An empirical Bayes method for estimating epistatic effects of quantitative trait loci

Summary .   The genetic variance of a quantitative trait is often controlled by the segregation of multiple interacting loci. Linear model regression analysis is usually applied to estimating and testing effects of these quantitative trait loci (QTL). Including all the main effects and the effects of interaction (epistatic effects), the dimension of the linear model can be extremely high. Variable selection via stepwise regression or stochastic search variable selection (SSVS) is the common procedure for epistatic effect QTL analysis. These methods are computationally intensive, yet they may not be optimal. The LASSO (least absolute shrinkage and selection operator) method is computationally more efficient than the above methods. As a result, it has been widely used in regression analysis for large models. However, LASSO has never been applied to genetic mapping for epistatic QTL, where the number of model effects is typically many times larger than the sample size. In this study, we developed an empirical Bayes method (E-BAYES) to map epistatic QTL under the mixed model framework. We also tested the feasibility of using LASSO to estimate epistatic effects, examined the fully Bayesian SSVS, and reevaluated the penalized likelihood (PENAL) methods in mapping epistatic QTL. Simulation studies showed that all the above methods performed satisfactorily well. However, E-BAYES appears to outperform all other methods in terms of minimizing the mean-squared error (MSE) with relatively short computing time. Application of the new method to real data was demonstrated using a barley dataset.     

A penalized maximum likelihood method for estimating epistatic effects of QTL

Although epistasis is an important phenomenon in the genetics and evolution of complex traits, epistatic effects are hard to estimate. The main problem is due to the overparameterized epistatic genetic models. An epistatic genetic model should include potential pair-wise interaction effects of all loci. However, the model is saturated quickly as the number of loci increases. Therefore, a variable selection technique is usually considered to exclude those interactions with negligible effects. With such techniques, we may run a high risk of missing some important interaction effects by not fully exploring the extremely large parameter space of models. We develop a penalized maximum likelihood method. The method developed here adopts a penalty that depends on the values of the parameters. The penalized likelihood method allows spurious QTL effects to be shrunk towards zero, while QTL with large effects are estimated with virtually no shrinkage. A simulation study shows that the new method can handle a model with a number of effects 15 times larger than the sample size. Simulation studies also show that results of the penalized likelihood method are comparable to the Bayesian shrinkage analysis, but the computational speed of the penalized method is orders of magnitude faster.     

A Model Selection Approach for the Identification of Quantitative Trait Loci in Experimental Crosses, Allowing Epistasis

The identification of quantitative trait loci (QTL) and their interactions is a crucial step toward the discovery of genes responsible for variation in experimental crosses. The problem is best viewed as one of model selection, and the most important aspect of the problem is the comparison of models of different sizes. We present a penalized likelihood approach, with penalties on QTL and pairwise interactions chosen to control false positive rates. This extends the work of Broman and Speed to allow for pairwise interactions among QTL. A conservative version of our penalized LOD score provides strict control over the rate of extraneous QTL and interactions; a more liberal criterion is more lenient on interactions but seeks to maintain control over the rate of inclusion of false loci. The key advance is that one needs only to specify a target false positive rate rather than a prior on the number of QTL and interactions. We illustrate the use of our model selection criteria as exploratory tools; simulation studies demonstrate reasonable power to detect QTL. Our liberal criterion is comparable in power to two Bayesian approaches.     

多QTL定位的压缩估计方法

本文综述了多标记分析和多QTL定位的压缩估计方法。对于前者,Xu（Genetics,2003,163：789—801）首先提出了Bayesian压缩估计方法。其关键在于让每个效应有一个特定的方差参数,而该方差又服从一定的先验分布,以致能从资料中估计之。由此,能够同时估计大量分子标记基因座的遗传效应,即使大多数标记的效应是可忽略的。然而,对于上位性遗传模型,其运算时间还是过长。为此,笔者将上述思想嵌入极大似然法,提出了惩罚最大似然方法。模拟研究显示：该方法能处理变量个数大于样本容量10倍左右的线性遗传模型。对于后者,本文详细介绍了基于固定区间和可变区间的Bayesian压缩估计方法。固定区间方法可处理中等密度的分子标记资料;可变区间方法则可分析高密度分子标记资料,甚至是上位性遗传模型。对于上位性检测,已介绍的惩罚最大似然方法和可变区间Bayesian压缩估计方法可供利用。应当指出,压缩估计方法在今后的eQTL和QTN定位以及基因互作网络分析等研究中也是有应用价值的。     

Multiple-Interval Mapping for Quantitative Trait Loci With a Spike in the Trait Distribution

For phenotypic distributions where many individuals share a common value—such as survival time following a pathogenic infection—a spike occurs at that common value. This spike affects quantitative trait loci (QTL) mapping methodologies and causes standard approaches to perform suboptimally. In this article, we develop a multiple-interval mapping (MIM) procedure based on mixture generalized linear models (GLIMs). An extended Bayesian information criterion (EBIC) is used for model selection. To demonstrate its utility, this new approach is compared to single-QTL models that appropriately handle the phenotypic distribution. The method is applied to data from Listeria infection as well as data from simulation studies. Compared to the single-QTL model, the findings demonstrate that the MIM procedure greatly improves the efficiency in terms of positive selection rate and false discovery rate. The method developed has been implemented using functions in R and is freely available to download and use.     

Shrinkage Estimation Method for Mapping Multiple Quantitative Trait Loci

In this article, shrinkage estimation method for multiple-marker analysis and for mapping multiple quantitative trait loci (QTL) was reviewed. For multiple-marker analysis, Xu (Genetics, 2003, 163:789-801) developed a Bayesian shrinkage estimation (BSE) method. The key to the success of this method is to allow each marker effect have its own variance parameter, which in turn has its own prior distribution so that the variance can be estimated from the data. Under this hierarchical model, a large number of markers can be handled although most of them may have negligible effects. Under epistatic genetic model, however, the running time is very long. To overcome this problem, a novel method of incorporating the idea described above into maximum likelihood, known as penalized likelihood method, was proposed. A simulated study showed that this method can handle a model with multiple effects, which are ten times larger than the sample size. For multiple QTL analysis, two modified versions for the BSE method were introduced: one is the fixed-interval method and another is the variable-interval method. The former deals with markers with intermediate density, and the latter can handle markers with extremely high density as well as model with epistatic effects. For the detection of epistatic effects, penalized likelihood method and the variable-interval approach of the BSE method are available.     

Bayesian joint mapping of quantitative trait loci for Gaussian and categorical characters in line crosses

Methodology for joint mapping of quantitative trait loci (QTL) affecting continuous and binary characters in experimental crosses is presented. The procedure consists of a Bayesian Gaussian-threshold model implemented via Markov chain Monte Carlo, which bypasses bottlenecks due to high-dimensional integrals required in maximum likelihood approaches. The method handles multiple binary traits and multiple QTL. Modeling of ordered categorical traits is discussed as well. Features of the method are illustrated using simulated datasets representing a backcross design, and the data are analyzed using mixed-trait and single-trait models. The mixed-trait analysis provides greater detection power of a QTL than a single-trait analysis when the QTL affects two or more traits. The number of QTL inferred in the mixed-trait analysis does not pertain to a specific trait, but the roles of each QTL on specific traits can be assessed from estimates of its effects. The impacts of varying incidence level and sample size on the mixed-trait QTL mapping analysis are investigated as well.     

Power calculations for selective genotyping in QTL mapping in backcross mice

Selective genotyping concerns the genotyping of a portion of individuals chosen on the basis of their phenotypic values. Often individuals are selected for genotyping from the high and low extremes of the phenotypic distribution. This procedure yields savings in cost and time by decreasing the total number of individuals genotyped. Previous work by Darvasi et al. (1993) has shown that the power to detect a QTL by genotyping 40-50 % of a population is roughly equivalent to genotyping the entire sample. However, these power studies have not accounted for different strategies of analysing the data when phenotypes of individuals in the middle are excluded, nor have they investigated the genome-wide type I error rate under these different strategies or different selection percentages. Further, these simulation studies have not considered markers over the entire genome. In this paper, we present simulation studies of power for the maximum likelihood approach to QTL mapping by Lander & Botstein (1989) in the context of selective genotyping. We calculate the power of selectively genotyping the individuals from the middle of the phenotypic distribution when performing QTL mapping over the whole mouse genome.     

Mapping of epistatic quantitative trait loci in four-way crosses

Four-way crosses (4WC) involving four different inbred lines often appear in plant and animal commercial breeding programs. Direct mapping of quantitative trait loci (QTL) in these commercial populations is both economical and practical. However, the existing statistical methods for mapping QTL in a 4WC population are built on the single-QTL genetic model. This simple genetic model fails to take into account QTL interactions, which play an important role in the genetic architecture of complex traits. In this paper, therefore, we attempted to develop a statistical method to detect epistatic QTL in 4WC population. Conditional probabilities of QTL genotypes, computed by the multi-point single locus method, were used to sample the genotypes of all putative QTL in the entire genome. The sampled genotypes were used to construct the design matrix for QTL effects. All QTL effects, including main and epistatic effects, were simultaneously estimated by the penalized maximum likelihood method. The proposed method was confirmed by a series of Monte Carlo simulation studies and real data analysis of cotton. The new method will provide novel tools for the genetic dissection of complex traits, construction of QTL networks, and analysis of heterosis.     

Screening methods for linear errors-in-variables models in high dimensions

Microarray studies, in order to identify genes associated with an outcome of interest, usually produce noisy measurements for a large number of gene expression features from a small number of subjects. One common approach to analyzing such high-dimensional data is to use linear errors-in-variables (EIV) models; however, current methods for fitting such models are computationally expensive. In this paper, we present two efficient screening procedures, namely, corrected penalized marginal screening (PMSc) and corrected sure independence screening (SISc), to reduce the number of variables for final model building. Both screening procedures are based on fitting corrected marginal regression models relating the outcome to each contaminated covariate separately, which can be computed efficiently even with a large number of features. Under mild conditions, we show that these procedures achieve screening consistency and reduce the number of features substantially, even when the number of covariates grows exponentially with sample size. In addition, if the true covariates are weakly correlated, we show that PMSc can achieve full variable selection consistency. Through a simulation study and an analysis of gene expression data for bone mineral density of Norwegian women, we demonstrate that the two new screening procedures make estimation of linear EIV models computationally scalable in high-dimensional settings, and improve finite sample estimation and selection performance compared with estimators that do not employ a screening stage.     

Multiple interval mapping for quantitative trait loci.

A new statistical method for mapping quantitative trait loci (QTL), called multiple interval mapping (MIM), is presented. It uses multiple marker intervals simultaneously to fit multiple putative QTL directly in the model for mapping QTL. The MIM model is based on Cockerham's model for interpreting genetic parameters and the method of maximum likelihood for estimating genetic parameters. With the MIM approach, the precision and power of QTL mapping could be improved. Also, epistasis between QTL, genotypic values of individuals, and heritabilities of quantitative traits can be readily estimated and analyzed. Using the MIM model, a stepwise selection procedure with likelihood ratio test statistic as a criterion is proposed to identify QTL. This MIM method was applied to a mapping data set of radiata pine on three traits: brown cone number, tree diameter, and branch quality scores. Based on the MIM result, seven, six, and five QTL were detected for the three traits, respectively. The detected QTL individually contributed from approximately 1 to 27% of the total genetic variation. Significant epistasis between four pairs of QTL in two traits was detected, and the four pairs of QTL contributed approximately 10.38 and 14.14% of the total genetic variation. The asymptotic variances of QTL positions and effects were also provided to construct the confidence intervals. The estimated heritabilities were 0.5606, 0.5226, and 0. 3630 for the three traits, respectively. With the estimated QTL effects and positions, the best strategy of marker-assisted selection for trait improvement for a specific purpose and requirement can be explored. The MIM FORTRAN program is available on the worldwide web (http://www.stat.sinica.edu.tw/chkao/).     

A penalized likelihood method for mapping epistatic quantitative trait Loci with one-dimensional genome searches

Epistasis is a common and important phenomenon, as indicated by results from a number of recent experiments. Unfortunately, the discovery of epistatic quantitative trait loci (QTL) is difficult since one must search for multiple QTL simultaneously in two or more dimensions. Such a multidimensional search necessitates many statistical tests, and a high statistical threshold must be adopted to avoid false positives. Furthermore, the large number of (interaction) parameters in comparison with the number of observations results in a serious danger of overfitting and overinterpretation of the data. In this article we present a new statistical framework for mapping epistasis in inbred line crosses. It is based on reducing the high dimensionality of the problem in two ways. First, epistatic QTL are mapped in a one-dimensional genome scan for high interactions between QTL and the genetic background. Second, the dimension of the search is bounded by penalized likelihood methods. We use simulated backcross data to illustrate the new approach.     

Simultaneous fine mapping of multiple closely linked quantitative trait Loci using combined linkage disequilibrium and linkage with a general pedigree

Within a small region (e.g., <10 cM), there can be multiple quantitative trait loci (QTL) underlying phenotypes of a trait. Simultaneous fine mapping of closely linked QTL needs an efficient tool to remove confounded shade effects among QTL within such a small region. We propose a variance component method using combined linkage disequilibrium (LD) and linkage information and a reversible jump Markov chain Monte Carlo (MCMC) sampling for model selection. QTL identity-by-descent (IBD) coefficients between individuals are estimated by a hybrid MCMC combining the random walk and the meiosis Gibbs sampler. These coefficients are used in a mixed linear model and an empirical Bayesian procedure combines residual maximum likelihood (REML) to estimate QTL effects and a reversible jump MCMC that samples the number of QTL and the posterior QTL intensities across the tested region. Note that two MCMC processes are used, i.e., an (internal) MCMC for IBD estimation and an (external) MCMC for model selection. In a simulation study, the use of the multiple-QTL model clearly removes the shade effects between three closely linked QTL located at 1.125, 3.875, and 7.875 cM across the region of 10 cM, using 40 markers at 0.25-cM intervals. It is shown that the use of combined LD and linkage information gives much more useful information compared to using linkage information alone for both single- and multiple-QTL analyses. When using a lower marker density (11 markers at 1-cM intervals), the signal of the second QTL can disappear. Extreme values of past effective size (resulting in extreme levels of LD) decrease the mapping accuracy.     

Biased estimators of quantitative trait locus heritability and location in interval mapping

In many empirical studies, it has been observed that genome scans yield biased estimates of heritability, as well as genetic effects. It is widely accepted that quantitative trait locus (QTL) mapping is a model selection procedure, and that the overestimation of genetic effects is the result of using the same data for model selection as estimation of parameters. There are two key steps in QTL modeling, each of which biases the estimation of genetic effects. First, test procedures are employed to select the regions of the genome for which there is significant evidence for the presence of QTL. Second, and most important for this demonstration, estimates of the genetic effects are reported only at the locations for which the evidence is maximal. We demonstrate that even when we know there is just one QTL present (ignoring the testing bias), and we use interval mapping to estimate its location and effect, the estimator of the effect will be biased. As evidence, we present results of simulations investigating the relative importance of the two sources of bias and the dependence of bias of heritability estimators on the true QTL heritability, sample size, and the length of the investigated part of the genome. Moreover, we present results of simulations demonstrating the skewness of the distribution of estimators of QTL locations and the resulting bias in estimation of location. We use computer simulations to investigate the dependence of this bias on the true QTL location, heritability, and the sample size.     

The analysis of QTL by simultaneous use of the full linkage map

An extension of interval mapping is presented that incorporates all intervals on the linkage map simultaneously. The approach uses a working model in which the sizes of putative QTL for all intervals across the genome are random effects. An outlier detection method is used to screen for possible QTL. Selected QTL are subsequently fitted as fixed effects. This screening and selection approach is repeated until the variance component for QTL sizes is not statistically significant. A comprehensive simulation study is conducted in which map uncertainty is included. The proposed method is shown to be superior to composite interval mapping in terms of power of detection of QTL. There is an increase in the rate of false positive QTL detected when using the new approach, but this rate decreases as the population size increases. The new approach is much simpler computationally. The analysis of flour milling yield in a doubled haploid population illustrates the improved power of detection of QTL using the approach, and also shows how vital it is to allow for sources of non-genetic variation in the analysis.     

Improving QTL Mapping Resolution Based on Genotypic Sampling—a Case Using a RIL Population

The QTL mapping results were compared with the genotypically selected and random samples of the same size on the base of a RIL population. The results demonstrated that there were no obvious differences in the trait distribution and marker segregation distortion between the genotypically selected and random samples with the same population size. However, a significant increase in QTL detection power, sensitivity, specificity, and QTL resolution in the genotypically selected samples were observed. Moreover, the highly significant effect was detected in small size of genotypically selected samples. In QTL mapping, phenotyping is a more sensitive limiting factor than genotyping so that the selection of samples could be an attractive strategy for increasing genome-wide QTL mapping resolution. The efficient selection of samples should be more helpful for QTL maker assistant selection, fine mapping, and QTL cloning.     

基于基因型选择提高QTL作图的精度——以一个RIL群体为例

以PCR为基础的分子标记以及其他检测技术的发展,使得大规模的标记分析成为现实。这也为通过大群体标记分析,然后基于基因型选择挑选合适的小群体,从而提高QTL定位准确性和精度提供了可能。以一个包含294个家系的重组自交系（RIL）群体为例,通过基因型选择和随机选择的办法产生了一系列大小不等的亚群体,比较了两类群体QTL定位的结果。分析表明：相同大小的基因型选择群体与随机群体相比性状的表型分布都符合正态分布;标记的偏分离情况也没有明显的差别,都随着群体大小的增大,偏分离的比例也逐渐增大。但同等大小的基因型选择群体比随机群体的交换富集率（CE）要大,且随着选择强度的增大不断增大,如群体大小为270时,CE=1．04,群体大小为30时,CE=1．45。总体上,随着群体大小的增加,不管是随机群体还是选择群体,其QTL检测能力、灵敏性和特异性也随之增加,但选择群体的检测能力、灵敏性和特异性总体上要好于随机群体。当群体大于或等于240时,其QTL检测能力基本没有差别;群体大小大于或等于210时,其QTL检测的灵敏性和特异性也没有什么差别。这也说明：选择强度越大,效果越明显。以QTLI—LOD区间作为衡量QTL精度的一个指标,结果显示所有基因型选择群体都比相同大小随机群体的QTL定位精度高。目前QTL定位研究中,基因型数据较表型数据而言更容易准确获得,因此通过基因型选择可以更好的优化群体结构,减少田间实验的工作量,提高全基因组水平QTL作图的精度,为随后的QTL辅助选择和精细定位以及克隆提供帮助。     

A Simple Regression-Based Method to Map Quantitative Trait Loci Underlying Function-Valued Phenotypes

Most statistical methods for quantitative trait loci (QTL) mapping focus on a single phenotype. However, multiple phenotypes are commonly measured, and recent technological advances have greatly simplified the automated acquisition of numerous phenotypes, including function-valued phenotypes, such as growth measured over time. While methods exist for QTL mapping with function-valued phenotypes, they are generally computationally intensive and focus on single-QTL models. We propose two simple, fast methods that maintain high power and precision and are amenable to extensions with multiple-QTL models using a penalized likelihood approach. After identifying multiple QTL by these approaches, we can view the function-valued QTL effects to provide a deeper understanding of the underlying processes. Our methods have been implemented as a package for R, funqtl.     

Look before you leap: a new approach to mapping QTL

In this paper, we present an innovative and powerful approach for mapping quantitative trait loci (QTL) in experimental populations. This deviates from the traditional approach of (composite) interval mapping which uses a QTL profile to simultaneously determine the number and location of QTL. Instead, we look before we leap by employing separate detection and localization stages. In the detection stage, we use an iterative variable selection process coupled with permutation to identify the number and synteny of QTL. In the localization stage, we position the detected QTL through a series of one-dimensional interval mapping scans. Results from a detailed simulation study and real analysis of wheat data are presented. We achieve impressive increases in the power of QTL detection compared to composite interval mapping. We also accurately estimate the size and position of QTL. An R library, DLMap, implements the methods described here and is freely available from CRAN ().     

Improving quantitative trait loci mapping resolution in experimental crosses by the use of genotypically selected samples

One of the key factors contributing to the success of a quantitative trait locus (QTL) mapping experiment is the precision with which QTL positions can be estimated. We show, using simulations, that QTL mapping precision for an experimental cross can be increased by the use of a genotypically selected sample of individuals rather than an unselected sample of the same size. Selection is performed using a previously described method that optimizes the complementarity of the crossover sites within the sample. Although the increase in precision is accompanied by a decrease in QTL detection power at markers distant from QTL, only a modest increase in marker density is needed to obtain equivalent power over the whole map. Selected samples also show a slight reduction in the number of false-positive QTL. We find that two features of selected samples independently contribute to these effects: an increase in the number of crossover sites and increased evenness in crossover spacing. We provide an empirical formula for crossover enrichment in selected samples that is useful in experimental design and data analysis. For QTL studies in which the phenotyping is more of a limiting factor than the generation of individuals and the scoring of genotypes, selective sampling is an attractive strategy for increasing genome-wide QTL map resolution.