49 in related traits). We conclude that the detection of local interactions requires adequate SNP coverage of the genome and that such interactions are only likely to be detectable between SNPs in low linkage disequilibrium. Analysing local interactions is a potentially valuable complement to GWAS and can provide new insights into the biology underlying variation in complex traits.

Gene-based testing of interactions in association studies of quantitative traits

Various methods have been developed for identifying gene–gene interactions in genome-wide association studies (GWAS). However, most methods focus on individual markers as the testing unit, and the large number of such tests drastically erodes statistical power. In this study, we propose novel interaction tests of quantitative traits that are gene-based and that confer advantage in both statistical power and biological interpretation. The framework of gene-based gene–gene interaction (GGG) tests combine marker-based interaction tests between all pairs of markers in two genes to produce a gene-level test for interaction between the two. The tests are based on an analytical formula we derive for the correlation between marker-based interaction tests due to linkage disequilibrium. We propose four GGG tests that extend the following P value combining methods: minimum P value, extended Simes procedure, truncated tail strength, and truncated P value product. Extensive simulations point to correct type I error rates of all tests and show that the two truncated tests are more powerful than the other tests in cases of markers involved in the underlying interaction not being directly genotyped and in cases of multiple underlying interactions. We applied our tests to pairs of genes that exhibit a protein–protein interaction to test for gene-level interactions underlying lipid levels using genotype data from the Atherosclerosis Risk in Communities study. We identified five novel interactions that are not evident from marker-based interaction testing and successfully replicated one of these interactions, between SMAD3 and NEDD9, in an independent sample from the Multi-Ethnic Study of Atherosclerosis. We conclude that our GGG tests show improved power to identify gene-level interactions in existing, as well as emerging, association studies.

Epistasis is likely to play a significant role in complex diseases or traits and is one of the many possible explanations for “missing heritability.” However, epistatic interactions have been difficult to detect in genome-wide association studies (GWAS) due to the limited power caused by the multiple-testing correction from the large number of tests conducted. Gene-based gene–gene interaction (GGG) tests might hold the key to relaxing the multiple-testing correction burden and increasing the power for identifying epistatic interactions in GWAS. Here, we developed GGG tests of quantitative traits by extending four P value combining methods and evaluated their type I error rates and power using extensive simulations. All four GGG tests are more powerful than a principal component-based test. We also applied our GGG tests to data from the Atherosclerosis Risk in Communities study and found five gene-level interactions associated with the levels of total cholesterol and high-density lipoprotein cholesterol (HDL-C). One interaction between SMAD3 and NEDD9 on HDL-C was further replicated in an independent sample from the Multi-Ethnic Study of Atherosclerosis.

Identification of shared genetic susceptibility locus for coronary artery disease, type 2 diabetes and obesity: a meta-analysis of genome-wide studies

Type 2 diabetes (2DM), obesity, and coronary artery disease (CAD) are frequently coexisted being as key components of metabolic syndrome. Whether there is shared genetic background underlying these diseases remained unclear. We performed a meta-analysis of 35 genome screens for 2DM, 36 for obesity or body mass index (BMI)-defined obesity, and 21 for CAD using genome search meta-analysis (GSMA), which combines linkage results to identify regions with only weak evidence and provide genetic interactions among different diseases. For each study, 120 genomic bins of approximately 30 cM were defined and ranked according to the best linkage evidence within each bin. For each disease, bin 6.2 achieved genomic significanct evidence, and bin 9.3, 10.5, 16.3 reached suggestive level for 2DM. Bin 11.2 and 16.3, and bin 10.5 and 9.3, reached suggestive evidence for obesity and CAD respectively. In pooled all three diseases, bin 9.3 and 6.5 reached genomic significant and suggestive evidence respectively, being relatively much weaker for 2DM/CAD or 2DM/obesity or CAD/obesity. Further, genomewide significant evidence was observed of bin 16.3 and 4.5 for 2DM/obesity, which is decreased when CAD was added. These findings indicated that bin 9.3 and 6.5 are most likely to be shared by 2DM, obesity and CAD. And bin 16.3 and 4.5 are potentially common regions to 2DM and obesity only. The observed shared susceptibility regions imply a partly overlapping genetic aspects of disease development. Fine scanning of these regions will definitely identify more susceptibility genes and causal variants.

BiForce Toolbox: powerful high-throughput computational analysis of gene-gene interactions in genome-wide association studies

Genome-wide association studies (GWAS) have discovered many loci associated with common disease and quantitative traits. However, most GWAS have not studied the gene–gene interactions (epistasis) that could be important in complex trait genetics. A major challenge in analysing epistasis in GWAS is the enormous computational demands of analysing billions of SNP combinations. Several methods have been developed recently to address this, some using computers equipped with particular graphical processing units, most restricted to binary disease traits and all poorly suited to general usage on the most widely used operating systems. We have developed the BiForce Toolbox to address the demand for high-throughput analysis of pairwise epistasis in GWAS of quantitative and disease traits across all commonly used computer systems. BiForce Toolbox is a stand-alone Java program that integrates bitwise computing with multithreaded parallelization and thus allows rapid full pairwise genome scans via a graphical user interface or the command line. Furthermore, BiForce Toolbox incorporates additional tests of interactions involving SNPs with significant marginal effects, potentially increasing the power of detection of epistasis. BiForce Toolbox is easy to use and has been applied in multiple studies of epistasis in large GWAS data sets, identifying interesting interaction signals and pathways.

Knowledge-driven analysis identifies a gene-gene interaction affecting high-density lipoprotein cholesterol levels in multi-ethnic populations

Total cholesterol, low-density lipoprotein cholesterol, triglyceride, and high-density lipoprotein cholesterol (HDL-C) levels are among the most important risk factors for coronary artery disease. We tested for gene–gene interactions affecting the level of these four lipids based on prior knowledge of established genome-wide association study (GWAS) hits, protein–protein interactions, and pathway information. Using genotype data from 9,713 European Americans from the Atherosclerosis Risk in Communities (ARIC) study, we identified an interaction between HMGCR and a locus near LIPC in their effect on HDL-C levels (Bonferroni corrected P_c = 0.002). Using an adaptive locus-based validation procedure, we successfully validated this gene–gene interaction in the European American cohorts from the Framingham Heart Study (P_c = 0.002) and the Multi-Ethnic Study of Atherosclerosis (MESA; P_c = 0.006). The interaction between these two loci is also significant in the African American sample from ARIC (P_c = 0.004) and in the Hispanic American sample from MESA (P_c = 0.04). Both HMGCR and LIPC are involved in the metabolism of lipids, and genome-wide association studies have previously identified LIPC as associated with levels of HDL-C. However, the effect on HDL-C of the novel gene–gene interaction reported here is twice as pronounced as that predicted by the sum of the marginal effects of the two loci. In conclusion, based on a knowledge-driven analysis of epistasis, together with a new locus-based validation method, we successfully identified and validated an interaction affecting a complex trait in multi-ethnic populations.

Genome-wide association studies (GWAS) have identified many loci associated with complex human traits or diseases. However, the fraction of heritable variation explained by these loci is often relatively low. Gene–gene interactions might play a significant role in complex traits or diseases and are one of the many possible factors contributing to the missing heritability. However, to date only a few interactions have been found and validated in GWAS due to the limited power caused by the need for multiple-testing correction for the very large number of tests conducted. Here, we used three types of prior knowledge, known GWAS hits, protein–protein interactions, and pathway information, to guide our search for gene–gene interactions affecting four lipid levels. We identified an interaction between HMGCR and a locus near LIPC in their effect on high-density lipoprotein cholesterol (HDL-C) and another pair of loci that interact in their effect on low-density lipoprotein cholesterol (LDL-C). We validated the interaction on HDL-C in a number of independent multiple-ethnic populations, while the interaction underlying LDL-C did not validate. The prior knowledge-driven searching approach and a locus-based validation procedure show the potential for dissecting and validating gene–gene interactions in current and future GWAS.

High-throughput analysis of epistasis in genome-wide association studies with BiForce

MOTIVATION

Gene-gene interactions (epistasis) are thought to be important in shaping complex traits, but they have been under-explored in genome-wide association studies (GWAS) due to the computational challenge of enumerating billions of single nucleotide polymorphism (SNP) combinations. Fast screening tools are needed to make epistasis analysis routinely available in GWAS.

RESULTS

We present BiForce to support high-throughput analysis of epistasis in GWAS for either quantitative or binary disease (case-control) traits. BiForce achieves great computational efficiency by using memory efficient data structures, Boolean bitwise operations and multithreaded parallelization. It performs a full pair-wise genome scan to detect interactions involving SNPs with or without significant marginal effects using appropriate Bonferroni-corrected significance thresholds. We show that BiForce is more powerful and significantly faster than published tools for both binary and quantitative traits in a series of performance tests on simulated and real datasets. We demonstrate BiForce in analysing eight metabolic traits in a GWAS cohort (323 697 SNPs, >4500 individuals) and two disease traits in another (>340 000 SNPs, >1750 cases and 1500 controls) on a 32-node computing cluster. BiForce completed analyses of the eight metabolic traits within 1 day, identified nine epistatic pairs of SNPs in five metabolic traits and 18 SNP pairs in two disease traits. BiForce can make the analysis of epistasis a routine exercise in GWAS and thus improve our understanding of the role of epistasis in the genetic regulation of complex traits.

AVAILABILITY AND IMPLEMENTATION

The software is free and can be downloaded from http://bioinfo.utu.fi/BiForce/.

CONTACT

wenhua.wei@igmm.ed.ac.uk

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.