A fast and noise-resilient approach to detect rare-variant associations with deep sequencing data for complex disorders

Excalibur: A new ensemble method based on an optimal combination of aggregation tests for rare-variant association testing for sequencing data

The development of high-throughput next-generation sequencing technologies and large-scale genetic association studies produced numerous advances in the biostatistics field. Various aggregation tests, i.e. statistical methods that analyze associations of a trait with multiple markers within a genomic region, have produced a variety of novel discoveries. Notwithstanding their usefulness, there is no single test that fits all needs, each suffering from specific drawbacks. Selecting the right aggregation test, while considering an unknown underlying genetic model of the disease, remains an important challenge. Here we propose a new ensemble method, called Excalibur, based on an optimal combination of 36 aggregation tests created after an in-depth study of the limitations of each test and their impact on the quality of result. Our findings demonstrate the ability of our method to control type I error and illustrate that it offers the best average power across all scenarios. The proposed method allows for novel advances in Whole Exome/Genome sequencing association studies, able to handle a wide range of association models, providing researchers with an optimal aggregation analysis for the genetic regions of interest.

An Integrative Approach to Analyze Seed Germination in Brassica napus

Seed germination is a complex trait determined by the interaction of hormonal, metabolic, genetic, and environmental components. Variability of this trait in crops has a big impact on seedling establishment and yield in the field. Classical studies of this trait in crops have focused mainly on the analyses of one level of regulation in the cascade of events leading to seed germination. We have carried out an integrative and extensive approach to deepen our understanding of seed germination in Brassica napus by generating transcriptomic, metabolic, and hormonal data at different stages upon seed imbibition. Deep phenotyping of different seed germination-associated traits in six winter-type B. napus accessions has revealed that seed germination kinetics, in particular seed germination speed, are major contributors to the variability of this trait. Metabolic profiling of these accessions has allowed us to describe a common pattern of metabolic change and to identify the levels of malate and aspartate metabolites as putative metabolic markers to estimate germination performance. Additionally, analysis of seed content of different hormones suggests that hormonal balance between ABA, GA, and IAA at crucial time points during this process might underlie seed germination differences in these accessions. In this study, we have also defined the major transcriptome changes accompanying the germination process in B. napus. Furthermore, we have observed that earlier activation of key germination regulatory genes seems to generate the differences in germination speed observed between accessions in B. napus. Finally, we have found that protein-protein interactions between some of these key regulator are conserved in B. napus, suggesting a shared regulatory network with other plant species. Altogether, our results provide a comprehensive and detailed picture of seed germination dynamics in oilseed rape. This new framework will be extremely valuable not only to evaluate germination performance of B. napus accessions but also to identify key targets for crop improvement in this important process.

Association detection between ordinal trait and rare variants based on adaptive combination of P values

Next-generation sequencing technology not only presents a new method for the detection of human genomic structural variation, but also provides a large number of genetic data of rare variants for us. Currently, how to detect association between human complex diseases and rare variants using genetical data has attracted extensive attention. In the field of medicine, many people's health and disease conditions are measured by ordinal response variables, namely, the trait value reflects the development stage or severity of a certain condition. However, most existing methods to test for association between rare variants and complex diseases are designed to deal with dichotomous or quantitative traits. Association analysis methods of ordinal traits are relatively fewer, and considering ordinal traits as dichotomous and quantitative traits will inevitably lose some valuable information in the original data. Therefore, in this paper, we extend an existing method of adaptive combination of P values (ADA) and propose a new method of association analysis for ordinal trait based on it (called OR-ADA) to test for possible association between ordinal trait and rare variants. In our method, we establish a cumulative logistic regression model, in which the regression coefficients are estimated by the Newton-Raphson algorithm and the likelihood ratio test is used to test the association. Through a large number of simulation studies and an example, we demonstrate the performance of the new method and compare it with several methods. The analysis results show that the OR-ADA strategy is robust to the signs of effects of causal variants and more powerful under many scenarios.

DoEstRare: A statistical test to identify local enrichments in rare genomic variants associated with disease

Next-generation sequencing technologies made it possible to assay the effect of rare variants on complex diseases. As an extension of the "common disease-common variant" paradigm, rare variant studies are necessary to get a more complete insight into the genetic architecture of human traits. Association studies of these rare variations show new challenges in terms of statistical analysis. Due to their low frequency, rare variants must be tested by groups. This approach is then hindered by the fact that an unknown proportion of the variants could be neutral. The risk level of a rare variation may be determined by its impact but also by its position in the protein sequence. More generally, the molecular mechanisms underlying the disease architecture may involve specific protein domains or inter-genic regulatory regions. While a large variety of methods are optimizing functionality weights for each single marker, few evaluate variant position differences between cases and controls. Here, we propose a test called DoEstRare, which aims to simultaneously detect clusters of disease risk variants and global allele frequency differences in genomic regions. This test estimates, for cases and controls, variant position densities in the genetic region by a kernel method, weighted by a function of allele frequencies. We compared DoEstRare with previously published strategies through simulation studies as well as re-analysis of real datasets. Based on simulation under various scenarios, DoEstRare was the sole to consistently show highest performance, in terms of type I error and power both when variants were clustered or not. DoEstRare was also applied to Brugada syndrome and early-onset Alzheimer's disease data and provided complementary results to other existing tests. DoEstRare, by integrating variant position information, gives new opportunities to explain disease susceptibility. DoEstRare is implemented in a user-friendly R package.

Selection Probability for Rare Variant Association Studies

In human genome research, genetic association studies of rare variants have been widely studied since the advent of high-throughput DNA sequencing platforms. However, detection of outcome-related rare variants still remains a statistically challenging problem because the number of observed genetic mutations is extremely rare. Recently, a power set-based statistical selection procedure has been proposed to locate both risk and protective rare variants within the outcome-related genes or genetic regions. Although it can perform an individual selection of rare variants, the procedure has a limitation that it cannot measure the certainty of selected rare variants. In this article, we propose a selection probability of individual rare variants, where selection frequencies of rare variants are computed based on bootstrap resampling. Therefore, it can quantify the certainty of both selected and unselected rare variants. Also, a new selection approach using a threshold of selection probability is introduced and compared with some existing selection procedures from extensive simulation studies and real sequencing data analysis. We have demonstrated that the proposed approach outperforms the existing methods in terms of a selection power.

Multiple Group Testing Procedures for Analysis of High-Dimensional Genomic Data

In genetic association studies with high-dimensional genomic data, multiple group testing procedures are often required in order to identify disease/trait-related genes or genetic regions, where multiple genetic sites or variants are located within the same gene or genetic region. However, statistical testing procedures based on an individual test suffer from multiple testing issues such as the control of family-wise error rate and dependent tests. Moreover, detecting only a few of genes associated with a phenotype outcome among tens of thousands of genes is of main interest in genetic association studies. In this reason regularization procedures, where a phenotype outcome regresses on all genomic markers and then regression coefficients are estimated based on a penalized likelihood, have been considered as a good alternative approach to analysis of high-dimensional genomic data. But, selection performance of regularization procedures has been rarely compared with that of statistical group testing procedures. In this article, we performed extensive simulation studies where commonly used group testing procedures such as principal component analysis, Hotelling's T² test, and permutation test are compared with group lasso (least absolute selection and shrinkage operator) in terms of true positive selection. Also, we applied all methods considered in simulation studies to identify genes associated with ovarian cancer from over 20,000 genetic sites generated from Illumina Infinium HumanMethylation27K Beadchip. We found a big discrepancy of selected genes between multiple group testing procedures and group lasso.

Conditioning adaptive combination of P-values method to analyze case-parent trios with or without population controls

Detection of rare causal variants can help uncover the etiology of complex diseases. Recruiting case-parent trios is a popular study design in family-based studies. If researchers can obtain data from population controls, utilizing them in trio analyses can improve the power of methods. The transmission disequilibrium test (TDT) is a well-known method to analyze case-parent trio data. It has been extended to rare-variant association testing (abbreviated as "rvTDT"), with the flexibility to incorporate population controls. The rvTDT method is robust to population stratification. However, power loss may occur in the conditioning process. Here we propose a "conditioning adaptive combination of P-values method" (abbreviated as "conADA"), to analyze trios with/without unrelated controls. By first truncating the variants with larger P-values, we decrease the vulnerability of conADA to the inclusion of neutral variants. Moreover, because the test statistic is developed by conditioning on parental genotypes, conADA generates valid statistical inference in the presence of population stratification. With regard to statistical methods for next-generation sequencing data analyses, validity may be hampered by population stratification, whereas power may be affected by the inclusion of neutral variants. We recommend conADA for its robustness to these two factors (population stratification and the inclusion of neutral variants).

Beyond Rare-Variant Association Testing: Pinpointing Rare Causal Variants in Case-Control Sequencing Study

Rare-variant association testing usually requires some method of aggregation. The next important step is to pinpoint individual rare causal variants among a large number of variants within a genetic region. Recently Ionita-Laza et al. propose a backward elimination (BE) procedure that can identify individual causal variants among the many variants in a gene. The BE procedure removes a variant if excluding this variant can lead to a smaller P-value for the BURDEN test (referred to as "BE-BURDEN") or the SKAT test (referred to as "BE-SKAT"). We here use the adaptive combination of P-values (ADA) method to pinpoint causal variants. Unlike most gene-based association tests, the ADA statistic is built upon per-site P-values of individual variants. It is straightforward to select important variants given the optimal P-value truncation threshold found by ADA. We performed comprehensive simulations to compare ADA with BE-SKAT and BE-BURDEN. Ranking these three approaches according to positive predictive values (PPVs), the percentage of truly causal variants among the total selected variants, we found ADA > BE-SKAT > BE-BURDEN across all simulation scenarios. We therefore recommend using ADA to pinpoint plausible rare causal variants in a gene.

Detecting association of rare and common variants by adaptive combination of P-values

Genome-wide association studies (GWAS) can detect common variants associated with diseases. Next generation sequencing technology has made it possible to detect rare variants. Most of association tests, including burden tests and nonburden tests, mainly target rare variants by upweighting rare variant effects and downweighting common variant effects. But there is increasing evidence that complex diseases are caused by both common and rare variants. In this paper, we extend the ADA method (adaptive combination of P-values; Lin et al., 2014) for rare variants only and propose a RC-ADA method (common and rare variants by adaptive combination of P-values). Our proposed method combines the per-site P-values with the weights based on minor allele frequencies (MAFs). The RC-ADA is robust to directions of effects of causal variants and inclusion of a high proportion of neutral variants. The performance of the RC-ADA method is compared with several other association methods. Extensive simulation studies show that the RC-ADA method is more powerful than other association methods over a wide range of models.

Group association test using a hidden Markov model

In the genomic era, group association tests are of great interest. Due to the overwhelming number of individual genomic features, the power of testing for association of a single genomic feature at a time is often very small, as are the effect sizes for most features. Many methods have been proposed to test association of a trait with a group of features within a functional unit as a whole, e.g. all SNPs in a gene, yet few of these methods account for the fact that generally a substantial proportion of the features are not associated with the trait. In this paper, we propose to model the association for each feature in the group as a mixture of features with no association and features with non-zero associations to explicitly account for the possibility that a fraction of features may not be associated with the trait while other features in the group are. The feature-level associations are first estimated by generalized linear models; the sequence of these estimated associations is then modeled by a hidden Markov chain. To test for global association, we develop a modified likelihood ratio test based on a log-likelihood function that ignores higher order dependency plus a penalty term. We derive the asymptotic distribution of the likelihood ratio test under the null hypothesis. Furthermore, we obtain the posterior probability of association for each feature, which provides evidence of feature-level association and is useful for potential follow-up studies. In simulations and data application, we show that our proposed method performs well when compared with existing group association tests especially when there are only few features associated with the outcome.

Reproducible simulations of realistic samples for next-generation sequencing studies using Variant Simulation Tools

Computer simulations have been widely used to validate and evaluate the power of statistical methods for genetic epidemiological studies. Although a large number of simulation methods and software packages have been developed for genome-wide association studies, methodological and bioinformatics challenges have limited their applications in simulating datasets for whole-genome and whole-exome sequencing studies. With the development of more sophisticated statistical methods that make fuller use of available data and our knowledge of the human genome, there is a pressing need for genetic simulators that capture more features of empirical data (e.g., multiallele variants, indels, use of the Variant Call Format) and the human genome (e.g., functional annotations of genetic variants). This article introduces Variant Simulation Tools (VST), a module of Variant Tools for the simulation of genetic variants for sequencing-based genetic epidemiological studies. Although multiple simulation engines are provided, the core of VST is a novel forward-time simulation engine that simulates real nucleotide sequences of the human genome using DNA mutation models, fine-scale recombination maps, and a selection model based on amino acid changes of translated protein sequences. The design of VST allows users to easily create and distribute simulation methods and simulated datasets for a variety of applications and encourages fair comparison between statistical methods through the use of existing or reproduced simulated datasets.

Adaptive combination of P-values for family-based association testing with sequence data

Family-based study design will play a key role in identifying rare causal variants, because rare causal variants can be enriched in families with multiple affected subjects. Furthermore, different from population-based studies, family studies are robust to bias induced by population substructure. It is well known that rare causal variants are difficult to detect from single-locus tests. Therefore, burden tests and non-burden tests have been developed, by combining signals of multiple variants in a chromosomal region or a functional unit. This inevitably incorporates some neutral variants into the test statistics, which can dilute the power of statistical methods. To guard against the noise caused by neutral variants, we here propose an 'adaptive combination of P-values method' (abbreviated as 'ADA'). This method combines per-site P-values of variants that are more likely to be causal. Variants with large P-values (which are more likely to be neutral variants) are discarded from the combined statistic. In addition to performing extensive simulation studies, we applied these tests to the Genetic Analysis Workshop 17 data sets, where real sequence data were generated according to the 1000 Genomes Project. Compared with some existing methods, ADA is more robust to the inclusion of neutral variants. This is a merit especially when dichotomous traits are analyzed. However, there are some limitations for ADA. First, it is more computationally intensive. Second, pedigree structures and founders' sequence data are required for the permutation procedure. Third, unrelated controls cannot be included. We here show that, for family-based studies, the application of ADA is limited to dichotomous trait analyses with full pedigree information.

A powerful and adaptive association test for rare variants

This article focuses on conducting global testing for association between a binary trait and a set of rare variants (RVs), although its application can be much broader to other types of traits, common variants (CVs), and gene set or pathway analysis. We show that many of the existing tests have deteriorating performance in the presence of many nonassociated RVs: their power can dramatically drop as the proportion of nonassociated RVs in the group to be tested increases. We propose a class of so-called sum of powered score (SPU) tests, each of which is based on the score vector from a general regression model and hence can deal with different types of traits and adjust for covariates, e.g., principal components accounting for population stratification. The SPU tests generalize the sum test, a representative burden test based on pooling or collapsing genotypes of RVs, and a sum of squared score (SSU) test that is closely related to several other powerful variance component tests; a previous study (Basu and Pan 2011) has demonstrated good performance of one, but not both, of the Sum and SSU tests in many situations. The SPU tests are versatile in the sense that one of them is often powerful, although its identity varies with the unknown true association parameters. We propose an adaptive SPU (aSPU) test to approximate the most powerful SPU test for a given scenario, consequently maintaining high power and being highly adaptive across various scenarios. We conducted extensive simulations to show superior performance of the aSPU test over several state-of-the-art association tests in the presence of many nonassociated RVs. Finally we applied the SPU and aSPU tests to the GAW17 mini-exome sequence data to compare its practical performance with some existing tests, demonstrating their potential usefulness.

Variant association tools for quality control and analysis of large-scale sequence and genotyping array data

Currently there is great interest in detecting associations between complex traits and rare variants. In this report, we describe Variant Association Tools (VAT) and the VAT pipeline, which implements best practices for rare-variant association studies. Highlights of VAT include variant-site and call-level quality control (QC), summary statistics, phenotype- and genotype-based sample selection, variant annotation, selection of variants for association analysis, and a collection of rare-variant association methods for analyzing qualitative and quantitative traits. The association testing framework for VAT is regression based, which readily allows for flexible construction of association models with multiple covariates and weighting themes based on allele frequencies or predicted functionality. Additionally, pathway analyses, conditional analyses, and analyses of gene-gene and gene-environment interactions can be performed. VAT is capable of rapidly scanning through data by using multi-process computation, adaptive permutation, and simultaneously conducting association analysis via multiple methods. Results are available in text or graphic file formats and additionally can be output to relational databases for further annotation and filtering. An interface to R language also facilitates user implementation of novel association methods. The VAT's data QC and association-analysis pipeline can be applied to sequence, imputed, and genotyping array, e.g., "exome chip," data, providing a reliable and reproducible computational environment in which to analyze small- to large-scale studies with data from the latest genotyping and sequencing technologies. Application of the VAT pipeline is demonstrated through analysis of data from the 1000 Genomes project.

A power set-based statistical selection procedure to locate susceptible rare variants associated with complex traits with sequencing data

MOTIVATION

Existing association methods for rare variants from sequencing data have focused on aggregating variants in a gene or a genetic region because of the fact that analysing individual rare variants is underpowered. However, these existing rare variant detection methods are not able to identify which rare variants in a gene or a genetic region of all variants are associated with the complex diseases or traits. Once phenotypic associations of a gene or a genetic region are identified, the natural next step in the association study with sequencing data is to locate the susceptible rare variants within the gene or the genetic region.

RESULTS

In this article, we propose a power set-based statistical selection procedure that is able to identify the locations of the potentially susceptible rare variants within a disease-related gene or a genetic region. The selection performance of the proposed selection procedure was evaluated through simulation studies, where we demonstrated the feasibility and superior power over several comparable existing methods. In particular, the proposed method is able to handle the mixed effects when both risk and protective variants are present in a gene or a genetic region. The proposed selection procedure was also applied to the sequence data on the ANGPTL gene family from the Dallas Heart Study to identify potentially susceptible rare variants within the trait-related genes.

AVAILABILITY AND IMPLEMENTATION

An R package 'rvsel' can be downloaded from http://www.columbia.edu/∼sw2206/ and http://statsun.pusan.ac.kr.

Association testing of clustered rare causal variants in case-control studies

Biological evidence suggests that multiple causal variants in a gene may cluster physically. Variants within the same protein functional domain or gene regulatory element would locate in close proximity on the DNA sequence. However, spatial information of variants is usually not used in current rare variant association analyses. We here propose a clustering method (abbreviated as "CLUSTER"), which is extended from the adaptive combination of P-values. Our method combines the association signals of variants that are more likely to be causal. Furthermore, the statistic incorporates the spatial information of variants. With extensive simulations, we show that our method outperforms several commonly-used methods in many scenarios. To demonstrate its use in real data analyses, we also apply this CLUSTER test to the Dallas Heart Study data. CLUSTER is among the best methods when the effects of causal variants are all in the same direction. As variants located in close proximity are more likely to have similar impact on disease risk, CLUSTER is recommended for association testing of clustered rare causal variants in case-control studies.

Rare variant association testing by adaptive combination of P-values

With the development of next-generation sequencing technology, there is a great demand for powerful statistical methods to detect rare variants (minor allele frequencies (MAFs)<1%) associated with diseases. Testing for each variant site individually is known to be underpowered, and therefore many methods have been proposed to test for the association of a group of variants with phenotypes, by pooling signals of the variants in a chromosomal region. However, this pooling strategy inevitably leads to the inclusion of a large proportion of neutral variants, which may compromise the power of association tests. To address this issue, we extend the -MidP method (Cheung et al., 2012, Genet Epidemiol 36: 675–685) and propose an approach (named ‘adaptive combination of P-values for rare variant association testing’, abbreviated as ‘ADA’) that adaptively combines per-site P-values with the weights based on MAFs. Before combining P-values, we first imposed a truncation threshold upon the per-site P-values, to guard against the noise caused by the inclusion of neutral variants. This ADA method is shown to outperform popular burden tests and non-burden tests under many scenarios. ADA is recommended for next-generation sequencing data analysis where many neutral variants may be included in a functional region.

The value of statistical or bioinformatics annotation for rare variant association with quantitative trait

In the past few years, a plethora of methods for rare variant association with phenotype have been proposed. These methods aggregate information from multiple rare variants across genomic region(s), but there is little consensus as to which method is most effective. The weighting scheme adopted when aggregating information across variants is one of the primary determinants of effectiveness. Here we present a systematic evaluation of multiple weighting schemes through a series of simulations intended to mimic large sequencing studies of a quantitative trait. We evaluate existing phenotype-independent and phenotype-dependent methods, as well as weights estimated by penalized regression approaches including Lasso, Elastic Net, and SCAD. We find that the difference in power between phenotype-dependent schemes is negligible when high-quality functional annotations are available. When functional annotations are unavailable or incomplete, all methods suffer from power loss; however, the variable selection methods outperform the others at the cost of increased computational time. Therefore, in the absence of good annotation, we recommend variable selection methods (which can be viewed as "statistical annotation") on top of regions implicated by a phenotype-independent weighting scheme. Further, once a region is implicated, variable selection can help to identify potential causal single nucleotide polymorphisms for biological validation. These findings are supported by an analysis of a high coverage targeted sequencing study of 1,898 individuals.

Applications of high-throughput DNA sequencing to benign hematology

The development of novel technologies for high-throughput DNA sequencing is having a major impact on our ability to measure and define normal and pathologic variation in humans. This review discusses advances in DNA sequencing that have been applied to benign hematologic disorders, including those affecting the red blood cell, the neutrophil, and other white blood cell lineages. Relevant examples of how these approaches have been used for disease diagnosis, gene discovery, and studying complex traits are provided. High-throughput DNA sequencing technology holds significant promise for impacting clinical care. This includes development of improved disease detection and diagnosis, better understanding of disease progression and stratification of risk of disease-specific complications, and development of improved therapeutic strategies, particularly patient-specific pharmacogenomics-based therapy, with monitoring of therapy by genomic biomarkers.

MetaSeq: privacy preserving meta-analysis of sequencing-based association studies

Human genetics recently transitioned from GWAS to studies based on NGS data. For GWAS, small effects dictated large sample sizes, typically made possible through meta-analysis by exchanging summary statistics across consortia. NGS studies groupwise-test for association of multiple potentially-causal alleles along each gene. They are subject to similar power constraints and therefore likely to resort to meta-analysis as well. The problem arises when considering privacy of the genetic information during the data-exchange process. Many scoring schemes for NGS association rely on the frequency of each variant thus requiring the exchange of identity of the sequenced variant. As such variants are often rare, potentially revealing the identity of their carriers and jeopardizing privacy. We have thus developed MetaSeq, a protocol for meta-analysis of genome-wide sequencing data by multiple collaborating parties, scoring association for rare variants pooled per gene across all parties. We tackle the challenge of tallying frequency counts of rare, sequenced alleles, for metaanalysis of sequencing data without disclosing the allele identity and counts, thereby protecting sample identity. This apparent paradoxical exchange of information is achieved through cryptographic means. The key idea is that parties encrypt identity of genes and variants. When they transfer information about frequency counts in cases and controls, the exchanged data does not convey the identity of a mutation and therefore does not expose carrier identity. The exchange relies on a 3rd party, trusted to follow the protocol although not trusted to learn about the raw data. We show applicability of this method to publicly available exome-sequencing data from multiple studies, simulating phenotypic information for powerful meta-analysis. The MetaSeq software is publicly available as open source.