Large-scale Genome-wide Association Study Data Research Articles

Schizophrenia and risk preference: a bidirectional two-sample mendelian randomization study.

Increasing evidence shows that risk preference is associated with schizophrenia. However, the causality and direction of this association are not clear; Therefore, we used Mendelian randomization (MR) to examine the potential bidirectional relationship between risk preference and schizophrenia. Genome-wide association studies (GWAS) summary data on risk preference of 939,908 participants from the UK Biobank and 23andMe were used to identify general risk preference. Data from 320,404 subjects (76,755 cases and 243,649 controls) from The Psychiatric Genomics Consortium were used to identify schizophrenia. The weighted median (WM), the inverse variance weighted (IVW), and the Mendelian randomization-Egger (MR-Egger) methods were used for the MR analysis to estimate the causal effect and detect the directional pleiotropy. The GWAS summary data were respectively from two combined samples, containing 939,908 and 320,404 subjects of European ancestry. Mendelian randomization evidence suggested that risk preference was associated with increased onset of schizophrenia (OR = 2.84, 95CI%: 1.77-4.56, P = 1.58*10 - 5) and that schizophrenia was also associated with raised risk preference (OR = 1.11, 95CI%: 1.07-1.15, P = 7.98*10 - 8). With the use of large-scale GWAS data, robust evidence suggests an interaction between risk preference and schizophrenia. This also indicates that early identification of and intervention for increased risk preference may improve the prognosis of schizophrenia.

Genetically predicted major depression causally increases the risk of temporomandibular joint disorders.

Observational studies have reported that mental disorders are comorbid with temporomandibular joint disorder (TMD). However, the causal relationship remains uncertain. To clarify the causal relationship between three common mental illnesses and TMD, we conduct this Mendelian Randomization (MR) study. The large-scale genome-wide association studies data of major depression, bipolar disorder and schizophrenia were retrieved from the Psychiatric Genomics Consortium. The summary data of TMD was obtained from the Finn-Gen consortium, including 211,023 subjects of European descent (5,668 cases and 205,355 controls). The main approach utilized was inverse variance weighting (IVW) to evaluate the causal association between the three mental disorders and TMD. Five sensitivity analyses including MR-Egger, Maximum Likelihood, Weighted median, MR. RAPS and MR-PRESSO were used as supplements. We conducted heterogeneity tests and pleiotropic tests to ensure the robustness. As shown by the IVW method, genetically determined major depression was associated with a 1.65-fold risk of TMD (95% CI = 1.10-2.47, p < 0.05). The direction and effect size remained consistent with sensitivity analyses. The odds ratios (ORs) were 1.51 (95% CI = 0.24-9.41, p > 0.05) for MR-Egger, 1.60 (95% CI = 0.98-2.61, p > 0.05) for Weighted median, 1.68 (95% CI = 1.19-2.38, p < 0.05) for Maximum likelihood, 1.56 (95% CI = 1.05-2.33, p < 0.05) for MR. RAPS, and 1.65 (95% CI = 1.10-2.47, p < 0.05) for MR-PRESSO, respectively. No pleiotropy was observed (both P for MR-Egger intercept and Global test >0.05). In addition, the IVW method identified no significant correlation between bipolar disorder, schizophrenia and TMD. Genetic evidence supports a causal relationship between major depression and TMD, instead of bipolar disorder and schizophrenia. These findings emphasize the importance of assessing a patient's depressive status in clinical settings.

A Mendelian randomization study of genetic liability to cutaneous melanoma and sunburns.

Some studies have reported that sunburns and cutaneous melanoma (CM) risk is increasing, but a clear causal link has yet to be established. This current study conducted a two-sample Mendelian randomization (MR) approach to clarify the association and causality between sunburn history and CM using large-scale genome-wide association study data. The inverse-variance weighted method result showed that sunburn might be associated with the risk of CM increasing (p = 2.21 × 10-23, OR = 1.034, 95% CI= 1.027-1.041), causally. The MR-Egger regression, weighted median method, simple mode method, and weighted mode method results showed similar results. This study offers evidence of sunburn history and increased risk of CM, and it shows that there might be common genetic basics regarding sunburns and CM susceptibility in Caucasian, European, or British ethnic groups.

Dissecting Complex Traits Using Omics Data: A Review on the Linear Mixed Models and Their Application in GWAS.

Genome-wide association study (GWAS) is the most popular approach to dissecting complex traits in plants, humans, and animals. Numerous methods and tools have been proposed to discover the causal variants for GWAS data analysis. Among them, linear mixed models (LMMs) are widely used statistical methods for regulating confounding factors, including population structure, resulting in increased computational proficiency and statistical power in GWAS studies. Recently more attention has been paid to pleiotropy, multi-trait, gene-gene interaction, gene-environment interaction, and multi-locus methods with the growing availability of large-scale GWAS data and relevant phenotype samples. In this review, we have demonstrated all possible LMMs-based methods available in the literature for GWAS. We briefly discuss the different LMM methods, software packages, and available open-source applications in GWAS. Then, we include the advantages and weaknesses of the LMMs in GWAS. Finally, we discuss the future perspective and conclusion. The present review paper would be helpful to the researchers for selecting appropriate LMM models and methods quickly for GWAS data analysis and would benefit the scientific society.

A Constrained Generalized Functional Linear Model for Multi-Loci Genetic Mapping

In genome-wide association studies (GWAS), efficient incorporation of linkage disequilibria (LD) among densely typed genetic variants into association analysis is a critical yet challenging problem. Functional linear models (FLM), which impose a smoothing structure on the coefficients of correlated covariates, are advantageous in genetic mapping of multiple variants with high LD. Here we propose a novel constrained generalized FLM (cGFLM) framework to perform simultaneous association tests on a block of linked SNPs with various trait types, including continuous, binary and zero-inflated count phenotypes. The new cGFLM applies a set of inequality constraints on the FLM to ensure model identifiability under different genetic codings. The method is implemented via B-splines, and an augmented Lagrangian algorithm is employed for parameter estimation. For hypotheses testing, a test statistic that accounts for the model constraints was derived, following a mixture of chi-square distributions. Simulation results show that cGFLM is effective in identifying causal loci and gene clusters compared to several competing methods based on single markers and SKAT-C. We applied the proposed method to analyze a candidate gene-based COGEND study and a large-scale GWAS data on dental caries risk.

Integration of summary data from GWAS and eQTL studies identified novel risk genes for coronary artery disease.

Several genetic loci have been reported to be significantly associated with coronary artery disease (CAD) by multiple genome-wide association studies (GWAS). Nevertheless, the biological and functional effects of these genetic variants on CAD remain largely equivocal. In the current study, we performed an integrative genomics analysis by integrating large-scale GWAS data (N = 459,534) and 2 independent expression quantitative trait loci (eQTL) datasets (N = 1890) to determine whether CAD-associated risk single nucleotide polymorphisms (SNPs) exert regulatory effects on gene expression. By using Sherlock Bayesian, MAGMA gene-based, multidimensional scaling (MDS), functional enrichment, and in silico permutation analyses for independent technical and biological replications, we highlighted 4 susceptible genes (CHCHD1, TUBG1, LY6G6C, and MRPS17) associated with CAD risk. Based on the protein–protein interaction (PPI) network analysis, these 4 genes were found to interact with each other. We detected a remarkably altered co-expression pattern among these 4 genes between CAD patients and controls. In addition, 3 genes of CHCHD1 (P = .0013), TUBG1 (P = .004), and LY6G6C (P = .038) showed significantly different expressions between CAD patients and controls. Together, we provide evidence to support that these identified genes such as CHCHD1 and TUBG1 are indicative factors of CAD.

Genome-wide association study reveals new insights into the heritability and genetic correlates of developmental dyslexia

Developmental dyslexia (DD) is a learning disorder affecting the ability to read, with a heritability of 40–60%. A notable part of this heritability remains unexplained, and large genetic studies are warranted to identify new susceptibility genes and clarify the genetic bases of dyslexia. We carried out a genome-wide association study (GWAS) on 2274 dyslexia cases and 6272 controls, testing associations at the single variant, gene, and pathway level, and estimating heritability using single-nucleotide polymorphism (SNP) data. We also calculated polygenic scores (PGSs) based on large-scale GWAS data for different neuropsychiatric disorders and cortical brain measures, educational attainment, and fluid intelligence, testing them for association with dyslexia status in our sample. We observed statistically significant (p < 2.8 × 10−6) enrichment of associations at the gene level, for LOC388780 (20p13; uncharacterized gene), and for VEPH1 (3q25), a gene implicated in brain development. We estimated an SNP-based heritability of 20–25% for DD, and observed significant associations of dyslexia risk with PGSs for attention deficit hyperactivity disorder (at pT = 0.05 in the training GWAS: OR = 1.23[1.16; 1.30] per standard deviation increase; p = 8 × 10−13), bipolar disorder (1.53[1.44; 1.63]; p = 1 × 10−43), schizophrenia (1.36[1.28; 1.45]; p = 4 × 10−22), psychiatric cross-disorder susceptibility (1.23[1.16; 1.30]; p = 3 × 10−12), cortical thickness of the transverse temporal gyrus (0.90[0.86; 0.96]; p = 5 × 10−4), educational attainment (0.86[0.82; 0.91]; p = 2 × 10−7), and intelligence (0.72[0.68; 0.76]; p = 9 × 10−29). This study suggests an important contribution of common genetic variants to dyslexia risk, and novel genomic overlaps with psychiatric conditions like bipolar disorder, schizophrenia, and cross-disorder susceptibility. Moreover, it revealed the presence of shared genetic foundations with a neural correlate previously implicated in dyslexia by neuroimaging evidence.

Integrative genomics analysis of various omics data and networks identify risk genes and variants vulnerable to childhood-onset asthma

BackgroundChildhood-onset asthma is highly affected by genetic components. In recent years, many genome-wide association studies (GWAS) have reported a large group of genetic variants and susceptible genes associated with asthma-related phenotypes including childhood-onset asthma. However, the regulatory mechanisms of these genetic variants for childhood-onset asthma susceptibility remain largely unknown.MethodsIn the current investigation, we conducted a two-stage designed Sherlock-based integrative genomics analysis to explore the cis- and/or trans-regulatory effects of genome-wide SNPs on gene expression as well as childhood-onset asthma risk through incorporating a large-scale GWAS data (N = 314,633) and two independent expression quantitative trait loci (eQTL) datasets (N = 1890). Furthermore, we applied various bioinformatics analyses, including MAGMA gene-based analysis, pathway enrichment analysis, drug/disease-based enrichment analysis, computer-based permutation analysis, PPI network analysis, gene co-expression analysis and differential gene expression analysis, to prioritize susceptible genes associated with childhood-onset asthma.ResultsBased on comprehensive genomics analyses, we found 31 genes with multiple eSNPs to be convincing candidates for childhood-onset asthma risk; such as, PSMB9 (cis-rs4148882 and cis-rs2071534) and TAP2 (cis-rs9267798, cis-rs4148882, cis-rs241456, and trans-10,447,456). These 31 genes were functionally interacted with each other in our PPI network analysis. Our pathway enrichment analysis showed that numerous KEGG pathways including antigen processing and presentation, type I diabetes mellitus, and asthma were significantly enriched to involve in childhood-onset asthma risk. The co-expression patterns among 31 genes were remarkably altered according to asthma status, and 25 of 31 genes (25/31 = 80.65%) showed significantly or suggestively differential expression between asthma group and control group.ConclusionsWe provide strong evidence to highlight 31 candidate genes for childhood-onset asthma risk, and offer a new insight into the genetic pathogenesis of childhood-onset asthma.

Abstract 29: Integrating computational epigenetic and statistical approaches to investigate how genome-wide transcription factor (TF)-DNA bindings affect breast cancer risk

Abstract Background: Fine-mapping and functional genomic studies of risk loci identified by genome wide association studies (GWAS) of breast cancer suggest that risk-associated regulatory variants may disrupt DNA binding affinities of transcription factors (TFs), such as FOXA1 and ESR1, thus altering gene expression and affecting breast cancer risk. To date, however, no studies have directly investigated how genome-wide TF-DNA bindings in a tissue-specific context affect breast cancer risk. Methods: We recently developed an analytic framework using computational epigenetic approaches and generalized mixed models to identify breast cancer risk associated cis-regulatory elements that are occupied by TFs. Summary statistics derived from GWAS of imputed data for approximately 11 million genetic variants were obtained from the Breast Cancer Association Consortium (BCAC). A total of 113 ChIP-seq datasets for TFs and chromatin features were annotated from the Encyclopedia of DNA Elements (ENCODE) and Roadmap projects and the Cistrome database (http://cistrome.org/) in multiple breast cancer cell lines..Chi-square values reported in the BCAC dataset were used to measure breast cancer risk associated with a genetic variant. To investigate TF-DNA bindings for breast cancer risk, we constructed generalized mixed models to evaluate the associations between the Chi-square values and binding sites of single TF or co-occupied by multiple TFs, given linkage disequilibrium (LD) blocks of variants to handle their dependence. To define approximately independent LD blocks similar to other studies, we defined LD blocks using non-overlapping segments of 100k bps. In addition, we investigated the effect of putative cis-regulatory elements characterized by TF-DNA bindings and chromatin features on breast cancer risk. Results: We identified a total of 22 TFs where their TF-DNA bindings were significantly associated with breast cancer risk at P &amp;lt; 1 × 10−5, with the top five TFs being FOXA1, SIN3AK, ESR1, TCF7l2, and AR (P &amp;lt; 1 × 10−10). We further observed stronger associations of breast cancer risk with co-occupied TF-TF DNA bindings, with the top five TF pairs (P &amp;lt; 1 × 10−16) being FOXA1+E2F1, FOXA1+NR2F2, FOXA1+ESR1, FOXA1+SIN3, and FOXA1+TCF12. The interactions of TF-TF DNA bindings for these five TF pairs were highly significant (P &amp;lt; 1 × 10−5). In addition, our results showed a significant interaction (P &amp;lt; 1 × 10−5) between chromatin features and the TF-DNA bindings. As compared with quiescent/low chromatin features, more TF-DNA bindings in enhancers, strong or weak transcription, or heterochromatin were associated with lower breast cancer risk. Conclusions: Our findings suggest that putative regulatory variants, which alter the TF-DNA binding affinities, particularly those located in TF-TF co-occupied sites, and TF-DNA binding colocalized with chromatin features, play an important role in breast cancer risk. Our approaches can be generalized to investigate the effect of genome-wide TF-DNA bindings on the risk of any human cancers that have comprehensive ChIP-seq and large-scale GWAS data. Citation Format: Wanqing Wen, Zhishan Chen, Quan Long, Wei Zheng, Xingyi Guo. Integrating computational epigenetic and statistical approaches to investigate how genome-wide transcription factor (TF)-DNA bindings affect breast cancer risk [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 29.

Genome-Wide Natural Selection Signatures Are Linked to Genetic Risk of Modern Phenotypes in the Japanese Population.

Elucidation of natural selection signatures and relationships with phenotype spectra is important to understand adaptive evolution of modern humans. Here, we conducted a genome-wide scan of selection signatures of the Japanese population by estimating locus-specific time to the most recent common ancestor using the ascertained sequentially Markovian coalescent (ASMC), from the biobank-based large-scale genome-wide association study data of 170,882 subjects. We identified 29 genetic loci with selection signatures satisfying the genome-wide significance. The signatures were most evident at the alcohol dehydrogenase (ADH) gene cluster locus at 4q23 (PASMC = 2.2 × 10−36), followed by relatively strong selection at the FAM96A (15q22), MYOF (10q23), 13q21, GRIA2 (4q32), and ASAP2 (2p25) loci (PASMC < 1.0 × 10−10). The additional analysis interrogating extended haplotypes (integrated haplotype score) showed robust concordance of the detected signatures, contributing to fine-mapping of the genes, and provided allelic directional insights into selection pressure (e.g., positive selection for ADH1B-Arg48His and HLA-DPB1*04:01). The phenome-wide selection enrichment analysis with the trait-associated variants identified a variety of the modern human phenotypes involved in the adaptation of Japanese. We observed population-specific evidence of enrichment with the alcohol-related phenotypes, anthropometric and biochemical clinical measurements, and immune-related diseases, differently from the findings in Europeans using the UK Biobank resource. Our study demonstrated population-specific features of the selection signatures in Japanese, highlighting a value of the natural selection study using the nation-wide biobank-scale genome and phenotype data.

Abstract LB-176: Transcriptome-wide association study revealed a novel gene associated with persistent hepatitis B virus infection among Chinese

Abstract The etiology of persistent hepatitis B virus (HBV) infection is complex and multifactorial, polygenic event with viral, environmental and genetic components being considered as the risk factors. Genome-wide association study (GWAS) has been applied into the exploration of genetic etiology of persistent hepatitis B virus infection among Chinese. However, the mechanisms beyond the genetic variants identified by GWAS located in noncoding regions are largely unknown. To further explore the causal gene of persistent hepatitis B virus (HBV) infection, we used a new approach developed by Gusev and Ko, leveraging expression imputation from large-scale GWAS data by using a small set of individuals with both genotype and gene expression data as a reference panel, to perform a transcriptome-wide association study (TWAS).Based on Genotype-Tissue Expression (GTEx) database, we included 97 individuals with both gene expression data in liver and SNP array data in blood as a reference panel. We first estimated the cis SNP heritability (cis-hg2) for each gene in 97 liver tissues using genome-wide complex trait analysis (GCTA) software, and kept cis-heritable genes for subsequent analysis. Then, we estimated the effect sizes of cis-SNPs on gene expression using genome-wide efficient mixed-model association (GEMMA) algorithm. Using the effect sizes trained from the reference panel, we imputed the gene expression levels for our existing NJMU HEP GWAS cohort (951 persistent hepatitis B virus (HBV) infection patients and 937 controls) and correlated the imputed gene expression values with persistent hepatitis B virus (HBV) infection trait. The cis-hg2 estimate was significantly nonzero for 2588 genes (P&amp;lt;0.05). The average cis-hg2value was 0.438. Using GTEx database as a reference panel. In the training set, After FDR correction, we identified three potentially persistent hepatitis B virus (HBV) infection -associated genes, those genes were located in 6p21.3(HLA-DPA1,PTWAS=2.02×10-7), 3q21.1(PARP9, PTWAS=2.2×10-4) and 22q11.21 (TMEM191A, PTWAS= 3.8×10-4)respectively. 6p21.3 and 22q11.21 as our previous reported. 3q21.1 was a novel region. Citation Format: Jing Han, Jing Ni, Meng Zhu, Jin Guang Fu, Zhi Bin Hu. Transcriptome-wide association study revealed a novel gene associated with persistent hepatitis B virus infection among Chinese [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr LB-176.

305-OR: Genomic Basis to Shared Influences on Pubertal Timing and Type 2 Diabetes

Earlier pubertal onset in girls is associated with later-life type 2 diabetes (T2D) risk. Studies have shown that age at menarche (AAM), a marker of girls’ pubertal timing, and T2D are genetically correlated genome-wide. Since both AAM and T2D also correlate with BMI, the biological link between the two traits could be mediated via BMI. As the most recent genome-wide association study (GWAS) for T2D also adjusted for BMI, we queried if AAM and T2D remain correlated after removing the effect of BMI on T2D. Leveraging such large-scale GWAS data, LD Score Regression revealed that the genetic correlation between AAM and T2D (rg (SE)=-0.24 (0.02) P=2.2x10-22) was only partly attenuated after BMI adjustment (rg (SE)=-0.1 (0.03) P=0.0002), suggesting that BMI does not entirely explain the puberty-T2D link. Five of the 13 significant GWAS-implicated loci associated with both AAM and T2D do not associate with BMI, further supporting that BMI is not the only mediator between the two traits. Given that causal effector genes are unknown at most GWAS loci, identification of target genes at these shared loci should provide key biological insights. To implicate effector genes, we first identified all SNPs in LD with the sentinel SNPs. We then extracted the subsets of proxy SNPs in open chromatin, determined by ATAC-seq, in primary cells or cell lines relevant to this trait area [beta cells (EndoC-βH1), adipose (SGBS and MSC-derived adipocytes), and—given recent findings in both AAM and BMI studies—brain (neural precursors, microglia, astrocytes and hypothalamic neurons)]. Using high resolution Capture-C, we detected consistent physical contacts between open-proxy SNPs and candidate effector genes. Our results support direct SNP-to-gene promoter contacts for at least four of these shared loci (one not associated with BMI) including TSPAN3 at ‘AC046168.1’ and TFAP2D at ‘TFAP2B’, along with leads at the ‘MTCH2/CELF1’ and ‘MAP2K5’ loci. Follow-up experiments are needed to determine the physiological roles of implicated target genes. Disclosure D.L. Cousminer: None. R. Mishra: None. M.A. Argenziano: None. E. Manduchi: None. K.M. Hodge: None. C. Su: None. M. Leonard: None. S. Lu: None. J.A. Pippin: None. M. Johnson: None. A.D. Wells: Research Support; Self; GlaxoSmithKline plc. A. Chesi: None. B.F. Voight: None. S.F. Grant: None. Funding American Diabetes Association (1-17-PDF-077 to D.L.C.); National Institutes of Health

Diabetes and cancer risk: A Mendelian randomization study.

Earlier cohort studies using conventional regression models have consistently shown an increased cancer risk among individuals with type 2 diabetes. However, reverse causality and residual confounding due to common risk factors could exist, and it remains unclear whether diabetes per se contributes to cancer development. Mendelian randomization analyses might clarify the true association between diabetes and cancer risk. We conducted a case–cohort study with 10,536 subcohort subjects and 3,541 newly diagnosed cancer cases derived from 32,949 eligible participants aged 40–69 years within the Japan Public Health Center‐based Prospective Study. With 29 known type 2 diabetes susceptibility variants, we used an inverse variance‐weighted method to estimate hazard ratios for the associations of diabetes with risks of total and site‐specific cancers. The hazard ratios of cancer per doubling of the probability of diabetes were 1.03 (95% confidence interval [CI], 0.92–1.15) overall, 1.08 (95% CI: 0.73–1.59) for the pancreas, 0.80 (95% CI: 0.57–1.14) for the liver and 0.90 (95% CI: 0.74–1.10) for the colorectum. Additional analyses, using publicly available large‐scale genome‐wide association study data on colorectal cancer in Japan, resulted in a narrower CI (hazard ratio: 1.00; 95% CI: 0.93–1.07). In this prospective Mendelian randomization study with a large number of incident cancer cases, we found no strong evidence to support associations between diabetes and overall and site‐specific cancer risks. Our findings suggest that there is little evidence to support the genetic role of type 2 diabetes in cancer development in the Japanese population.

LPG: A four-group probabilistic approach to leveraging pleiotropy in genome-wide association studies

BackgroundTo date, genome-wide association studies (GWAS) have successfully identified tens of thousands of genetic variants among a variety of traits/diseases, shedding light on the genetic architecture of complex disease. The polygenicity of complex diseases is a widely accepted phenomenon through which a vast number of risk variants, each with a modest individual effect, collectively contribute to the heritability of complex diseases. This imposes a major challenge on fully characterizing the genetic bases of complex diseases. An immediate implication of polygenicity is that a much larger sample size is required to detect individual risk variants with weak/moderate effects. Meanwhile, accumulating evidence suggests that different complex diseases can share genetic risk variants, a phenomenon known as pleiotropy.ResultsIn this study, we propose a statistical framework for Leveraging Pleiotropic effects in large-scale GWAS data (LPG). LPG utilizes a variational Bayesian expectation-maximization (VBEM) algorithm, making it computationally efficient and scalable for genome-wide-scale analysis. To demonstrate the advantages of LPG over existing methods that do not leverage pleiotropy, we conducted extensive simulation studies and applied LPG to analyze two pairs of disorders (Crohn’s disease and Type 1 diabetes, as well as rheumatoid arthritis and Type 1 diabetes). The results indicate that by levelaging pleiotropy, LPG can improve the power of prioritization of risk variants and the accuracy of risk prediction.ConclusionsOur methodology provides a novel and efficient tool to detect pleiotropy among GWAS data for multiple traits/diseases collected from different studies. The software is available at https://github.com/Shufeyangyi2015310117/LPG.

Re-assessment of multiple testing strategies for more efficient genome-wide association studies.

Although enormous costs have been dedicated to discovering relevant disease-related genetic variants, especially in genome-wide association studies (GWASs), only a small fraction of estimated heritability can be explained by these results. This is the so-called missing heritability problem. The conventional use of overly conservative multiple testing strategies based on controlling the familywise error rate (FWER), in particular with a genome-wide significance threshold of P <5 × 10-8, is one of the most important issues from a statistical perspective. To help resolve this problem, we performed comprehensive re-assessments of currently available strategies using recently published, extremely large-scale GWAS data sets of rheumatoid arthritis and schizophrenia (>50,000 subjects). The estimates of statistical power averaged for all disease-related genetic variants of the standard FWER-based strategy were only 0.09% for the rheumatoid arthritis data and 0.04% for the schizophrenia data. To design more efficient strategies, we also conducted an extensive comparison of multiple testing strategies by applying false discovery rate (FDR)-controlling procedures to these data sets and simulations, and found that the FDR-based procedures achieved higher power than the FWER-based strategy, even at a strict FDR level (e.g., FDR = 1%). We also discuss a useful alternative measure, namely "partial power," which is an averaged power for detecting the clinically and biologically meaningful genetic factors with the largest effects. Simulation results suggest that the FDR-based procedures can achieve sufficient partial power (>80%) for detecting these factors (odds ratios of >1.05) with 80,000 subjects, and thus this may be a useful measure for defining realistic objectives of future GWASs.

High-performance epistasis detection in quantitative trait GWAS

epiSNP is a program for identifying pairwise single nucleotide polymorphism (SNP) interactions (epistasis) in quantitative-trait genome-wide association studies (GWAS). A parallel MPI version (EPISNPmpi) was created in 2008 to address this computationally expensive analysis on large data sets with many quantitative traits and SNP markers. However, the falling cost of genotyping has led to an explosion of large-scale GWAS data sets that challenge EPISNPmpi’s ability to compute results in a reasonable amount of time. Therefore, we optimized epiSNP for modern multi-core and highly parallel many-core processors to efficiently handle these large data sets. This paper describes the serial optimizations, dynamic load balancing using MPI-3 RMA operations, and shared-memory parallelization with OpenMP to further enhance load balancing and allow execution on the Intel Xeon Phi coprocessor (MIC). For a large GWAS data set, our optimizations provided a 38.43× speedup over EPISNPmpi on 126 nodes using 2 MICs on TACC’s Stampede Supercomputer. We also describe a Coarray Fortran (CAF) version that demonstrates the suitability of PGAS languages for problems with this computational pattern. We show that the Coarray version performs competitively with the MPI version on the NERSC Edison Cray XC30 supercomputer. Finally, the performance benefits of hyper-threading for this application on Edison (average 1.35× speedup) are demonstrated.

Summarizing polygenic risks for complex diseases in a clinical whole-genome report.

Disease-causing mutations and pharmacogenomic variants are of primary interest for clinical whole-genome sequencing. However, estimating genetic liability for common complex diseases using established risk alleles might one day prove clinically useful. We compared polygenic scoring methods using a case-control data set with independently discovered risk alleles in the MedSeq Project. For eight traits of clinical relevance in both the primary-care and cardiomyopathy study cohorts, we estimated multiplicative polygenic risk scores using 161 published risk alleles and then normalized them using the population median estimated from the 1000 Genomes Project. Our polygenic score approach identified the overrepresentation of independently discovered risk alleles in cases as compared with controls using a large-scale genome-wide association study data set. In addition to normalized multiplicative polygenic risk scores and rank in a population, the disease prevalence and proportion of heritability explained by known common risk variants provide important context in the interpretation of modern multilocus disease risk models. Our approach in the MedSeq Project demonstrates how complex trait risk variants from an individual genome can be summarized and reported for the general clinician and also highlights the need for definitive clinical studies to obtain reference data for such estimates and to establish clinical utility.

Chronic Periodontitis Genome-wide Association Studies

Recent genome-wide association studies (GWAS) of chronic periodontitis (CP) offer rich data sources for the investigation of candidate genes, functional elements, and pathways. We used GWAS data of CP (n = 4,504) and periodontal pathogen colonization (n = 1,020) from a cohort of adult Americans of European descent participating in the Atherosclerosis Risk in Communities study and employed a MAGENTA approach (i.e., meta-analysis gene set enrichment of variant associations) to obtain gene-centric and gene set association results corrected for gene size, number of single-nucleotide polymorphisms, and local linkage disequilibrium characteristics based on the human genome build 18 (National Center for Biotechnology Information build 36). We used the Gene Ontology, Ingenuity, KEGG, Panther, Reactome, and Biocarta databases for gene set enrichment analyses. Six genes showed evidence of statistically significant association: 4 with severe CP (NIN, p = 1.6 × 10−7; ABHD12B, p = 3.6 × 10−7; WHAMM, p = 1.7 × 10−6; AP3B2, p = 2.2 × 10−6) and 2 with high periodontal pathogen colonization (red complex–KCNK1, p = 3.4 × 10−7; Porphyromonas gingivalis–DAB2IP, p = 1.0 × 10−6). Top-ranked genes for moderate CP were HGD (p = 1.4 × 10−5), ZNF675 (p = 1.5 × 10−5), TNFRSF10C (p = 2.0 × 10−5), and EMR1 (p = 2.0 × 10−5). Loci containing NIN, EMR1, KCNK1, and DAB2IP had showed suggestive evidence of association in the earlier single-nucleotide polymorphism–based analyses, whereas WHAMM and AP2B2 emerged as novel candidates. The top gene sets included severe CP (“endoplasmic reticulum membrane,” “cytochrome P450,” “microsome,” and “oxidation reduction”) and moderate CP (“regulation of gene expression,” “zinc ion binding,” “BMP signaling pathway,” and “ruffle”). Gene-centric analyses offer a promising avenue for efficient interrogation of large-scale GWAS data. These results highlight genes in previously identified loci and new candidate genes and pathways possibly associated with CP, which will need to be validated via replication and mechanistic studies.

CARAT-GxG: CUDA-Accelerated Regression Analysis Toolkit for Large-Scale Gene-Gene Interaction with GPU Computing System.

In genome-wide association studies (GWAS), regression analysis has been most commonly used to establish an association between a phenotype and genetic variants, such as single nucleotide polymorphism (SNP). However, most applications of regression analysis have been restricted to the investigation of single marker because of the large computational burden. Thus, there have been limited applications of regression analysis to multiple SNPs, including gene–gene interaction (GGI) in large-scale GWAS data. In order to overcome this limitation, we propose CARAT-GxG, a GPU computing system-oriented toolkit, for performing regression analysis with GGI using CUDA (compute unified device architecture). Compared to other methods, CARAT-GxG achieved almost 700-fold execution speed and delivered highly reliable results through our GPU-specific optimization techniques. In addition, it was possible to achieve almost-linear speed acceleration with the application of a GPU computing system, which is implemented by the TORQUE Resource Manager. We expect that CARAT-GxG will enable large-scale regression analysis with GGI for GWAS data.