Accuracy of genomic selection methods in a standard data set of loblolly pine (Pinus taeda L.)

Genomic selection can increase genetic gain per generation through early selection. Genomic selection is expected to be particularly valuable for traits that are costly to phenotype and expressed late in the life cycle of long-lived species. Alternative approaches to genomic selection prediction models may perform differently for traits with distinct genetic properties. Here the performance of four different original methods of genomic selection that differ with respect to assumptions regarding distribution of marker effects, including (i) ridge regression-best linear unbiased prediction (RR-BLUP), (ii) Bayes A, (iii) Bayes Cπ, and (iv) Bayesian LASSO are presented. In addition, a modified RR-BLUP (RR-BLUP B) that utilizes a selected subset of markers was evaluated. The accuracy of these methods was compared across 17 traits with distinct heritabilities and genetic architectures, including growth, development, and disease-resistance properties, measured in a Pinus taeda (loblolly pine) training population of 951 individuals genotyped with 4853 SNPs. The predictive ability of the methods was evaluated using a 10-fold, cross-validation approach, and differed only marginally for most method/trait combinations. Interestingly, for fusiform rust disease-resistance traits, Bayes Cπ, Bayes A, and RR-BLUB B had higher predictive ability than RR-BLUP and Bayesian LASSO. Fusiform rust is controlled by few genes of large effect. A limitation of RR-BLUP is the assumption of equal contribution of all markers to the observed variation. However, RR-BLUP B performed equally well as the Bayesian approaches.The genotypic and phenotypic data used in this study are publically available for comparative analysis of genomic selection prediction models.     

Using whole-genome sequence data to predict quantitative trait phenotypes in Drosophila melanogaster

Predicting organismal phenotypes from genotype data is important for plant and animal breeding, medicine, and evolutionary biology. Genomic-based phenotype prediction has been applied for single-nucleotide polymorphism (SNP) genotyping platforms, but not using complete genome sequences. Here, we report genomic prediction for starvation stress resistance and startle response in Drosophila melanogaster, using ～2.5 million SNPs determined by sequencing the Drosophila Genetic Reference Panel population of inbred lines. We constructed a genomic relationship matrix from the SNP data and used it in a genomic best linear unbiased prediction (GBLUP) model. We assessed predictive ability as the correlation between predicted genetic values and observed phenotypes by cross-validation, and found a predictive ability of 0.239±0.008 (0.230±0.012) for starvation resistance (startle response). The predictive ability of BayesB, a Bayesian method with internal SNP selection, was not greater than GBLUP. Selection of the 5% SNPs with either the highest absolute effect or variance explained did not improve predictive ability. Predictive ability decreased only when fewer than 150,000 SNPs were used to construct the genomic relationship matrix. We hypothesize that predictive power in this population stems from the SNP-based modeling of the subtle relationship structure caused by long-range linkage disequilibrium and not from population structure or SNPs in linkage disequilibrium with causal variants. We discuss the implications of these results for genomic prediction in other organisms.     

Prediction of genetic values of quantitative traits with epistatic effects in plant breeding populations

Though epistasis has long been postulated to have a critical role in genetic regulation of important pathways as well as provide a major source of variation in the process of speciation, the importance of epistasis for genomic selection in the context of plant breeding is still being debated. In this paper, we report the results on the prediction of genetic values with epistatic effects for 280 accessions in the Nebraska Wheat Breeding Program using adaptive mixed least absolute shrinkage and selection operator (LASSO). The development of adaptive mixed LASSO, originally designed for association mapping, for the context of genomic selection is reported. The results show that adaptive mixed LASSO can be successfully applied to the prediction of genetic values while incorporating both marker main effects and epistatic effects. Especially, the prediction accuracy is substantially improved by the inclusion of two-locus epistatic effects (more than onefold in some cases as measured by cross-validation correlation coefficient), which is observed for multiple traits and planting locations. This points to significant potential in using non-additive genetic effects for genomic selection in crop breeding practices.     

Including dominance effects in the prediction model through locus-specific weights on heterozygous genotypes can greatly improve genomic predictive abilities

The dominance effect is considered to be a key factor affecting complex traits. However, previous studies have shown that the improvement of the model, including the dominance effect, is usually less than 1%. This study proposes a novel genomic prediction method called CADM, which combines additive and dominance genetic effects through locus-specific weights on heterozygous genotypes. To the best of our knowledge, this is the first study of weighting dominance effects for genomic prediction. This method was applied to the analysis of chicken (511 birds) and pig (3534 animals) datasets. A 5-fold cross-validation method was used to evaluate the genomic predictive ability. The CADM model was compared with typical models considering additive and dominance genetic effects (ADM) and the model considering only additive genetic effects (AM). Based on the chicken data, using the CADM model, the genomic predictive abilities were improved for all three traits (body weight at 12th week, eviscerating percentage, and breast muscle percentage), and the average improvement in prediction accuracy was 27.1% compared with the AM model, while the ADM model was not better than the AM model. Based on the pig data, the CADM model increased the genomic predictive ability for all the three pig traits (trait names are masked, here designated as T1, T2, and T3), with an average increase of 26.3%, and the ADM model did not improve, or even slightly decreased, compared with the AM model. The results indicate that dominant genetic variation is one of the important sources of phenotypic variation, and the novel prediction model significantly improves the accuracy of genomic prediction.Subject terms: Animal breeding, Quantitative trait     

Causation and causal inference for genetic effects

Over the past three decades, substantial developments have been made on how to infer the causal effect of an exposure on an outcome, using data from observational studies, with the randomized experiment as the golden standard. These developments have reshaped the paradigm of how to build statistical models, how to adjust for confounding, how to assess direct effects, mediated effects and interactions, and even how to analyze data from randomized experiments. The congruence of random transmission of alleles during meiosis and the randomization in controlled experiments/trials, suggests that genetic studies may lend themselves naturally to a causal analysis. In this contribution, we will reflect on this and motivate, through illustrative examples, where insights from the causal inference literature may help to understand and correct for typical biases in genetic effect estimates.     

The Effects of Demography and Long-Term Selection on the Accuracy of Genomic Prediction with Sequence Data

The use of dense SNPs to predict the genetic value of an individual for a complex trait is often referred to as “genomic selection” in livestock and crops, but is also relevant to human genetics to predict, for example, complex genetic disease risk. The accuracy of prediction depends on the strength of linkage disequilibrium (LD) between SNPs and causal mutations. If sequence data were used instead of dense SNPs, accuracy should increase because causal mutations are present, but demographic history and long-term negative selection also influence accuracy. We therefore evaluated genomic prediction, using simulated sequence in two contrasting populations: one reducing from an ancestrally large effective population size (N_e) to a small one, with high LD common in domestic livestock, while the second had a large constant-sized N_e with low LD similar to that in some human or outbred plant populations. There were two scenarios in each population; causal variants were either neutral or under long-term negative selection. For large N_e, sequence data led to a 22% increase in accuracy relative to ∼600K SNP chip data with a Bayesian analysis and a more modest advantage with a BLUP analysis. This advantage increased when causal variants were influenced by negative selection, and accuracy persisted when 10 generations separated reference and validation populations. However, in the reducing N_e population, there was little advantage for sequence even with negative selection. This study demonstrates the joint influence of demography and selection on accuracy of prediction and improves our understanding of how best to exploit sequence for genomic prediction.     

Ultra-high dimensional variable selection for doubly robust causal inference

Causal inference has been increasingly reliant on observational studies with rich covariate information. To build tractable causal procedures, such as the doubly robust estimators, it is imperative to first extract important features from high or even ultra-high dimensional data. In this paper, we propose causal ball screening for confounder selection from modern ultra-high dimensional data sets. Unlike the familiar task of variable selection for prediction modeling, our confounder selection procedure aims to control for confounding while improving efficiency in the resulting causal effect estimate. Previous empirical and theoretical studies suggest excluding causes of the treatment that are not confounders. Motivated by these results, our goal is to keep all the predictors of the outcome in both the propensity score and outcome regression models. A distinctive feature of our proposal is that we use an outcome model-free procedure for propensity score model selection, thereby maintaining double robustness in the resulting causal effect estimator. Our theoretical analyses show that the proposed procedure enjoys a number of properties, including model selection consistency and pointwise normality. Synthetic and real data analysis show that our proposal performs favorably with existing methods in a range of realistic settings. Data used in preparation of this paper were obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database.     

Enhancing Genome-Enabled Prediction by Bagging Genomic BLUP

We examined whether or not the predictive ability of genomic best linear unbiased prediction (GBLUP) could be improved via a resampling method used in machine learning: bootstrap aggregating sampling (“bagging”). In theory, bagging can be useful when the predictor has large variance or when the number of markers is much larger than sample size, preventing effective regularization. After presenting a brief review of GBLUP, bagging was adapted to the context of GBLUP, both at the level of the genetic signal and of marker effects. The performance of bagging was evaluated with four simulated case studies including known or unknown quantitative trait loci, and an application was made to real data on grain yield in wheat planted in four environments. A metric aimed to quantify candidate-specific cross-validation uncertainty was proposed and assessed; as expected, model derived theoretical reliabilities bore no relationship with cross-validation accuracy. It was found that bagging can ameliorate predictive performance of GBLUP and make it more robust against over-fitting. Seemingly, 25–50 bootstrap samples was enough to attain reasonable predictions as well as stable measures of individual predictive mean squared errors.     

What Can Causal Networks Tell Us about Metabolic Pathways?

Graphical models describe the linear correlation structure of data and have been used to establish causal relationships among phenotypes in genetic mapping populations. Data are typically collected at a single point in time. Biological processes on the other hand are often non-linear and display time varying dynamics. The extent to which graphical models can recapitulate the architecture of an underlying biological processes is not well understood. We consider metabolic networks with known stoichiometry to address the fundamental question: “What can causal networks tell us about metabolic pathways?”. Using data from an Arabidopsis BaySha population and simulated data from dynamic models of pathway motifs, we assess our ability to reconstruct metabolic pathways using graphical models. Our results highlight the necessity of non-genetic residual biological variation for reliable inference. Recovery of the ordering within a pathway is possible, but should not be expected. Causal inference is sensitive to subtle patterns in the correlation structure that may be driven by a variety of factors, which may not emphasize the substrate-product relationship. We illustrate the effects of metabolic pathway architecture, epistasis and stochastic variation on correlation structure and graphical model-derived networks. We conclude that graphical models should be interpreted cautiously, especially if the implied causal relationships are to be used in the design of intervention strategies.     

Genomic prediction ability for yield-related traits in German winter barley elite material

Key message

Genomic prediction was evaluated in German winter barley breeding lines. In this material, prediction ability is strongly influenced by population structure and main determinant of prediction ability is the close genetic relatedness of the breeding material.

Abstract

To ensure breeding progress under changing environmental conditions the implementation and evaluation of new breeding methods is of crucial importance. Modern breeding approaches like genomic selection may significantly accelerate breeding progress. We assessed the potential of genomic prediction in a training population of 750 genotypes, consisting of multiple six-rowed winter barley (Hordeum vulgare L.) elite material families and old cultivars, which reflect the breeding history of barley in Germany. Crosses of parents selected from the training set were used to create a set of double-haploid families consisting of 750 genotypes. Those were used to confirm prediction ability estimates based on a cross-validation with the training set material using 11 different genomic prediction models. Population structure was inferred with dimensionality reduction methods like discriminant analysis of principle components and the influence of population structure on prediction ability was investigated. In addition to the size of the training set, marker density is of crucial importance for genomic prediction. We used genome-wide linkage disequilibrium and persistence of linkage phase as indicators to estimate that 11,203 evenly spaced markers are required to capture all QTL effects. Although a 9k SNP array does not contain a sufficient number of polymorphic markers for long-term genomic selection, we obtained fairly high prediction accuracies ranging from 0.31 to 0.71 for the traits earing, hectoliter weight, spikes per square meter, thousand kernel weight and yield and show that they result from the close genetic relatedness of the material. Our work contributes to designing long-term genetic prediction programs for barley breeding.

     

Using Genetic Distance to Infer the Accuracy of Genomic Prediction

The prediction of phenotypic traits using high-density genomic data has many applications such as the selection of plants and animals of commercial interest; and it is expected to play an increasing role in medical diagnostics. Statistical models used for this task are usually tested using cross-validation, which implicitly assumes that new individuals (whose phenotypes we would like to predict) originate from the same population the genomic prediction model is trained on. In this paper we propose an approach based on clustering and resampling to investigate the effect of increasing genetic distance between training and target populations when predicting quantitative traits. This is important for plant and animal genetics, where genomic selection programs rely on the precision of predictions in future rounds of breeding. Therefore, estimating how quickly predictive accuracy decays is important in deciding which training population to use and how often the model has to be recalibrated. We find that the correlation between true and predicted values decays approximately linearly with respect to either F_ST or mean kinship between the training and the target populations. We illustrate this relationship using simulations and a collection of data sets from mice, wheat and human genetics.     

The Effect of Linkage on Establishment and Survival of Locally Beneficial Mutations

We study invasion and survival of weakly beneficial mutations arising in linkage to an established migration–selection polymorphism. Our focus is on a continent–island model of migration, with selection at two biallelic loci for adaptation to the island environment. Combining branching and diffusion processes, we provide the theoretical basis for understanding the evolution of islands of divergence, the genetic architecture of locally adaptive traits, and the importance of so-called “divergence hitchhiking” relative to other mechanisms, such as “genomic hitchhiking”, chromosomal inversions, or translocations. We derive approximations to the invasion probability and the extinction time of a de novo mutation. Interestingly, the invasion probability is maximized at a nonzero recombination rate if the focal mutation is sufficiently beneficial. If a proportion of migrants carries a beneficial background allele, the mutation is less likely to become established. Linked selection may increase the survival time by several orders of magnitude. By altering the timescale of stochastic loss, it can therefore affect the dynamics at the focal site to an extent that is of evolutionary importance, especially in small populations. We derive an effective migration rate experienced by the weakly beneficial mutation, which accounts for the reduction in gene flow imposed by linked selection. Using the concept of the effective migration rate, we also quantify the long-term effects on neutral variation embedded in a genome with arbitrarily many sites under selection. Patterns of neutral diversity change qualitatively and quantitatively as the position of the neutral locus is moved along the chromosome. This will be useful for population-genomic inference. Our results strengthen the emerging view that physically linked selection is biologically relevant if linkage is tight or if selection at the background locus is strong.     

Joint-multiple family linkage analysis predicts within-family variation better than single-family analysis of the maize nested association mapping population

Quantitative trait locus (QTL) mapping has been used to dissect the genetic architecture of complex traits and predict phenotypes for marker-assisted selection. Many QTL mapping studies in plants have been limited to one biparental family population. Joint analysis of multiple biparental families offers an alternative approach to QTL mapping with a wider scope of inference. Joint-multiple population analysis should have higher power to detect QTL shared among multiple families, but may have lower power to detect rare QTL. We compared prediction ability of single-family and joint-family QTL analysis methods with fivefold cross-validation for 6 diverse traits using the maize nested association mapping population, which comprises 25 biparental recombinant inbred families. Joint-family QTL analysis had higher mean prediction abilities than single-family QTL analysis for all traits at most significance thresholds, and was always better at more stringent significance thresholds. Most robust QTL (detected in >50% of data samples) were restricted to one family and were often not detected at high frequency by joint-family analysis, implying substantial genetic heterogeneity among families for complex traits in maize. The superior predictive ability of joint-family QTL models despite important genetic differences among families suggests that joint-family models capture sufficient smaller effect QTL that are shared across families to compensate for missing some rare large-effect QTL.     

Genomic models with genotype × environment interaction for predicting hybrid performance: an application in maize hybrids

Key message

A new genomic model that incorporates genotype?×?environment interaction gave increased prediction accuracy of untested hybrid response for traits such as percent starch content, percent dry matter content and silage yield of maize hybrids.

Abstract

The prediction of hybrid performance (HP) is very important in agricultural breeding programs. In plant breeding, multi-environment trials play an important role in the selection of important traits, such as stability across environments, grain yield and pest resistance. Environmental conditions modulate gene expression causing genotype?×?environment interaction (G?×?E), such that the estimated genetic correlations of the performance of individual lines across environments summarize the joint action of genes and environmental conditions. This article proposes a genomic statistical model that incorporates G?×?E for general and specific combining ability for predicting the performance of hybrids in environments. The proposed model can also be applied to any other hybrid species with distinct parental pools. In this study, we evaluated the predictive ability of two HP prediction models using a cross-validation approach applied in extensive maize hybrid data, comprising 2724 hybrids derived from 507 dent lines and 24 flint lines, which were evaluated for three traits in 58 environments over 12 years; analyses were performed for each year. On average, genomic models that include the interaction of general and specific combining ability with environments have greater predictive ability than genomic models without interaction with environments (ranging from 12 to 22%, depending on the trait). We concluded that including G?×?E in the prediction of untested maize hybrids increases the accuracy of genomic models.

     

Bridging the gap between marker-assisted and genomic selection of heading time and plant height in hybrid wheat

Based on data from field trials with a large collection of 135 elite winter wheat inbred lines and 1604 F₁ hybrids derived from them, we compared the accuracy of prediction of marker-assisted selection and current genomic selection approaches for the model traits heading time and plant height in a cross-validation approach. For heading time, the high accuracy seen with marker-assisted selection severely dropped with genomic selection approaches RR-BLUP (ridge regression best linear unbiased prediction) and BayesCπ, whereas for plant height, accuracy was low with marker-assisted selection as well as RR-BLUP and BayesCπ. Differences in the linkage disequilibrium structure of the functional and single-nucleotide polymorphism markers relevant for the two traits were identified in a simulation study as a likely explanation for the different trends in accuracies of prediction. A new genomic selection approach, weighted best linear unbiased prediction (W-BLUP), designed to treat the effects of known functional markers more appropriately, proved to increase the accuracy of prediction for both traits and thus closes the gap between marker-assisted and genomic selection.     

Selection of haplotype variables from a high-density marker map for genomic prediction

Background

Using haplotype blocks as predictors rather than individual single nucleotide polymorphisms (SNPs) may improve genomic predictions, since haplotypes are in stronger linkage disequilibrium with the quantitative trait loci than are individual SNPs. It has also been hypothesized that an appropriate selection of a subset of haplotype blocks can result in similar or better predictive ability than when using the whole set of haplotype blocks. This study investigated genomic prediction using a set of haplotype blocks that contained the SNPs with large effects estimated from an individual SNP prediction model. We analyzed protein yield, fertility and mastitis of Nordic Holstein cattle, and used high-density markers (about 770k SNPs). To reach an optimum number of haplotype variables for genomic prediction, predictions were performed using subsets of haplotype blocks that contained a range of 1000 to 50 000 main SNPs.

Results

The use of haplotype blocks improved the prediction reliabilities, even when selection focused on only a group of haplotype blocks. In this case, the use of haplotype blocks that contained the 20 000 to 50 000 SNPs with the highest effect was sufficient to outperform the model that used all individual SNPs as predictors (up to 1.3 % improvement in prediction reliability for mastitis, compared to individual SNP approach), and the achieved reliabilities were similar to those using all haplotype blocks available in the genome data (from 0.6 % lower to 0.8 % higher reliability).

Conclusions

Haplotype blocks used as predictors can improve the reliability of genomic prediction compared to the individual SNP model. Furthermore, the use of a subset of haplotype blocks that contains the main SNP effects from genomic data could be a feasible approach to genomic prediction in dairy cattle, given an increase in density of genotype data available. The predictive ability of the models that use a subset of haplotype blocks was similar to that obtained using either all haplotype blocks or all individual SNPs, with the benefit of having a much lower computational demand.     

A bayesian mixed regression based prediction of quantitative traits from molecular marker and gene expression data

Both molecular marker and gene expression data were considered alone as well as jointly to serve as additive predictors for two pathogen-activity-phenotypes in real recombinant inbred lines of soybean. For unobserved phenotype prediction, we used a bayesian hierarchical regression modeling, where the number of possible predictors in the model was controlled by different selection strategies tested. Our initial findings were submitted for DREAM5 (the 5th Dialogue on Reverse Engineering Assessment and Methods challenge) and were judged to be the best in sub-challenge B3 wherein both functional genomic and genetic data were used to predict the phenotypes. In this work we further improve upon this previous work by considering various predictor selection strategies and cross-validation was used to measure accuracy of in-data and out-data predictions. The results from various model choices indicate that for this data use of both data types (namely functional genomic and genetic) simultaneously improves out-data prediction accuracy. Adequate goodness-of-fit can be easily achieved with more complex models for both phenotypes, since the number of potential predictors is large and the sample size is not small. We also further studied gene-set enrichment (for continuous phenotype) in the biological process in question and chromosomal enrichment of the gene set. The methodological contribution of this paper is in exploration of variable selection techniques to alleviate the problem of over-fitting. Different strategies based on the nature of covariates were explored and all methods were implemented under the bayesian hierarchical modeling framework with indicator-based covariate selection. All the models based in careful variable selection procedure were found to produce significant results based on permutation test.     

Spending degrees of freedom in a poor economy: A case study of building a sightability model for moose in northeastern Minnesota

Sightability models are binary logistic-regression models used to estimate and adjust for visibility bias in wildlife-population surveys. Like many models in wildlife and ecology, sightability models are typically developed from small observational datasets with many candidate predictors. Aggressive model-selection methods are often employed to choose a best model for prediction and effect estimation, despite evidence that such methods can lead to overfitting (i.e., selected models may describe random error or noise rather than true predictor–response curves) and poor predictive ability. We used moose (Alces alces) sightability data from northeastern Minnesota (2005–2007) as a case study to illustrate an alternative approach, which we refer to as degrees-of-freedom (df) spending: sample-size guidelines are used to determine an acceptable level of model complexity and then a pre-specified model is fit to the data and used for inference. For comparison, we also constructed sightability models using Akaike's Information Criterion (AIC) step-down procedures and model averaging (based on a small set of models developed using df-spending guidelines). We used bootstrap procedures to mimic the process of model fitting and prediction, and to compute an index of overfitting, expected predictive accuracy, and model-selection uncertainty. The index of overfitting increased 13% when the number of candidate predictors was increased from three to eight and a best model was selected using step-down procedures. Likewise, model-selection uncertainty increased when the number of candidate predictors increased. Model averaging (based on R = 30 models with 1–3 predictors) effectively shrunk regression coefficients toward zero and produced similar estimates of precision to our 3-df pre-specified model. As such, model averaging may help to guard against overfitting when too many predictors are considered (relative to available sample size). The set of candidate models will influence the extent to which coefficients are shrunk toward zero, which has implications for how one might apply model averaging to problems traditionally approached using variable-selection methods. We often recommend the df-spending approach in our consulting work because it is easy to implement and it naturally forces investigators to think carefully about their models and predictors. Nonetheless, similar concepts should apply whether one is fitting 1 model or using multi-model inference. For example, model-building decisions should consider the effective sample size, and potential predictors should be screened (without looking at their relationship to the response) for missing data, narrow distributions, collinearity, potentially overly influential observations, and measurement errors (e.g., via logical error checks). © 2011 The Wildlife Society.     

The importance of phenotypic data analysis for genomic prediction - a case study comparing different spatial models in rye

Background

Genomic prediction is becoming a daily tool for plant breeders. It makes use of genotypic information to make predictions used for selection decisions. The accuracy of the predictions depends on the number of genotypes used in the calibration; hence, there is a need of combining data across years. A proper phenotypic analysis is a crucial prerequisite for accurate calibration of genomic prediction procedures. We compared stage-wise approaches to analyse a real dataset of a multi-environment trial (MET) in rye, which was connected between years only through one check, and used different spatial models to obtain better estimates, and thus, improved predictive abilities for genomic prediction. The aims of this study were to assess the advantage of using spatial models for the predictive abilities of genomic prediction, to identify suitable procedures to analyse a MET weakly connected across years using different stage-wise approaches, and to explore genomic prediction as a tool for selection of models for phenotypic data analysis.

Results

Using complex spatial models did not significantly improve the predictive ability of genomic prediction, but using row and column effects yielded the highest predictive abilities of all models. In the case of MET poorly connected between years, analysing each year separately and fitting year as a fixed effect in the genomic prediction stage yielded the most realistic predictive abilities. Predictive abilities can also be used to select models for phenotypic data analysis. The trend of the predictive abilities was not the same as the traditionally used Akaike information criterion, but favoured in the end the same models.

Conclusions

Making predictions using weakly linked datasets is of utmost interest for plant breeders. We provide an example with suggestions on how to handle such cases. Rather than relying on checks we show how to use year means across all entries for integrating data across years. It is further shown that fitting of row and column effects captures most of the heterogeneity in the field trials analysed.

Electronic supplementary material

The online version of this article (doi:10.1186/1471-2164-15-646) contains supplementary material, which is available to authorized users.     

Accounting for Genetic Architecture Improves Sequence Based Genomic Prediction for a Drosophila Fitness Trait

The ability to predict quantitative trait phenotypes from molecular polymorphism data will revolutionize evolutionary biology, medicine and human biology, and animal and plant breeding. Efforts to map quantitative trait loci have yielded novel insights into the biology of quantitative traits, but the combination of individually significant quantitative trait loci typically has low predictive ability. Utilizing all segregating variants can give good predictive ability in plant and animal breeding populations, but gives little insight into trait biology. Here, we used the Drosophila Genetic Reference Panel to perform both a genome wide association analysis and genomic prediction for the fitness-related trait chill coma recovery time. We found substantial total genetic variation for chill coma recovery time, with a genetic architecture that differs between males and females, a small number of molecular variants with large main effects, and evidence for epistasis. Although the top additive variants explained 36% (17%) of the genetic variance among lines in females (males), the predictive ability using genomic best linear unbiased prediction and a relationship matrix using all common segregating variants was very low for females and zero for males. We hypothesized that the low predictive ability was due to the mismatch between the infinitesimal genetic architecture assumed by the genomic best linear unbiased prediction model and the true genetic architecture of chill coma recovery time. Indeed, we found that the predictive ability of the genomic best linear unbiased prediction model is markedly improved when we combine quantitative trait locus mapping with genomic prediction by only including the top variants associated with main and epistatic effects in the relationship matrix. This trait-associated prediction approach has the advantage that it yields biologically interpretable prediction models.