A comparison of multivariate genome-wide association methods

Maize green leaf area index dynamics: genetic basis of a new secondary trait for grain yield in optimal and drought conditions

KEY MESSAGE

Green Leaf Area Index dynamics is a promising secondary trait for grain yield and drought tolerance. Multivariate GWAS is particularly well suited to identify the genetic determinants of the green leaf area index dynamics. Improvement of maize grain yield is impeded by important genotype-environment interactions, especially under drought conditions. The use of secondary traits, that are correlated with yield, more heritable and less prone to genotype-environment interactions, can increase breeding efficiency. Here, we studied the genetic basis of a new secondary trait: the green leaf area index (GLAI) dynamics over the maize life cycle. For this, we used an unmanned aerial vehicle to characterize the GLAI dynamics of a diverse panel in well-watered and water-deficient trials in two years. From the dynamics, we derived 24 traits (slopes, durations, areas under the curve), and showed that six of them were heritable traits representative of the panel diversity. To identify the genetic determinants of GLAI, we compared two genome-wide association approaches: a univariate (single-trait) method and a multivariate (multi-trait) method combining GLAI traits, grain yield, and precocity. The explicit modeling of correlation structure between secondary traits and grain yield in the multivariate mixed model led to 2.5 times more associations detected. A total of 475 quantitative trait loci (QTLs) were detected. The genetic architecture of GLAI traits appears less complex than that of yield with stronger-effect QTLs that are more stable between environments. We also showed that a subset of GLAI QTLs explains nearly one fifth of yield variability across a larger environmental network of 11 water-deficient trials. GLAI dynamics is a promising grain yield secondary trait in optimal and drought conditions, and the detected QTLs could help to increase breeding efficiency through a marker-assisted approach.

Joint regression analysis of multiple traits based on genetic relationships

Motivation

Polygenic scores (PGSs) are widely available and employed in genomic data analyses for predicting and understanding genetic architectures. Existing approaches either require information on SNP level, do not infer clusters of traits sharing genetic characteristic, or do not have any immediate predictive properties.

Results

Here, we present geneJAM, which is a novel clustering and estimation method using PGSs for inferring a genetic relationship among multiple, simultaneously measured and potentially correlated traits in a multivariate GWAS.Using graphical lasso, we estimate a sparse covariance matrix of the PGSs and obtain clusters of traits sharing genetic characteristics. We use the clusters to specify the structure of the error covariance matrix of a generalized least squares (GLS) model and use the feasible GLS estimator for estimating a linear regression model with a certain unknown degree of correlation between the residuals.The method suits many biology studies well with traits embedded in some genetic functioning groups and facilitates development of the PGS research. We compare the method with fully parametric techniques on simulated data and illustrate the utility of the methods by examining a heterogeneous stock mouse data set from the Wellcome Trust Centre for Human Genetics. We demonstrate that the method successfully identifies clusters of traits and increases precision, power, and computational efficiency.

Availability and implementation

GeneJAM is implemented in R and available at: https://github.com/abuchardt/geneJAM.

Joint analysis of multiple phenotypes for extremely unbalanced case-control association studies using multi-layer network

MOTIVATION

Genome-wide association studies is an essential tool for analyzing associations between phenotypes and single nucleotide polymorphisms (SNPs). Most of binary phenotypes in large biobanks are extremely unbalanced, which leads to inflated type I error rates for many widely used association tests for joint analysis of multiple phenotypes. In this article, we first propose a novel method to construct a Multi-Layer Network (MLN) using individuals with at least one case status among all phenotypes. Then, we introduce a computationally efficient community detection method to group phenotypes into disjoint clusters based on the MLN. Finally, we propose a novel approach, MLN with Omnibus (MLN-O), to jointly analyse the association between phenotypes and a SNP. MLN-O uses the score test to test the association of each merged phenotype in a cluster and a SNP, then uses the Omnibus test to obtain an overall test statistic to test the association between all phenotypes and a SNP.

RESULTS

We conduct extensive simulation studies to reveal that the proposed approach can control type I error rates and is more powerful than some existing methods. Meanwhile, we apply the proposed method to a real data set in the UK Biobank. Using phenotypes in Chapter XIII (Diseases of the musculoskeletal system and connective tissue) in the UK Biobank, we find that MLN-O identifies more significant SNPs than other methods we compare with.

AVAILABILITY AND IMPLEMENTATION

https://github.com/Hongjing-Xie/Multi-Layer-Network-with-Omnibus-MLN-O.

A fast non-parametric test of association for multiple traits

The increasing availability of multidimensional phenotypic data in large cohorts of genotyped individuals requires efficient methods to identify genetic effects on multiple traits. Permutational multivariate analysis of variance (PERMANOVA) offers a powerful non-parametric approach. However, it relies on permutations to assess significance, which hinders the analysis of large datasets. Here, we derive the limiting null distribution of the PERMANOVA test statistic, providing a framework for the fast computation of asymptotic p values. Our asymptotic test presents controlled type I error and high power, often outperforming parametric approaches. We illustrate its applicability in the context of QTL mapping and GWAS.

Multivariate GWAS of Alzheimer's disease CSF biomarker profiles implies GRIN2D in synaptic functioning

BACKGROUND

Genome-wide association studies (GWAS) of Alzheimer's disease (AD) have identified several risk loci, but many remain unknown. Cerebrospinal fluid (CSF) biomarkers may aid in gene discovery and we previously demonstrated that six CSF biomarkers (β-amyloid, total/phosphorylated tau, NfL, YKL-40, and neurogranin) cluster into five principal components (PC), each representing statistically independent biological processes. Here, we aimed to (1) identify common genetic variants associated with these CSF profiles, (2) assess the role of associated variants in AD pathophysiology, and (3) explore potential sex differences.

METHODS

We performed GWAS for each of the five biomarker PCs in two multi-center studies (EMIF-AD and ADNI). In total, 973 participants (n = 205 controls, n = 546 mild cognitive impairment, n = 222 AD) were analyzed for 7,433,949 common SNPs and 19,511 protein-coding genes. Structural equation models tested whether biomarker PCs mediate genetic risk effects on AD, and stratified and interaction models probed for sex-specific effects.

RESULTS

Five loci showed genome-wide significant association with CSF profiles, two were novel (rs145791381 [inflammation] and GRIN2D [synaptic functioning]) and three were previously described (APOE, TMEM106B, and CHI3L1). Follow-up analyses of the two novel signals in independent datasets only supported the GRIN2D locus, which contains several functionally interesting candidate genes. Mediation tests indicated that variants in APOE are associated with AD status via processes related to amyloid and tau pathology, while markers in TMEM106B and CHI3L1 are associated with AD only via neuronal injury/inflammation. Additionally, seven loci showed sex-specific associations with AD biomarkers.

CONCLUSIONS

These results suggest that pathway and sex-specific analyses can improve our understanding of AD genetics and may contribute to precision medicine.

Multivariate genome-wide associations for immune traits in two maternal pig lines

BACKGROUND

Immune traits are considered to serve as potential biomarkers for pig's health. Medium to high heritabilities have been observed for some of the immune traits suggesting genetic variability of these phenotypes. Consideration of previously established genetic correlations between immune traits can be used to identify pleiotropic genetic markers. Therefore, genome-wide association study (GWAS) approaches are required to explore the joint genetic foundation for health biomarkers. Usually, GWAS explores phenotypes in a univariate (uv), trait-by-trait manner. Besides two uv GWAS methods, four multivariate (mv) GWAS approaches were applied on combinations out of 22 immune traits for Landrace (LR) and Large White (LW) pig lines.

RESULTS

In total 433 (LR: 351, LW: 82) associations were identified with the uv approach implemented in PLINK and a Bayesian linear regression uv approach (BIMBAM) software. Single Nucleotide Polymorphisms (SNPs) that were identified with both uv approaches (n = 32) were mostly associated with immune traits such as haptoglobin, red blood cell characteristics and cytokines, and were located in protein-coding genes. Mv GWAS approaches detected 647 associations for different mv immune trait combinations which were summarized to 133 Quantitative Trait Loci (QTL). SNPs for different trait combinations (n = 66) were detected with more than one mv method. Most of these SNPs are associated with red blood cell related immune trait combinations. Functional annotation of these QTL revealed 453 immune-relevant protein-coding genes. With uv methods shared markers were not observed between the breeds, whereas mv approaches were able to detect two conjoint SNPs for LR and LW. Due to unmapped positions for these markers, their functional annotation was not clarified.

CONCLUSIONS

This study evaluated the joint genetic background of immune traits in LR and LW piglets through the application of various uv and mv GWAS approaches. In comparison to uv methods, mv methodologies identified more significant associations, which might reflect the pleiotropic background of the immune system more accurately. In genetic research of complex traits, the SNP effects are generally small. Furthermore, one genetic variant can affect several correlated immune traits at the same time, termed pleiotropy. As mv GWAS methods consider strong dependencies among traits, the power to detect SNPs can be boosted. Both methods revealed immune-relevant potential candidate genes. Our results indicate that one single test is not able to detect all the different types of genetic effects in the most powerful manner and therefore, the methods should be applied complementary.

Simultaneous modeling of multivariate heterogeneous responses and heteroskedasticity via a two-stage composite likelihood

Multivariate heterogeneous responses and heteroskedasticity have attracted increasing attention in recent years. In genome-wide association studies, effective simultaneous modeling of multiple phenotypes would improve statistical power and interpretability. However, a flexible common modeling system for heterogeneous data types can pose computational difficulties. Here we build upon a previous method for multivariate probit estimation using a two-stage composite likelihood that exhibits favorable computational time while retaining attractive parameter estimation properties. We extend this approach to incorporate multivariate responses of heterogeneous data types (binary and continuous), and possible heteroskedasticity. Although the approach has wide applications, it would be particularly useful for genomics, precision medicine, or individual biomedical prediction. Using a genomics example, we explore statistical power and confirm that the approach performs well for hypothesis testing and coverage percentages under a wide variety of settings. The approach has the potential to better leverage genomics data and provide interpretable inference for pleiotropy, in which a locus is associated with multiple traits.

Cross-phenotype association analysis of gastric cancer: in-silico functional annotation based on the disease-gene network

BACKGROUND

A gene or variant has pleiotropic effects, and genetic variant identification across multiple phenotypes can provide a comprehensive understanding of biological pathways shared among different diseases or phenotypes. Discovery of genetic loci associated with multiple diseases can simultaneously support general interventions. Several meta-analyses have shown genetic associations with gastric cancer (GC); however, no study has identified associations with other phenotypes using this approach.

METHODS

Here, we applied disease network analysis and gene-based analysis (GBA) to examine genetic variants linked to GC and simultaneously associated with other phenotypes. We conducted a single-nucleotide polymorphism (SNP) level meta-analysis and GBA through a systematic genome-wide association study (GWAS) linked to GC, to integrate published results for the SNP variants and group them into major GC-associated genes. We then performed disease network and expression quantitative trait loci (eQTL) analyses to evaluate cross-phenotype associations and expression levels of GC-related genes.

RESULTS

Seven genes (MTX1, GBAP1, MUC1, TRIM46, THBS3, PSCA, and ABO) were associated with GC as well as blood urea nitrogen (BUN), glomerular filtration rate (GFR), and uric acid (UA). In addition, 17 SNPs regulated the expression of genes located on 1q22, 24 SNPs regulated the expression of PSCA on 8q24.3, and rs7849820 regulated the expression of ABO on 9q34.2. Furthermore, rs1057941 and rs2294008 had the highest posterior causal probabilities of being a causal candidate SNP in 1q22, and 8q24.3, respectively.

CONCLUSIONS

These findings identified seven GC-associated genes exhibiting a cross-association with GFR, BUN, and UA.

Univariate and multivariate genome-wide association studies for hematological traits in Murciano-Granadina goats

Hematological traits are important indicators of health status, and they are frequently used as criteria for clinical diagnosis. In humans, the genomic architecture of blood traits has been investigated in depth and thousands of associations with genetic variants have been found. In contrast, the association between marker genotypes and the variation of hematological traits has not been investigated in goats yet. Herewith, we have recorded 12 hematological parameters in 882 Murciano-Granadina goats that were also genotyped with the Goat SNP50 BeadChip (Illumina). Performance of a univariate genome-wide association study (GWAS) made it possible to detect one genomic region on goat chromosome (CHI) 21 (19.2-19.5 Mb) associated, at the genome-wide level of significance, with 4 red blood cell traits. The three markers displaying the highest significances were rs268272996 (CHI21: 19225290 bp), rs268273004 (CHI21: 19565629 bp) and rs268239059 (CHI13: 9615190 bp). Consistently, a multivariate GWAS indicated that the rs268273004 marker on chromosome 21 is associated with seven blood cell traits. Interestingly, this marker maps close to the FA Complementation Group I (FANCI) gene (CHI21: 20021947-20077025 bp), which is functionally related to Fanconi anemia, a syndrome characterized by bone marrow failure, aplastic anemia, and congenital disorders. We have also uncovered additional chromosome-wide significant associations between genetic markers and erythrocyte and leukocyte traits in the univariate GWAS. These findings evidence that the phenotypic variation of hematological traits in goats is regulated, at least to some extent, by polygenic determinants distributed in multiple chromosomes.

Comparison of two multi-trait association testing methods and sequence-based fine mapping of six additive QTL in Swiss Large White pigs

BACKGROUND

Genetic correlations between complex traits suggest that pleiotropic variants contribute to trait variation. Genome-wide association studies (GWAS) aim to uncover the genetic underpinnings of traits. Multivariate association testing and the meta-analysis of summary statistics from single-trait GWAS enable detecting variants associated with multiple phenotypes. In this study, we used array-derived genotypes and phenotypes for 24 reproduction, production, and conformation traits to explore differences between the two methods and used imputed sequence variant genotypes to fine-map six quantitative trait loci (QTL).

RESULTS

We considered genotypes at 44,733 SNPs for 5,753 pigs from the Swiss Large White breed that had deregressed breeding values for 24 traits. Single-trait association analyses revealed eleven QTL that affected 15 traits. Multi-trait association testing and the meta-analysis of the single-trait GWAS revealed between 3 and 6 QTL, respectively, in three groups of traits. The multi-trait methods revealed three loci that were not detected in the single-trait GWAS. Four QTL that were identified in the single-trait GWAS, remained undetected in the multi-trait analyses. To pinpoint candidate causal variants for the QTL, we imputed the array-derived genotypes to the sequence level using a sequenced reference panel consisting of 421 pigs. This approach provided genotypes at 16 million imputed sequence variants with a mean accuracy of imputation of 0.94. The fine-mapping of six QTL with imputed sequence variant genotypes revealed four previously proposed causal mutations among the top variants.

CONCLUSIONS

Our findings in a medium-size cohort of pigs suggest that multivariate association testing and the meta-analysis of summary statistics from single-trait GWAS provide very similar results. Although multi-trait association methods provide a useful overview of pleiotropic loci segregating in mapping populations, the investigation of single-trait association studies is still advised, as multi-trait methods may miss QTL that are uncovered in single-trait GWAS.

Multivariate genome-wide association analysis by iterative hard thresholding

MOTIVATION

In a genome-wide association study, analyzing multiple correlated traits simultaneously is potentially superior to analyzing the traits one by one. Standard methods for multivariate genome-wide association study operate marker-by-marker and are computationally intensive.

RESULTS

We present a sparsity constrained regression algorithm for multivariate genome-wide association study based on iterative hard thresholding and implement it in a convenient Julia package MendelIHT.jl. In simulation studies with up to 100 quantitative traits, iterative hard thresholding exhibits similar true positive rates, smaller false positive rates, and faster execution times than GEMMA's linear mixed models and mv-PLINK's canonical correlation analysis. On UK Biobank data with 470 228 variants, MendelIHT completed a three-trait joint analysis (n=185 656) in 20 h and an 18-trait joint analysis (n=104 264) in 53 h with an 80 GB memory footprint. In short, MendelIHT enables geneticists to fit a single regression model that simultaneously considers the effect of all SNPs and dozens of traits.

AVAILABILITY AND IMPLEMENTATION

Software, documentation, and scripts to reproduce our results are available from https://github.com/OpenMendel/MendelIHT.jl.

Bivariate quantitative Bayesian LASSO for detecting association of rare haplotypes with two correlated continuous phenotypes

In genetic association studies, the multivariate analysis of correlated phenotypes offers statistical and biological advantages compared to analyzing one phenotype at a time. The joint analysis utilizes additional information contained in the correlation and avoids multiple testing. It also provides an opportunity to investigate and understand shared genetic mechanisms of multiple phenotypes. Bivariate logistic Bayesian LASSO (LBL) was proposed earlier to detect rare haplotypes associated with two binary phenotypes or one binary and one continuous phenotype jointly. There is currently no haplotype association test available that can handle multiple continuous phenotypes. In this study, by employing the framework of bivariate LBL, we propose bivariate quantitative Bayesian LASSO (QBL) to detect rare haplotypes associated with two continuous phenotypes. Bivariate QBL removes unassociated haplotypes by regularizing the regression coefficients and utilizing a latent variable to model correlation between two phenotypes. We carry out extensive simulations to investigate the performance of bivariate QBL and compare it with that of a standard (univariate) haplotype association test, Haplo.score (applied twice to two phenotypes individually). Bivariate QBL performs better than Haplo.score in all simulations with varying degrees of power gain. We analyze Genetic Analysis Workshop 19 exome sequencing data on systolic and diastolic blood pressures and detect several rare haplotypes associated with the two phenotypes.

Multivariate GWAS analysis reveals loci associated with liver functions in continental African populations

BACKGROUND

Liver disease is any condition that causes liver damage and inflammation and may likely affect the function of the liver. Vital biochemical screening tools that can be used to evaluate the health of the liver and help diagnose, prevent, monitor, and control the development of liver disease are known as liver function tests (LFT). LFTs are performed to estimate the level of liver biomarkers in the blood. Several factors are associated with differences in concentration levels of LFTs in individuals, such as genetic and environmental factors. The aim of our study was to identify genetic loci associated with liver biomarker levels with a shared genetic basis in continental Africans, using a multivariate genome-wide association study (GWAS) approach.

METHODS

We used two distinct African populations, the Ugandan Genome Resource (UGR = 6,407) and South African Zulu cohort (SZC = 2,598). The six LFTs used in our analysis were: aspartate transaminase (AST), alanine transaminase (ALT), alkaline phosphatase (ALP), gamma-glutamyl transferase (GGT), total bilirubin, and albumin. A multivariate GWAS of LFTs was conducted using the exact linear mixed model (mvLMM) approach implemented in GEMMA and the resulting P-values were presented in Manhattan and quantile-quantile (QQ) plots. First, we attempted to replicate the findings of the UGR cohort in SZC. Secondly, given that the genetic architecture of UGR is different from that of SZC, we further undertook similar analysis in the SZC and discussed the results separately.

RESULTS

A total of 59 SNPs reached genome-wide significance (P = 5x10-8) in the UGR cohort and with 13 SNPs successfully replicated in SZC. These included a novel lead SNP near the RHPN1 locus (lead SNP rs374279268, P-value = 4.79x10-9, Effect Allele Frequency (EAF) = 0.989) and a lead SNP at the RGS11 locus (lead SNP rs148110594, P-value = 2.34x10-8, EAF = 0.928). 17 SNPs were significant in the SZC, while all the SNPs fall within a signal on chromosome 2, rs1976391 mapped to UGT1A was identified as the lead SNP within this region.

CONCLUSIONS

Using multivariate GWAS method improves the power to detect novel genotype-phenotype associations for liver functions not found with the standard univariate GWAS in the same dataset.

Generalization of cortical MOSTest genome-wide associations within and across samples

Genome-Wide Association studies have typically been limited to univariate analysis in which a single outcome measure is tested against millions of variants. Recent work demonstrates that a Multivariate Omnibus Statistic Test (MOSTest) is well powered to discover genomic effects distributed across multiple phenotypes. Applied to cortical brain MRI morphology measures, MOSTest has resulted in a drastic improvement in power to discover loci when compared to established approaches (min-P). One question that arises is how well these discovered loci replicate in independent data. Here we perform 10 times cross validation within 34,973 individuals from UK Biobank for imaging measures of cortical area, thickness and sulcal depth (>1,000 dimensionality for each). By deploying a replication method that aggregates discovered effects distributed across multiple phenotypes, termed PolyVertex Score (MOSTest-PVS), we demonstrate a higher replication yield and comparable replication rate of discovered loci for MOSTest (# replicated loci: 242-496, replication rate: 96-97%) in independent data when compared with the established min-P approach (# replicated loci: 26-55, replication rate: 91-93%). An out-of-sample replication of discovered loci was conducted with a sample of 4,069 individuals from the Adolescent Brain Cognitive Development® (ABCD) study, who are on average 50 years younger than UK Biobank individuals. We observe a higher replication yield and comparable replication rate of MOSTest-PVS compared to min-P. This finding underscores the importance of using well-powered multivariate techniques for both discovery and replication of high dimensional phenotypes in Genome-Wide Association studies.

A Novel Framework for Analysis of the Shared Genetic Background of Correlated Traits

We propose a novel effective framework for the analysis of the shared genetic background for a set of genetically correlated traits using SNP-level GWAS summary statistics. This framework called SHAHER is based on the construction of a linear combination of traits by maximizing the proportion of its genetic variance explained by the shared genetic factors. SHAHER requires only full GWAS summary statistics and matrices of genetic and phenotypic correlations between traits as inputs. Our framework allows both shared and unshared genetic factors to be effectively analyzed. We tested our framework using simulation studies, compared it with previous developments, and assessed its performance using three real datasets: anthropometric traits, psychiatric conditions and lipid concentrations. SHAHER is versatile and applicable to summary statistics from GWASs with arbitrary sample sizes and sample overlaps, allows for the incorporation of different GWAS models (Cox, linear and logistic), and is computationally fast.

The UK Biobank: A Shining Example of Genome-Wide Association Study Science with the Power to Detect the Murky Complications of Real-World Epidemiology

Genome-wide association studies (GWASs) have successfully identified thousands of genetic variants that are reliably associated with human traits. Although GWASs are restricted to certain variant frequencies, they have improved our understanding of the genetic architecture of complex traits and diseases. The UK Biobank (UKBB) has brought substantial analytical opportunity and performance to association studies. The dramatic expansion of many GWAS sample sizes afforded by the inclusion of UKBB data has improved the power of estimation of effect sizes but, critically, has done so in a context where phenotypic depth and precision enable outcome dissection and the application of epidemiological approaches. However, at the same time, the availability of such a large, well-curated, and deeply measured population-based collection has the capacity to increase our exposure to the many complications and inferential complexities associated with GWASs and other analyses. In this review, we discuss the impact that UKBB has had in the GWAS era, some of the opportunities that it brings, and exemplar challenges that illustrate the reality of using data from this world-leading resource.

A genome-wide association study of total child psychiatric problems scores

Substantial genetic correlations have been reported across psychiatric disorders and numerous cross-disorder genetic variants have been detected. To identify the genetic variants underlying general psychopathology in childhood, we performed a genome-wide association study using a total psychiatric problem score. We analyzed 6,844,199 common SNPs in 38,418 school-aged children from 20 population-based cohorts participating in the EAGLE consortium. The SNP heritability of total psychiatric problems was 5.4% (SE = 0.01) and two loci reached genome-wide significance: rs10767094 and rs202005905. We also observed an association of SBF2, a gene associated with neuroticism in previous GWAS, with total psychiatric problems. The genetic effects underlying the total score were shared with common psychiatric disorders only (attention-deficit/hyperactivity disorder, anxiety, depression, insomnia) (rG > 0.49), but not with autism or the less common adult disorders (schizophrenia, bipolar disorder, or eating disorders) (rG < 0.01). Importantly, the total psychiatric problem score also showed at least a moderate genetic correlation with intelligence, educational attainment, wellbeing, smoking, and body fat (rG > 0.29). The results suggest that many common genetic variants are associated with childhood psychiatric symptoms and related phenotypes in general instead of with specific symptoms. Further research is needed to establish causality and pleiotropic mechanisms between related traits.

Multivariate phenotype analysis enables genome-wide inference of mammalian gene function

The function of the majority of genes in the human and mouse genomes is unknown. Investigating and illuminating this dark genome is a major challenge for the biomedical sciences. The International Mouse Phenotyping Consortium (IMPC) is addressing this through the generation and broad-based phenotyping of a knockout (KO) mouse line for every protein-coding gene, producing a multidimensional data set that underlies a genome-wide annotation map from genes to phenotypes. Here, we develop a multivariate (MV) statistical approach and apply it to IMPC data comprising 148 phenotypes measured across 4,548 KO lines. There are 4,256 (1.4% of 302,997 observed data measurements) hits called by the univariate (UV) model analysing each phenotype separately, compared to 31,843 (10.5%) hits in the observed data results of the MV model, corresponding to an estimated 7.5-fold increase in power of the MV model relative to the UV model. One key property of the data set is its 55.0% rate of missingness, resulting from quality control filters and incomplete measurement of some KO lines. This raises the question of whether it is possible to infer perturbations at phenotype-gene pairs at which data are not available, i.e., to infer some in vivo effects using statistical analysis rather than experimentation. We demonstrate that, even at missing phenotypes, the MV model can detect perturbations with power comparable to the single-phenotype analysis, thereby filling in the complete gene-phenotype map with good sensitivity. A factor analysis of the MV model's fitted covariance structure identifies 20 clusters of phenotypes, with each cluster tending to be perturbed collectively. These factors cumulatively explain 75% of the KO-induced variation in the data and facilitate biological interpretation of perturbations. We also demonstrate that the MV approach strengthens the correspondence between IMPC phenotypes and existing gene annotation databases. Analysis of a subset of KO lines measured in replicate across multiple laboratories confirms that the MV model increases power with high replicability.

Rare variants in IFFO1, DTNB, NLRC3 and SLC22A10 associate with Alzheimer's disease CSF profile of neuronal injury and inflammation

Alzheimer's disease (AD) biomarkers represent several neurodegenerative processes, such as synaptic dysfunction, neuronal inflammation and injury, as well as amyloid pathology. We performed an exome-wide rare variant analysis of six AD biomarkers (β-amyloid, total/phosphorylated tau, NfL, YKL-40, and Neurogranin) to discover genes associated with these markers. Genetic and biomarker information was available for 480 participants from two studies: EMIF-AD and ADNI. We applied a principal component (PC) analysis to derive biomarkers combinations, which represent statistically independent biological processes. We then tested whether rare variants in 9576 protein-coding genes associate with these PCs using a Meta-SKAT test. We also tested whether the PCs are intermediary to gene effects on AD symptoms with a SMUT test. One PC loaded on NfL and YKL-40, indicators of neuronal injury and inflammation. Four genes were associated with this PC: IFFO1, DTNB, NLRC3, and SLC22A10. Mediation tests suggest, that these genes also affect dementia symptoms via inflammation/injury. We also observed an association between a PC loading on Neurogranin, a marker for synaptic functioning, with GABBR2 and CASZ1, but no mediation effects. The results suggest that rare variants in IFFO1, DTNB, NLRC3, and SLC22A10 heighten susceptibility to neuronal injury and inflammation, potentially by altering cytoskeleton structure and immune activity disinhibition, resulting in an elevated dementia risk. GABBR2 and CASZ1 were associated with synaptic functioning, but mediation analyses suggest that the effect of these two genes on synaptic functioning is not consequential for AD development.

Multivariate, region-based genetic analyses of facets of reproductive aging in White and Black women

BACKGROUND

Age at final menstrual period (FMP) and the accompanying hormone trajectories across the menopause transition do not occur in isolation, but likely share molecular pathways. Understanding the genetics underlying the endocrinology of the menopause transition may be enhanced by jointly analyzing multiple interrelated traits.

METHODS

In a sample of 347 White and 164 Black women from the Study of Women's Health Across the Nation (SWAN), we investigated pleiotropic effects of 54 candidate genetic regions of interest (ROI) on 5 menopausal traits (age at FMP and premenopausal and postmenopausal levels of follicle stimulation hormone and estradiol) using multivariate kernel regression (Multi-SKAT). A backward elimination procedure was used to identify which subset of traits were most strongly associated with a specific ROI.

RESULTS

In White women, the 20 kb ROI around rs10734411 was significantly associated with the multivariate distribution of age at FMP, premenopausal estradiol, and postmenopausal estradiol (omnibus p-value = .00004). This association did not replicate in the smaller sample of Black women.

CONCLUSION

This study using a region-based, multiple-trait approach suggests a shared genetic basis among multiple facets of reproductive aging.

Multivariate GWAS of Structural Dental Anomalies and Dental Caries in a Multi-Ethnic Cohort

Odontogenesis is a complex process, where disruption can result in dental anomalies and/or increase the risk of developing dental caries. Based on previous studies, certain dental anomalies tend to co-occur in patients, suggesting that these traits may share common genetic and etiological components. The main goal of this study was to implement a multivariate genome-wide association study approach to identify genetic variants shared between correlated structural dental anomalies and dental caries. Our cohort (N = 3,579) was derived from the Pittsburgh Orofacial Clefts Study, where multiple dental traits were assessed in both the unaffected relatives of orofacial cleft (OFC) cases (n = 2,187) and unaffected controls (n = 1,392). We identified four multivariate patterns of correlated traits in this data: tooth agenesis, impaction, and rotation (AIR); enamel hypoplasia, displacement, and rotation (HDR); displacement, rotation, and mamelon (DRM); and dental caries, tooth agenesis and enamel hypoplasia (CAH). We analyzed each of these four models using genome-wide multivariate tests of association. No genome-wide statistically significant results were found, but we identified multiple suggestive association signals (P < 10-5) near genes with known biological roles during tooth development, including ADAMTS9 and PRICKLE2 associated with AIR; GLIS3, WDR72, and ROR2 associated with HDR and DRM; ROBO2 associated with DRM; BMP7 associated with HDR; and ROBO1, SMAD2, and MSX2 associated with CAH. This is the first study to investigate genetic associations for multivariate patterns of correlated dental anomalies and dental caries. Further studies are needed to replicate these results in independent cohorts.

A web-based survey on various symptoms of computer vision syndrome and the genetic understanding based on a multi-trait genome-wide association study

A variety of eye-related symptoms due to the overuse of digital devices is collectively referred to as computer vision syndrome (CVS). In this study, a web-based survey about mind and body functions, including eye strain, was conducted on 1998 Japanese volunteers. To investigate the biological mechanisms behind CVS, a multi-trait genome-wide association study (GWAS), a multivariate analysis on individual-level multivariate data, was performed based on the structural equation modeling methodology assuming a causal pathway for a genetic variant to influence each symptom via a single common latent variable. Twelve loci containing lead variants with a suggestive level of significance were identified. Two loci showed relatively strong signals and were associated with TRABD2B relative to the Wnt signaling pathway and SDK1 having neuronal adhesion and immune functions, respectively. By utilizing publicly available eQTL data, colocalization between GWAS and eQTL signals for four loci was detected, and a locus on 2p25.3 showed a strong colocalization (PPH₄ > 0.9) on retinal MYT1L, known to play an important role in neuronal differentiation. This study suggested that the use of multivariate questionnaire data and multi-trait GWAS can lead to biologically reasonable findings and enhance our genetic understanding of complex relationships among symptoms related to CVS.

AGNEP: An Agglomerative Nesting Clustering Algorithm for Phenotypic Dimension Reduction in Joint Analysis of Multiple Phenotypes

Genome-wide association study (GWAS) has identified thousands of genetic variants associated with complex traits and diseases. Compared with analyzing a single phenotype at a time, the joint analysis of multiple phenotypes can improve statistical power by taking into account the information from phenotypes. However, most established joint algorithms ignore the different level of correlations between multiple phenotypes; instead of that, they simultaneously analyze all phenotypes in a genetic model. Thus, they may fail to capture the genetic structure of phenotypes and consequently reduce the statistical power. In this study, we develop a novel method agglomerative nesting clustering algorithm for phenotypic dimension reduction analysis (AGNEP) to jointly analyze multiple phenotypes for GWAS. First, AGNEP uses an agglomerative nesting clustering algorithm to group correlated phenotypes and then applies principal component analysis (PCA) to generate representative phenotypes for each group. Finally, multivariate analysis is employed to test associations between genetic variants and the representative phenotypes rather than all phenotypes. We perform three simulation experiments with various genetic structures and a real dataset analysis for 19 Arabidopsis phenotypes. Compared to established methods, AGNEP is more powerful in terms of statistical power, computing time, and the number of quantitative trait nucleotides (QTNs). The analysis of the Arabidopsis real dataset further illustrates the efficiency of AGNEP for detecting QTNs, which are confirmed by The Arabidopsis Information Resource gene bank.

Multivariate linear mixed model enhanced the power of identifying genome-wide association to poplar tree heights in a randomized complete block design

With the advances in high-throughput sequencing technologies, it is not difficult to extract tens of thousands of single-nucleotide polymorphisms (SNPs) across many individuals in a fast and cheap way, making it possible to perform genome-wide association studies (GWAS) of quantitative traits in outbred forest trees. It is very valuable to apply traditional breeding experiments in GWAS for identifying genome variants associated with ecologically and economically important traits in Populus. Here, we reported a GWAS of tree height measured at multiple time points from a randomized complete block design (RCBD), which was established with clones from an F₁ hybrid population of Populus deltoides and Populus simonii. A total of 22,670 SNPs across 172 clones in the RCBD were obtained with restriction site-associated DNA sequencing (RADseq) technology. The multivariate mixed linear model was applied by incorporating the pedigree relationship matrix of individuals to test the association of each SNP to the tree heights over 8 time points. Consequently, 41 SNPs were identified significantly associated with the tree height under the P-value threshold determined by Bonferroni correction at the significant level of 0.01. These SNPs were distributed on all but two chromosomes (Chr02 and Chr18) and explained the phenotypic variance ranged from 0.26% to 2.64%, amounting to 63.68% in total. Comparison with previous mapping studies for poplar height as well as the candidate genes of these detected SNPs were also investigated. We therefore showed that the application of multivariate linear mixed model to the longitudinal phenotypic data from the traditional breeding experimental design facilitated to identify far more genome-wide variants for tree height in poplar. The significant SNPs identified in this study would enhance understanding of molecular mechanism for growth traits and would accelerate marker-assisted breeding programs in Populus.

Statistical Impact of Sample Size and Imbalance on Multivariate Analysis in silico and A Case Study in the UK Biobank

Large-scale biobank cohorts coupled with electronic health records offer unprecedented opportunities to study genotype-phenotype relationships. Genome-wide association studies uncovered disease-associated loci through univariate methods, with the focus on one trait at a time. With genetic variants being identifiedfor thousands of traits, researchers found that 90% of human genetic loci are associated with more than one trait, highlighting the ubiquity of pleiotropy. Recently, multivariate methods have been proposed to effectively identify pleiotropy. However, the statistical performance in natural biomedical data, which often have unbalanced case-control sample sizes, is largely known. In this work, we designed 21 scenarios of real-data informed simulations to thoroughly evaluate the statistical characteristics of univariate and multivariate methods. Our results can serve as a reference guide for the application of multivariate methods. We also investigated potential pleiotropy across type II diabetes, Alzheimer's disease, atherosclerosis of arteries, depression, and atherosclerotic heart disease in the UK Biobank.

A two-step approach to testing overall effect of gene-environment interaction for multiple phenotypes

MOTIVATION

While gene-environment (GxE) interactions contribute importantly to many different phenotypes, detecting such interactions requires well-powered studies and has proven difficult. To address this, we combine two approaches to improve GxE power: simultaneously evaluating multiple phenotypes and using a two-step analysis approach. Previous work shows that the power to identify a main genetic effect can be improved by simultaneously analyzing multiple related phenotypes. For a univariate phenotype, two-step methods produce higher power for detecting a GxE interaction compared to single step analysis. Therefore, we propose a two-step approach to test for an overall GxE effect for multiple phenotypes.

RESULTS

Using simulations we demonstrate that, when more than one phenotype has GxE effect (i.e., GxE pleiotropy), our approach offers substantial gain in power (18%-43%) to detect an aggregate-level GxE effect for a multivariate phenotype compared to an analogous two-step method to identify GxE effect for a univariate phenotype. We applied the proposed approach to simultaneously analyze three lipids, LDL, HDL and Triglyceride with the frequency of alcohol consumption as environmental factor in the UK Biobank. The method identified two loci with an overall GxE effect on the vector of lipids, one of which was missed by the competing approaches.

AVAILABILITY

We provide an R package MPGE implementing the proposed approach which is available from CRAN: https://cran.r-project.org/web/packages/MPGE/index.html.

Detecting rare haplotype association with two correlated phenotypes of binary and continuous types

Multiple correlated traits/phenotypes are often collected in genetic association studies and they may share a common genetic mechanism. Joint analysis of correlated phenotypes has well-known advantages over one-at-a-time analysis including gain in power and better understanding of genetic etiology. However, when the phenotypes are of discordant types such as binary and continuous, the joint modeling is more challenging. Another research area of current interest is discovery of rare genetic variants. Currently there is no method available for detecting association of rare (or common) haplotypes with multiple discordant phenotypes jointly. Our goal is to fill this gap specifically for two discordant phenotypes. We consider a rare haplotype association method for a binary phenotype, logistic Bayesian LASSO (univariate LBL) and its extension for two correlated binary phenotypes (bivariate LBL-2B). Under this framework, we propose a haplotype association test with binary and continuous phenotypes jointly (bivariate LBL-BC). Specifically, we use a latent variable to induce correlation between the two phenotypes. We carry out extensive simulations to investigate bivariate LBL-BC and compare it with univariate LBL and bivariate LBL-2B. In most settings, bivariate LBL-BC performs the best. In only two situations, bivariate LBL-BC has similar performance-when the two phenotypes are (1) weakly or not correlated and the target haplotype affects the binary phenotype only and (2) strongly positively correlated and the target haplotype affects both phenotypes in positive direction. Finally, we apply the method to a data set on lung cancer and nicotine dependence and detect several haplotypes including a rare one.

How Well Can Multivariate and Univariate GWAS Distinguish Between True and Spurious Pleiotropy?

Quantification of the simultaneous contributions of loci to multiple traits, a phenomenon called pleiotropy, is facilitated by the increased availability of high-throughput genotypic and phenotypic data. To understand the prevalence and nature of pleiotropy, the ability of multivariate and univariate genome-wide association study (GWAS) models to distinguish between pleiotropic and non-pleiotropic loci in linkage disequilibrium (LD) first needs to be evaluated. Therefore, we used publicly available maize and soybean genotypic data to simulate multiple pairs of traits that were either (i) controlled by quantitative trait nucleotides (QTNs) on separate chromosomes, (ii) controlled by QTNs in various degrees of LD with each other, or (iii) controlled by a single pleiotropic QTN. We showed that multivariate GWAS could not distinguish between QTNs in LD and a single pleiotropic QTN. In contrast, a unique QTN detection rate pattern was observed for univariate GWAS whenever the simulated QTNs were in high LD or pleiotropic. Collectively, these results suggest that multivariate and univariate GWAS should both be used to infer whether or not causal mutations underlying peak GWAS associations are pleiotropic. Therefore, we recommend that future studies use a combination of multivariate and univariate GWAS models, as both models could be useful for identifying and narrowing down candidate loci with potential pleiotropic effects for downstream biological experiments.

Insights into the genetic architecture of the human face

The human face is complex and multipartite, and characterization of its genetic architecture remains challenging. Using a multivariate genome-wide association study meta-analysis of 8,246 European individuals, we identified 203 genome-wide-significant signals (120 also study-wide significant) associated with normal-range facial variation. Follow-up analyses indicate that the regions surrounding these signals are enriched for enhancer activity in cranial neural crest cells and craniofacial tissues, several regions harbor multiple signals with associations to different facial phenotypes, and there is evidence for potential coordinated actions of variants. In summary, our analyses provide insights into the understanding of how complex morphological traits are shaped by both individual and coordinated genetic actions.

Comparison of Single-Trait and Multi-Trait Genome-Wide Association Models and Inclusion of Correlated Traits in the Dissection of the Genetic Architecture of a Complex Trait in a Breeding Program

Unknown genetic architecture makes it difficult to characterize the genetic basis of traits and associated molecular markers because of the complexity of small effect quantitative trait loci (QTLs), environmental effects, and difficulty in phenotyping. Seedling emergence of wheat (Triticum aestivum L.) from deep planting, has a poorly understood genetic architecture, is a vital factor affecting stand establishment and grain yield, and is historically correlated with coleoptile length. This study aimed to dissect the genetic architecture of seedling emergence while accounting for correlated traits using one multi-trait genome-wide association study (MT-GWAS) model and three single-trait GWAS (ST-GWAS) models. The ST-GWAS models included one single-locus model [mixed-linear model (MLM)] and two multi-locus models [fixed and random model circulating probability unification (FarmCPU) and Bayesian information and linkage-disequilibrium iteratively nested keyway (BLINK)]. We conducted GWAS using two populations. The first population consisted of 473 varieties from a diverse association mapping panel phenotyped from 2015 to 2019. The second population consisted of 279 breeding lines phenotyped in 2015 in Lind, WA, with 40,368 markers. We also compared the inclusion of coleoptile length and markers associated with reduced height as covariates in our ST-GWAS models. ST-GWAS found 107 significant markers across 19 chromosomes, while MT-GWAS found 82 significant markers across 14 chromosomes. The FarmCPU and BLINK models, including covariates, were able to identify many small effect markers while identifying large effect markers on chromosome 5A. By using multi-locus model breeding, programs can uncover the complex nature of traits to help identify candidate genes and the underlying architecture of a trait, such as seedling emergence.

Genetic Pleiotropy of Bone-Related Phenotypes: Insights from Osteoporosis

PURPOSE OF REVIEW

We summarize recent evidence on the shared genetics within and outside the musculoskeletal system (mostly related to bone density and osteoporosis).

RECENT FINDINGS

Osteoporosis is determined by an interplay between multiple genetic and environmental factors. Significant progress has been made regarding its genetic background revealing a number of robustly validated loci and respective pathways. However, pleiotropic factors affecting bone and other tissues are not well understood. The analytical methods proposed to test for potential associations between genetic variants and multiple phenotypes can be applied to bone-related data. A number of recent genetic studies have shown evidence of pleiotropy between bone density and other different phenotypes (traits, conditions, or diseases), within and outside the musculoskeletal system. Power benefits of combining correlated phenotypes, as well as unbiased discovery, make these studies promising. Studies in humans are supported by evidence from animal models. Drug development and repurposing should benefit from the pleiotropic approach. We believe that future studies should take into account shared genetics between the bone and related traits.

Fine-mapping genetic associations

Whilst thousands of genetic variants have been associated with human traits, identifying the subset of those variants that are causal requires a further 'fine-mapping' step. We review the basic fine-mapping approach, which is computationally fast and requires only summary data, but depends on an assumption of a single causal variant per associated region which is recognized as biologically unrealistic. We discuss different ways that the approach has been built upon to accommodate multiple causal variants in a region and to incorporate additional layers of functional annotation data. We further review methods for simultaneous fine-mapping of multiple datasets, either exploiting different linkage disequilibrium (LD) structures across ancestries or borrowing information between distinct but related traits. Finally, we look to the future and the opportunities that will be offered by increasingly accurate maps of causal variants for a multitude of human traits.

PCA-based GRS analysis enhances the effectiveness for genetic correlation detection

Genetic risk score (GRS, also known as polygenic risk score) analysis is an increasingly popular method for exploring genetic architectures and relationships of complex diseases. However, complex diseases are usually measured by multiple correlated phenotypes. Analyzing each disease phenotype individually is likely to reduce statistical power due to multiple testing correction. In order to conquer the disadvantage, we proposed a principal component analysis (PCA)-based GRS analysis approach. Extensive simulation studies were conducted to compare the performance of PCA-based GRS analysis and traditional GRS analysis approach. Simulation results observed significantly improved performance of PCA-based GRS analysis compared to traditional GRS analysis under various scenarios. For the sake of verification, we also applied both PCA-based GRS analysis and traditional GRS analysis to a real Caucasian genome-wide association study (GWAS) data of bone geometry. Real data analysis results further confirmed the improved performance of PCA-based GRS analysis. Given that GWAS have flourished in the past decades, our approach may help researchers to explore the genetic architectures and relationships of complex diseases or traits.

Multi-Trait Genome-Wide Association Studies Reveal Loci Associated with Maize Inflorescence and Leaf Architecture

Maize inflorescence is a complex phenotype that involves the physical and developmental interplay of multiple traits. Given the evidence that genes could pleiotropically contribute to several of these traits, we used publicly available maize data to assess the ability of multivariate genome-wide association study (GWAS) approaches to identify pleiotropic quantitative trait loci (pQTL). Our analysis of 23 publicly available inflorescence and leaf-related traits in a diversity panel of n = 281 maize lines genotyped with 376,336 markers revealed that the two multivariate GWAS approaches we tested were capable of identifying pQTL in genomic regions coinciding with similar associations found in previous studies. We then conducted a parallel simulation study on the same individuals, where it was shown that multivariate GWAS approaches yielded a higher true-positive quantitative trait nucleotide (QTN) detection rate than comparable univariate approaches for all evaluated simulation settings except for when the correlated simulated traits had a heritability of 0.9. We therefore conclude that the implementation of state-of-the-art multivariate GWAS approaches is a useful tool for dissecting pleiotropy and their more widespread implementation could facilitate the discovery of genes and other biological mechanisms underlying maize inflorescence.

Multi-trait analysis for genome-wide association study of five psychiatric disorders

We conducted a cross-trait meta-analysis of genome-wide association study on schizophrenia (SCZ) (n = 65,967), bipolar disorder (BD) (n = 41,653), autism spectrum disorder (ASD) (n = 46,350), attention deficit hyperactivity disorder (ADHD) (n = 55,374), and depression (DEP) (n = 688,809). After the meta-analysis, the number of genomic loci increased from 14 to 19 in ADHD, from 3 to 10 in ASD, from 45 to 57 in DEP, from 8 to 54 in BD, and from 64 to 87 in SCZ. We observed significant enrichment of overlapping genes among different disorders and identified a panel of cross-disorder genes. A total of seven genes were found being commonly associated with four out of five psychiatric conditions, namely GABBR1, GLT8D1, HIST1H1B, HIST1H2BN, HIST1H4L, KCNB1, and DCC. The SORCS3 gene was highlighted due to the fact that it was involved in all the five conditions of study. Analysis of correlations unveiled the existence of two clusters of related psychiatric conditions, SCZ and BD that were separate from the other three traits, and formed another group. Our results may provide a new insight for genetic basis of the five psychiatric disorders.

Use of multivariate factor analysis of detailed milk fatty acid profile to perform a genome-wide association study in Italian Simmental and Italian Holstein

Milk fatty acid (FA) profile is a clear example of complex and multiple correlated traits whose genetic basis is difficult to assess. Although genome-wide association (GWA) studies have been successful in the identification of significant genetic variants for complex traits, when correlated phenotypes are analysed separately, the outcomes are difficult to compare and interpret in a metabolic context. Here, we performed a multivariate factor analysis (MFA) on Italian Simmental and Italian Holstein milk fat profiles to extract latent unobserved factors able to explain correlation structure and common metabolic function among different FAs. Individual factor scores obtained by MFA were used to perform a single-SNP based GWA. In both breeds, MFA was able to extract ten latent factors with specific biological meaning, notably: de novo synthesis, desaturation activity and biohydrogenation. The GWA result confirmed the increased power of joint association analysis on multiple correlated traits and allowed us to identify major candidate genes with well-documented function consistent with the metabolic classification of factors obtained, such as DGAT1, FASN and SCD.

Integrate multiple traits to detect novel trait-gene association using GWAS summary data with an adaptive test approach

MOTIVATION

Genetics hold great promise to precision medicine by tailoring treatment to the individual patient based on their genetic profiles. Toward this goal, many large-scale genome-wide association studies (GWAS) have been performed in the last decade to identify genetic variants associated with various traits and diseases. They have successfully identified tens of thousands of disease-related variants. However they have explained only a small proportion of the overall trait heritability for most traits and are of very limited clinical use. This is partly owing to the small effect sizes of most genetic variants, and the common practice of testing association between one trait and one genetic variant at a time in most GWAS, even when multiple related traits are often measured for each individual. Increasing evidence suggests that many genetic variants can influence multiple traits simultaneously, and we can gain more power by testing association of multiple traits simultaneously. It is appealing to develop novel multi-trait association test methods that need only GWAS summary data, since it is generally very hard to access the individual-level GWAS phenotype and genotype data.

RESULTS

Many existing GWAS summary data-based association test methods have relied on ad hoc approach or crude Monte Carlo approximation. In this article, we develop rigorous statistical methods for efficient and powerful multi-trait association test. We develop robust and efficient methods to accurately estimate the marginal trait correlation matrix using only GWAS summary data. We construct the principal component (PC)-based association test from the summary statistics. PC-based test has optimal power when the underlying multi-trait signal can be captured by the first PC, and otherwise it will have suboptimal performance. We develop an adaptive test by optimally weighting the PC-based test and the omnibus chi-square test to achieve robust performance under various scenarios. We develop efficient numerical algorithms to compute the analytical P-values for all the proposed tests without the need of Monte Carlo sampling. We illustrate the utility of proposed methods through application to the GWAS meta-analysis summary data for multiple lipids and glycemic traits. We identify multiple novel loci that were missed by individual trait-based association test.

AVAILABILITY AND IMPLEMENTATION

All the proposed methods are implemented in an R package available at http://www.github.com/baolinwu/MTAR. The developed R programs are extremely efficient: it takes less than 2 min to compute the list of genome-wide significant single nucleotide polymorphisms (SNPs) for all proposed multi-trait tests for the lipids GWAS summary data with 2.5 million SNPs on a single Linux desktop.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

mTADA is a framework for identifying risk genes from de novo mutations in multiple traits

Joint analysis of multiple traits can result in the identification of associations not found through the analysis of each trait in isolation. Studies of neuropsychiatric disorders and congenital heart disease (CHD) which use de novo mutations (DNMs) from parent-offspring trios have reported multiple putatively causal genes. However, a joint analysis method designed to integrate DNMs from multiple studies has yet to be implemented. We here introduce multiple-trait TADA (mTADA) which jointly analyzes two traits using DNMs from non-overlapping family samples. We first demonstrate that mTADA is able to leverage genetic overlaps to increase the statistical power of risk-gene identification. We then apply mTADA to large datasets of >13,000 trios for five neuropsychiatric disorders and CHD. We report additional risk genes for schizophrenia, epileptic encephalopathies and CHD. We outline some shared and specific biological information of intellectual disability and CHD by conducting systems biology analyses of genes prioritized by mTADA.

A Multiple-Trait Bayesian Lasso for Genome-Enabled Analysis and Prediction of Complex Traits

A multiple-trait Bayesian LASSO (MBL) for genome-based analysis and prediction of quantitative traits is presented and applied to two real data sets. The data-generating model is a multivariate linear Bayesian regression on possibly a huge number of molecular markers, and with a Gaussian residual distribution posed. Each (one per marker) of the [Formula: see text] vectors of regression coefficients (T: number of traits) is assigned the same T-variate Laplace prior distribution, with a null mean vector and unknown scale matrix Σ. The multivariate prior reduces to that of the standard univariate Bayesian LASSO when [Formula: see text] The covariance matrix of the residual distribution is assigned a multivariate Jeffreys prior, and Σ is given an inverse-Wishart prior. The unknown quantities in the model are learned using a Markov chain Monte Carlo sampling scheme constructed using a scale-mixture of normal distributions representation. MBL is demonstrated in a bivariate context employing two publicly available data sets using a bivariate genomic best linear unbiased prediction model (GBLUP) for benchmarking results. The first data set is one where wheat grain yields in two different environments are treated as distinct traits. The second data set comes from genotyped Pinus trees, with each individual measured for two traits: rust bin and gall volume. In MBL, the bivariate marker effects are shrunk differentially, i.e., "short" vectors are more strongly shrunk toward the origin than in GBLUP; conversely, "long" vectors are shrunk less. A predictive comparison was carried out as well in wheat, where the comparators of MBL were bivariate GBLUP and bivariate Bayes Cπ-a variable selection procedure. A training-testing layout was used, with 100 random reconstructions of training and testing sets. For the wheat data, all methods produced similar predictions. In Pinus, MBL gave better predictions that either a Bayesian bivariate GBLUP or the single trait Bayesian LASSO. MBL has been implemented in the Julia language package JWAS, and is now available for the scientific community to explore with different traits, species, and environments. It is well known that there is no universally best prediction machine, and MBL represents a new resource in the armamentarium for genome-enabled analysis and prediction of complex traits.

Genome-Wide Association Study of Topsoil Root System Architecture in Field-Grown Soybean [Glycine max (L.) Merr.]

Water and nutrient acquisition is a critical function of plant root systems. Root system architecture (RSA) traits are often complex and controlled by many genes. This is the first genome-wide association study reporting genetic loci for RSA traits for field-grown soybean (Glycine max). A collection of 289 soybean genotypes was grown in three environments, root crowns were excavated, and 12 RSA traits assessed. The first two components of a principal component analysis of these 12 traits were used as additional aggregate traits for a total of 14 traits. Marker-trait association for RSA traits were identified using 31,807 single-nucleotide polymorphisms (SNPs) by a genome-wide association analysis. In total, 283 (non-unique) SNPs were significantly associated with one or more of the 14 root traits. Of these, 246 were unique SNPs and 215 SNPs were associated with a single root trait, while 26, four, and one SNPs were associated with two, three, and four root traits, respectively. The 246 SNPs marked 67 loci associated with at least one of the 14 root traits. Seventeen loci on 13 chromosomes were identified by SNPs associated with more than one root trait. Several genes with annotation related to processes that could affect root architecture were identified near these 67 loci. Additional follow-up studies will be needed to confirm the markers and candidate genes identified for RSA traits and to examine the importance of the different root characteristics for soybean productivity under a range of soil and environmental conditions.

Multivariate Genome-wide Association Analysis of a Cytokine Network Reveals Variants with Widespread Immune, Haematological, and Cardiometabolic Pleiotropy

Cytokines are essential regulatory components of the immune system, and their aberrant levels have been linked to many disease states. Despite increasing evidence that cytokines operate in concert, many of the physiological interactions between cytokines, and the shared genetic architecture that underlies them, remain unknown. Here, we aimed to identify and characterize genetic variants with pleiotropic effects on cytokines. Using three population-based cohorts (n = 9,263), we performed multivariate genome-wide association studies (GWAS) for a correlation network of 11 circulating cytokines, then combined our results in meta-analysis. We identified a total of eight loci significantly associated with the cytokine network, of which two (PDGFRB and ABO) had not been detected previously. In addition, conditional analyses revealed a further four secondary signals at three known cytokine loci. Integration, through the use of Bayesian colocalization analysis, of publicly available GWAS summary statistics with the cytokine network associations revealed shared causal variants between the eight cytokine loci and other traits; in particular, cytokine network variants at the ABO, SERPINE2, and ZFPM2 loci showed pleiotropic effects on the production of immune-related proteins, on metabolic traits such as lipoprotein and lipid levels, on blood-cell-related traits such as platelet count, and on disease traits such as coronary artery disease and type 2 diabetes.

Bayesian multivariate reanalysis of large genetic studies identifies many new associations

Genome-wide association studies (GWAS) have now been conducted for hundreds of phenotypes of relevance to human health. Many such GWAS involve multiple closely-related phenotypes collected on the same samples. However, the vast majority of these GWAS have been analyzed using simple univariate analyses, which consider one phenotype at a time. This is despite the fact that, at least in simulation experiments, multivariate analyses have been shown to be more powerful at detecting associations. Here, we conduct multivariate association analyses on 13 different publicly-available GWAS datasets that involve multiple closely-related phenotypes. These data include large studies of anthropometric traits (GIANT), plasma lipid traits (GlobalLipids), and red blood cell traits (HaemgenRBC). Our analyses identify many new associations (433 in total across the 13 studies), many of which replicate when follow-up samples are available. Overall, our results demonstrate that multivariate analyses can help make more effective use of data from both existing and future GWAS.

Genome-wide association studies (GWAS) have become a common and powerful tool for identifying significant correlations between markers of genetic variation and physical traits of interest. Often these studies are conducted by comparing genetic variation against single traits one at a time (‘univariate’); however, it has previously been shown that it is possible to increase your power to detect significant associations by comparing genetic variation against multiple traits simultaneously (‘multivariate’). Despite this apparent increase in power though, researchers still rarely conduct multivariate GWAS, even when studies have multiple traits readily available. Here, we reanalyze 13 previously published GWAS using a multivariate method and find >400 additional associations. Our method makes use of univariate GWAS summary statistics and is available as a software package, thus making it accessible to other researchers interested in conducting the same analyses. We also show, using studies that have multiple releases, that our new associations have high rates of replication. Overall, we argue multivariate approaches in GWAS should no longer be overlooked and how, often, there is low-hanging fruit in the form of new associations by running these methods on data already collected.

Bivariate logistic Bayesian LASSO for detecting rare haplotype association with two correlated phenotypes

In genetic association studies, joint modeling of related traits/phenotypes can utilize the correlation between them and thereby provide more power and uncover additional information about genetic etiology. Moreover, detecting rare genetic variants are of current scientific interest as a key to missing heritability. Logistic Bayesian LASSO (LBL) has been proposed recently to detect rare haplotype variants using case-control data, that is, a single binary phenotype. As there is currently no haplotype association method that can handle multiple binary phenotypes, we extend LBL to fill this gap. We develop a bivariate model by using a latent variable to induce correlation between the two outcomes. We carry out extensive simulations to investigate the bivariate LBL and compare with the univariate LBL. The bivariate LBL performs better or similar to the univariate LBL in most settings. It has the highest gain in power when a haplotype is associated with both traits and it affects at least one trait in a direction opposite to the direction of the correlation between the traits. We analyze two data sets-Genetic Analysis Workshop 19 sequence data on systolic and diastolic blood pressures and a genome-wide association data set on lung cancer and smoking and detect several associated rare haplotypes.

Multivariate generalized linear model for genetic pleiotropy

When a single gene influences more than one trait, known as pleiotropy, it is important to detect pleiotropy to improve the biological understanding of a gene. This can lead to improved screening, diagnosis, and treatment of diseases. Yet, most current multivariate methods to evaluate pleiotropy test the null hypothesis that none of the traits are associated with a variant; departures from the null could be driven by just one associated trait. A formal test of pleiotropy should assume a null hypothesis that one or fewer traits are associated with a genetic variant. We recently developed statistical methods to analyze pleiotropy for quantitative traits having a multivariate normal distribution. We now extend this approach to traits that can be modeled by generalized linear models, such as analysis of binary, ordinal, or quantitative traits, or a mixture of these types of traits. Based on methods from estimating equations, we developed a new test for pleiotropy. We then extended the testing framework to a sequential approach to test the null hypothesis that $k+1$ traits are associated, given that the null of $k$ associated traits was rejected. This provides a testing framework to determine the number of traits associated with a genetic variant, as well as which traits, while accounting for correlations among the traits. By simulations, we illustrate the Type-I error rate and power of our new methods, describe how they are influenced by sample size, the number of traits, and the trait correlations, and apply the new methods to a genome-wide association study of multivariate traits measuring symptoms of major depression. Our new approach provides a quantitative assessment of pleiotropy, enhancing current analytic practice.