The data set simulated for Genetic Analysis Workshop 17 was designed to mimic a subset of data that might be produced in a full exome screen for a complex disorder and related risk factors in order to permit workshop participants to investigate issues of study design and statistical genetic analysis. Real sequence data from the 1000 Genomes Project formed the basis for simulating a common disease trait with a prevalence of 30% and three related quantitative risk factors in a sample of 697 unrelated individuals and a second sample of 697 individuals in large, extended pedigrees. Called genotypes for 24,487 autosomal markers assigned to 3,205 genes and simulated affection status, quantitative traits, age, sex, pedigree relationships, and cigarette smoking were provided to workshop participants. The simulating model included both common and rare variants with minor allele frequencies ranging from 0.07% to 25.8% and a wide range of effect sizes for these variants. Genotype-smoking interaction effects were included for variants in one gene. Functional variants were concentrated in genes selected from specific biological pathways and were selected on the basis of the predicted deleteriousness of the coding change. For each sample, unrelated individuals and family, 200 replicates of the phenotypes were simulated.
Next-generation sequencing technologies are rapidly changing the field of genetic epidemiology and enabling exploration of the full allele frequency spectrum underlying complex diseases. Although sequencing technologies have shifted our focus toward rare genetic variants, statistical methods traditionally used in genetic association studies are inadequate for estimating effects of low minor allele frequency variants. Four our study we use the Genetic Analysis Workshop 17 data from 697 unrelated individuals (genotypes for 24,487 autosomal variants from 3,205 genes). We apply a Bayesian hierarchical mixture model to identify genes associated with a simulated binary phenotype using a transformed genotype design matrix weighted by allele frequencies. A Metropolis Hasting algorithm is used to jointly sample each indicator variable and additive genetic effect pair from its conditional posterior distribution, and remaining parameters are sampled by Gibbs sampling. This method identified 58 genes with a posterior probability greater than 0.8 for being associated with the phenotype. One of these 58 genes, PIK3C2B was correctly identified as being associated with affected status based on the simulation process. This project demonstrates the utility of Bayesian hierarchical mixture models using a transformed genotype matrix to detect genes containing rare and common variants associated with a binary phenotype.
We use least absolute shrinkage and selection operator (LASSO) regression to select genetic markers and phenotypic features that are most informative with respect to a trait of interest. We compare several strategies for applying LASSO methods in risk prediction models, using the Genetic Analysis Workshop 17 exome simulation data consisting of 697 individuals with information on genotypic and phenotypic features (smoking, age, sex) in 5-fold cross-validated fashion. The cross-validated averages of the area under the receiver operating curve range from 0.45 to 0.63 for different strategies using only genotypic markers. The same values are improved to 0.69–0.87 when both genotypic and phenotypic information are used. The ability of the LASSO method to find true causal markers is limited, but the method was able to discover several common variants (e.g., FLT1) under certain conditions.
Rare causal variants are believed to significantly contribute to the genetic basis of common diseases or quantitative traits. Appropriate statistical methods are required to discover the highest possible number of disease-relevant variants in a genome-wide screening study. The publicly available Genetic Analysis Workshop 17 data set consists of 697 individuals and 24,487 genetic variants. It includes a simulated complex disease model with intermediate quantitative phenotypes. We compare four gene-wise scoring methods with respect to ranking of causal genes under variable allele frequency thresholds for collapsing of rare variants and considering whether or not rare variants were included. We also compare causal genes for which the ranks differ clearly between scoring methods regarding such characteristics as number and strength of causal variants. We corroborated our findings with additional simulations. We found that the maximum statistics method was superior in assigning high ranks to genes with a single strong causal variant. Hotelling’s T2 test was superior for genes with several independent causal variants. This was consistent for all phenotypes and was confirmed by single-gene analyses and additional simulations. The multivariate analysis performed similarly to Hotelling’s T2 test. The least absolute shrinkage and selection operator (LASSO) analysis was widely comparable with the maximum statistics method. We conclude that the maximum statistics method is a superior alternative to Hotelling’s T2 test if one expects only one independent causal variant per gene with a dominating effect. Such a variant could also be a supermarker derived by collapsing rare variants. Because the true nature of the genetic effect is unknown for real data, both methods need to be taken into consideration.
We compared family-based single-marker association analysis using Merlin and multi-marker analysis using LASSO (least absolute shrinkage and selection operator) for the low-density lipoprotein phenotype at the first visit for all 200 replicates of the Genetic Analysis Workshop 16 Framingham simulated data sets. Using "answers," we selected single-nucleotide polymorphisms (SNPs) on chromosome 22 for comparison of results between single-marker and multi-marker analyses. For the major causal SNP rs2294207 on chromosome 22, both single-marker and multi-marker analyses provided similar results, indicating the importance of this SNP. For the 12 polygenic SNPs on the same chromosome, both single-marker and multi-marker analyses failed to provide statistically significant associations, indicating that their effects were too weak to be detected by either method. The main difference between the two methods was that for the 14 SNPs near the causal SNPs, p-values from Merlin were the next smallest, whereas LASSO often excluded these non-causal neighboring SNPs entirely from the first 10,000 models.
Identifying rare variants that are responsible for complex disease has been promoted by advances in sequencing technologies. However, statistical methods that can handle the vast amount of data generated and that can interpret the complicated relationship between disease and these variants have lagged. We apply a zero-inflated Poisson regression model to take into account the excess of zeros caused by the extremely low frequency of the 24,487 exonic variants in the Genetic Analysis Workshop 17 data. We grouped the 697 subjects in the data set as Europeans, Asians, and Africans based on principal components analysis and found the total number of rare variants per gene for each individual. We then analyzed these collapsed variants based on the assumption that rare variants are enriched in a group of people affected by a disease compared to a group of unaffected people. We also tested the hypothesis with quantitative traits Q1, Q2, and Q4. Analyses performed on the combined 697 individuals and on each ethnic group yielded different results. For the combined population analysis, we found that UGT1A1, which was not part of the simulation model, was associated with disease liability and that FLT1, which was a causal locus in the simulation model, was associated with Q1. Of the causal loci in the simulation models, FLT1 and KDR were associated with Q1 and VNN1 was correlated with Q2. No significant genes were associated with Q4. These results show the feasibility and capability of our new statistical model to detect multiple rare variants influencing disease risk.
Methods that can evaluate aggregate effects of rare and common variants are limited. Therefore, we applied a two-stage approach to evaluate aggregate gene effects in the 1000 Genomes Project data, which contain 24,487 single-nucleotide polymorphisms (SNPs) in 697 unrelated individuals from 7 populations. In stage 1, we identified potentially interesting genes (PIGs) as those having at least one SNP meeting Bonferroni correction using univariate, multiple regression models. In stage 2, we evaluate aggregate PIG effects on trait, Q1, by modeling each gene as a latent construct, which is defined by multiple common and rare variants, using the multivariate statistical framework of structural equation modeling (SEM). In stage 1, we found that PIGs varied markedly between a randomly selected replicate (replicate 137) and 100 other replicates, with the exception of FLT1. In stage 1, collapsing rare variants decreased false positives but increased false negatives. In stage 2, we developed a good-fitting SEM model that included all nine genes simulated to affect Q1 (FLT1, KDR, ARNT, ELAV4, FLT4, HIF1A, HIF3A, VEGFA, VEGFC) and found that FLT1 had the largest effect on Q1 (βstd = 0.33 ± 0.05). Using replicate 137 estimates as population values, we found that the mean relative bias in the parameters (loadings, paths, residuals) and their standard errors across 100 replicates was on average, less than 5%. Our latent variable SEM approach provides a viable framework for modeling aggregate effects of rare and common variants in multiple genes, but more elegant methods are needed in stage 1 to minimize type I and type II error.
We use a novel penalized approach for genome-wide association study that accounts for the linkage disequilibrium between adjacent markers. This method uses a penalty on the difference of the genetic effect at adjacent single-nucleotide polymorphisms and combines it with the minimax concave penalty, which has been shown to be superior to the least absolute shrinkage and selection operator (LASSO) in terms of estimator bias and selection consistency. Our method is implemented using a coordinate descent algorithm. The value of the tuning parameters is determined by extended Bayesian information criteria. The leave-one-out method is used to compute p-values of selected single-nucleotide polymorphisms. Its applicability to a simulated data from Genetic Analysis Workshop 17 replication one is illustrated. Our method selects three SNPs (C13S522, C13S523, and C13S524), whereas the LASSO method selects two SNPs (C13S522 and C13S523).
Variable selection in genome-wide association studies can be a daunting task and statistically challenging because there are more variables than subjects. We propose an approach that uses principal-component analysis (PCA) and least absolute shrinkage and selection operator (LASSO) to identify gene-gene interaction in genome-wide association studies. A PCA was used to first reduce the dimension of the single-nucleotide polymorphisms (SNPs) within each gene. The interaction of the gene PCA scores were placed into LASSO to determine whether any gene-gene signals exist. We have extended the PCA-LASSO approach using the bootstrap to estimate the standard errors and confidence intervals of the LASSO coefficient estimates. This method was compared to placing the raw SNP values into the LASSO and the logistic model with individual gene-gene interaction. We demonstrated these methods with the Genetic Analysis Workshop 16 rheumatoid arthritis genome-wide association study data and our results identified a few gene-gene signals. Based on our results, the PCA-LASSO method shows promise in identifying gene-gene interactions, and, at this time we suggest using it with other conventional approaches, such as generalized linear models, to narrow down genetic signals.
Because multiple loci control complex diseases, there is great interest in testing markers simultaneously instead of one by one. In this paper, we applied two model selection algorithms: the stochastic search variable selection (SSVS) and the least absolute shrinkage and selection operator (LASSO) to two quantitative phenotypes related to rheumatoid arthritis (RA).
The Genetic Analysis Workshop 16 data includes 2,062 unrelated individuals and 545,080 single-nucleotide polymorphism markers from the Illumina 550 k chip. We performed our analyses on the cases as the quantitative phenotype data was not provided for the controls. The performance of the two algorithms was compared. Using sure independence screening as the prescreening procedure, both SSVS and LASSO give small models. No markers are identified in the human leukocyte antigen region of chromosome 6 that was shown to be associated with RA. SSVS and LASSO identify seven common loci, and some of them are on genes LRRC8D, LRP1B, and COLEC12. These genes have not been reported to be associated with RA. LASSO also identified a common locus on gene KTCD21 for the two phenotypes (marker rs230662 and rs483731, respectively).
SSVS outperforms LASSO in simulation studies. Both SSVS and LASSO give small models on the RA data, however this depends on model parameters. We also demonstrate the ability of both LASSO and SSVS to handle more markers than the number of samples.
Genome-wide association studies are a powerful approach used to identify common variants for complex disease. However, the traditional genome-wide association methods may not be optimal when they are applied to rare variants because of the rare variants’ low frequencies and weak signals. To alleviate the difficulty, investigators have proposed many methods that collapse rare variants. In this paper, we propose a novel ranking method, which we call stability selection based on random collapsing, to rank the candidate rare variants. We use the simulated mini-exome data sets of unrelated individuals from Genetic Analysis Workshop 17 for the analysis. The numerical results suggest that the selection based on a random collapsing method is promising for identifying functional rare variants in genome-wide association studies. Further research to examine the error control property of the proposed method is underway.
The Genetic Analysis Workshop 15 Problem 3 simulated rheumatoid arthritis data set provided 100 replicates of simulated single-nucleotide polymorphism (SNP) and covariate data sets for 1500 families with an affected sib pair and 2000 controls, modeled after real rheumatoid arthritis data. The data generation model included nine unobserved trait loci, most of which have one or more of the generated SNPs associated with them. These data sets provide an ideal experimental test bed for evaluating new and old algorithms for selecting SNPs and covariates that can separate cases from controls, because the cases and controls are known as well as the identities of the trait loci. LASSO-Patternsearch is a new multi-step algorithm with a LASSO-type penalized likelihood method at its core specifically designed to detect and model interactions between important predictor variables. In this article the original LASSO-Patternsearch algorithm is modified to handle the large number of SNPs plus covariates. We start with a screen step within the framework of parametric logistic regression. The patterns that survived the screen step were further selected by a penalized logistic regression with the LASSO penalty. And finally, a parametric logistic regression model were built on the patterns that survived the LASSO step. In our analysis of Genetic Analysis Workshop 15 Problem 3 data we have identified most of the associated SNPs and relevant covariates. Upon using the model as a classifier, very competitive error rates were obtained.
The least absolute shrinkage and selection operator (LASSO) can be used to predict SNP effects. This operator has the desirable feature of including in the model only a subset of explanatory SNPs, which can be useful both in QTL detection and GWS studies. LASSO solutions can be obtained by the least angle regression (LARS) algorithm. The big issue with this procedure is to define the best constraint (t), i.e. the upper bound of the sum of absolute value of the SNP effects which roughly corresponds to the number of SNPs to be selected. Usai et al. (2009) dealt with this problem by a cross-validation approach and defined t as the average number of selected SNPs overall replications. Nevertheless, in small size populations, such estimator could give underestimated values of t. Here we propose two alternative ways to define t and compared them with the "classical" one.
The first (strategy 1), was based on 1,000 cross-validations carried out by randomly splitting the reference population (2,000 individuals with performance) into two halves. The value of t was the number of SNPs which occurred in more than 5% of replications. The second (strategy 2), which did not use cross-validations, was based on the minimization of the Cp-type selection criterion which depends on the number of selected SNPs and the expected residual variance.
The size of the subset of selected SNPs was 46, 189 and 64 for the classical approach, strategy 1 and 2 respectively. Classical and strategy 2 gave similar results and indicated quite clearly the regions were QTL with additive effects were located. Strategy 1 confirmed such regions and added further positions which gave a less clear scenario. Correlation between GEBVs estimated with the three strategies and TBVs in progenies without phenotypes were 0.9237, 0.9000 and 0.9240 for classical, strategy 1 and 2 respectively.
This suggests that the Cp-type selection criterion is a valid alternative to the cross-validations to define the best constraint for selecting subsets of predicting SNPs by LASSO-LARS procedure.
Gene-based and single-nucleotide polymorphism (SNP) set association studies provide an important complement to SNP analysis. Kernel-based nonparametric regression has recently emerged as a powerful and flexible tool for this purpose. Our goal is to explore whether this approach can be extended to incorporate and test for interaction effects, especially for genes containing rare variant SNPs. Here, we construct nonparametric regression models that can be used to include a gene-environment interaction effect under the framework of the least-squares kernel machine and examine the performance of the proposed method on the Genetic Analysis Workshop 17 unrelated individuals data set. Two hundred simulated replicates were used to explore the power for detecting interaction. We demonstrate through a genome scan of the quantitative phenotype Q1 that the simulated gene-environment interaction effect in the data can be detected with reasonable power by using the least-squares kernel machine method.
Using the exome sequencing data from 697 unrelated individuals and their simulated disease phenotypes from Genetic Analysis Workshop 17, we develop and apply a gene-based method to identify the relationship between a gene with multiple rare genetic variants and a phenotype. The method is based on the Mantel test, which assesses the correlation between two distance matrices using a permutation procedure. Using up to 100,000 permutations to estimate the statistical significance in 200 replicate data sets, we found that the method had 5.1% type I error at an α level of 0.05 and had various power to detect genes with simulated genetic associations. FLT1 and KDR had the most significant correlations with Q1 and were replicated 170 and 24 times, respectively, in 200 simulated data sets using a Bonferroni corrected p-value of 0.05 as a threshold. These results suggest that the distance correlation method can be used to identify genotype-phenotype association when multiple rare genetic variants in a gene are involved.
We implemented least absolute shrinkage and selection operator (LASSO) regression to evaluate gene effects in genome-wide association studies (GWAS) of brain images, using an MRI-derived temporal lobe volume measure from 729 subjects scanned as part of the Alzheimer’s Disease Neuroimaging Initiative (ADNI). Sparse groups of SNPs in individual genes were selected by LASSO, which identifies efficient sets of variants influencing the data. These SNPs were considered jointly when assessing their association with neuroimaging measures. We discovered 22 genes that passed genome-wide significance for influencing temporal lobe volume. This was a substantially greater number of significant genes compared to those found with standard, univariate GWAS. These top genes are all expressed in the brain and include genes previously related to brain function or neuropsychiatric disorders such as MACROD2, SORCS2, GRIN2B, MAGI2, NPAS3, CLSTN2, GABRG3, NRXN3, PRKAG2, GAS7, RBFOX1, ADARB2, CHD4, and CDH13. The top genes we identified with this method also displayed significant and widespread post hoc effects on voxelwise, tensor-based morphometry (TBM) maps of the temporal lobes. The most significantly associated gene was an autism susceptibility gene known as MACROD2. We were able to successfully replicate the effect of the MACROD2 gene in an independent cohort of 564 young, Australian healthy adult twins and siblings scanned with MRI (mean age: 23.8 ± 2.2 SD years). Our approach powerfully complements univariate techniques in detecting influences of genes on the living brain.
neuroimaging; MRI; imaging genetics; GWAS; LASSO; MACROD2
Genetic Analysis Workshop 17 used real sequence data from the 1000 Genomes Project and simulated phenotypes influenced by a large number of rare variants. Our aim is to evaluate the performance of various collapsing methods that were developed for analysis of multiple rare variants. We apply collapsing methods to continuous phenotypes Q1 and Q2 for all 200 replicates of the unrelated individuals data. Within each gene, we collapse (1) all SNPs, (2) all SNPs with minor allele frequency (MAF) < 0.05, and (3) nonsynonymous SNPs with MAF < 0.05. We consider two tests when collapsing variants: using the proportion of variants and using the presence/absence of any variant. We also compare our results to a single-marker analysis using PLINK. For phenotype Q1, the proportion test for collapsing rare nonsynonymous SNPs often performed the best. Two genes (FLT1 and KDR) had statistically significant results. A single-marker analysis using PLINK also provided statistically significant results for some SNPs within these two genes. For phenotype Q2, collapsing rare nonsynonymous SNPs performed the best, with almost no difference between proportion and presence tests. However, neither collapsing methods nor a single-marker analysis provided statistically significant results at the true genes for Q2. We also found that a large number of noncausal genes had high correlations with causal genes for Q1 and Q2, which may account for inflated false positives.
Currently there is a great deal of interest in developing methods for testing the role that rare variation plays in disease development. Here we propose a weighted association test that accumulates genetic variation across a signaling pathway. We evaluate our approach by analyzing simulated phenotype data from an exome sequencing study of 697 unrelated individuals from the Genetic Analysis Workshop 17 (GAW17) data set. Although our weighted approach identifies several interesting pathways associated with phenotype Q1, so does an alternative unweighted accumulation approach. Such a result is not unexpected because there is no systematic relationship between the allele frequency of a variant and its effect on phenotype in the GAW17 simulation model.
Motivation: In ordinary regression, imposition of a lasso penalty makes continuous model selection straightforward. Lasso penalized regression is particularly advantageous when the number of predictors far exceeds the number of observations.
Method: The present article evaluates the performance of lasso penalized logistic regression in case–control disease gene mapping with a large number of SNPs (single nucleotide polymorphisms) predictors. The strength of the lasso penalty can be tuned to select a predetermined number of the most relevant SNPs and other predictors. For a given value of the tuning constant, the penalized likelihood is quickly maximized by cyclic coordinate ascent. Once the most potent marginal predictors are identified, their two-way and higher order interactions can also be examined by lasso penalized logistic regression.
Results: This strategy is tested on both simulated and real data. Our findings on coeliac disease replicate the previous SNP results and shed light on possible interactions among the SNPs.
Availability: The software discussed is available in Mendel 9.0 at the UCLA Human Genetics web site.
Supplementary information: Supplementary data are available at Bioinformatics online.
We evaluate four association tests for rare variants—the combined multivariate and collapsing (CMC) method, two weighted-sum methods, and a variable threshold method—by applying them to the simulated data sets of unrelated individuals in the Genetic Analysis Workshop 17 (GAW17) data. The family-wise error rate (FWER) and average power are used as criteria for evaluation. Our results show that when all nonsynonymous SNPs (rare variants and common variants) in a gene are jointly analyzed, the CMC method fails to control the FWER; when only rare variants (single-nucleotide polymorphisms with minor allele frequency less than 0.05) are analyzed, all four methods can control FWER well. All four methods have comparable power, which is low for the analysis of the GAW17 data sets. Three of the methods (not including the CMC method) involve estimation of p-values using permutation procedures that either can be computationally intensive or generate inflated FWERs. We adapt a fast permutation procedure into these three methods. The results show that using the fast permutation procedure can produce FWERs and average powers close to the values obtained from the standard permutation procedure on the GAW17 data sets. The standard permutation procedure is computationally intensive.
There have been relatively few publications using linear regression models to predict a continuous response based on microarray expression profiles. Standard linear regression methods are problematic when the number of predictor variables exceeds the number of cases. We have evaluated three linear regression algorithms that can be used for the prediction of a continuous response based on high dimensional gene expression data. The three algorithms are the least angle regression (LAR), the least absolute shrinkage and selection operator (LASSO), and the averaged linear regression method (ALM). All methods are tested using simulations based on a real gene expression dataset and analyses of two sets of real gene expression data and using an unbiased complete cross validation approach. Our results show that the LASSO algorithm often provides a model with somewhat lower prediction error than the LAR method, but both of them perform more efficiently than the ALM predictor. We have developed a plug-in for BRB-ArrayTools that implements the LAR and the LASSO algorithms with complete cross-validation.
regression model; gene expression; continuous outcome
Existing methods for analyzing rare variant data focus on collapsing a group of rare variants into a single common variant; collapsing is based on an intuitive function of the rare variant genotype information, such as an indicator function or a weighted sum. It is more natural, however, to take into account the single-nucleotide polymorphism (SNP) interactions informed directly by the data. We propose a novel tree-based method that automatically detects SNP interactions and generates candidate markers from the original pool of rare variants. In addition, we utilize the advantage of having 200 phenotype replications in the Genetic Analysis Workshop 17 data to assess the candidate markers by means of repeated logistic regressions. This new approach shows potential in the rare variant analysis. We correctly identify the association between gene FLT1 and phenotype Affect, although there exist other false positives in our results. Our analyses are performed without knowledge of the underlying simulating model.
Rare variants are believed to play an important role in disease etiology. Recent advances in high-throughput sequencing technology enable investigators to systematically characterize the genetic effects of both common and rare variants. We introduce several approaches that simultaneously test the effects of common and rare variants within a single-nucleotide polymorphism (SNP) set based on logistic regression models and logistic kernel machine models. Gene-environment interactions and SNP-SNP interactions are also considered in some of these models. We illustrate the performance of these methods using the unrelated individuals data from Genetic Analysis Workshop 17. Three true disease genes (FLT1, PIK3C3, and KDR) were consistently selected using the proposed methods. In addition, compared to logistic regression models, the logistic kernel machine models were more powerful, presumably because they reduced the effective number of parameters through regularization. Our results also suggest that a screening step is effective in decreasing the number of false-positive findings, which is often a big concern for association studies.
We describe the use of the l1 norm for selection of a sparse set of model parameters that are used in
the prediction of viral drug response, based on genetic sequence data
of the Human Immunodeficiency Virus (HIV) reverse-transcriptase enzyme. We
discuss the use of the l1 norm in the Least Absolute Selection and Shrinkage Operator (LASSO) regression
model and the Support Vector Machine model. When tested by cross-validation
with laboratory measurements, these models predict viral
phenotype, or resistance, in response to Reverse-Transcriptase Inhibitors (RTIs) more
accurately than other known models. The l1 norm is the most selective convex function, which sets a large proportion
of the parameters to zero and also assures that a single optimal solution
will be found, given a particular model formulation and training
data set. A statistical model that reliably predicts viral drug response
is an important tool in the selection of Anti-Retroviral Therapy. These
techniques have general application to modeling phenotype from
complex genetic data.
Phenotype prediction; genetic data analysis; LASSO; support vector machines; l1 norm; sparse regression; Human Immunodeficiency Virus
The advance of high-throughput next-generation sequencing technology makes possible the analysis of rare variants. However, the investigation of rare variants in unrelated-individuals data sets faces the challenge of low power, and most methods circumvent the difficulty by using various collapsing procedures based on genes, pathways, or gene clusters. We suggest a new way to identify causal rare variants using the F-statistic and sliced inverse regression. The procedure is tested on the data set provided by the Genetic Analysis Workshop 17 (GAW17). After preliminary data reduction, we ranked markers according to their F-statistic values. Top-ranked markers were then subjected to sliced inverse regression, and those with higher absolute coefficients in the most significant sliced inverse regression direction were selected. The procedure yields good false discovery rates for the GAW17 data and thus is a promising method for future study on rare variants.