Genomic Restricted Maximum Likelihood Research Articles

SNP-Based Heritability of Osteochondrosis Dissecans in Hanoverian Warmblood Horses.

Before the genomics era, heritability estimates were performed using pedigree data. Data collection for pedigree analysis is time consuming and holds the risk of incorrect or incomplete data. With the availability of SNP-based arrays, heritability can now be estimated based on genotyping data. We used SNP array and 1.6 million imputed genotype data with different minor allele frequency restrictions to estimate heritabilities for osteochondrosis dissecans in the fetlock, hock and stifle joints of 446 Hanoverian warmblood horses. SNP-based heritabilities were estimated using a genomic restricted maximum likelihood (GREML) method and accounting for patterns of regional linkage disequilibrium in the equine genome. In addition, we employed GREML for family data to account for different degrees of relatedness in the study population. Our results indicate that we were able to capture a larger proportion of additive genetic variance compared to pedigree-based estimates in the same population of Hanoverian horses. Heritability estimates on the linear scale for fetlock-, hock- and stifle-osteochondrosis dissecans were 0.41-0.43, 0.62-0.63, and 0.23-0.25, respectively, with standard errors of 0.11-0.14. Accounting for linkage disequilibrium patterns had an upward effect on the imputed data and a downward impact on the SNP array genotype data. GREML for family data resulted in higher heritability estimates for fetlock-osteochondrosis dissecans and slightly higher estimates for hock-osteochondrosis dissecans, but had no effect on stifle-osteochondrosis dissecans. The largest and most consistent heritability estimates were obtained when we employed GREML for family data with genomic relationship matrices weighted through patterns of regional linkage disequilibrium. Estimation of SNP-based heritability should be recommended for traits that can only be phenotyped in smaller samples or are cost-effective.

GVCHAP: A Computing Pipeline for Genomic Prediction and Variance Component Estimation Using Haplotypes and SNP Markers.

Haplotype prediction models open many possibilities to improve the accuracy of genomic selection but require more data processing and computing time than single-SNP prediction models. To facilitate haplotype analysis for genomic prediction and estimation using structural and functional genomic information, we developed a computing pipeline to implement haplotype analysis with capabilities for preparation of input data for haplotype analysis, genomic prediction and estimation using GVCHAP, and analysis of GVCHAP results. Data preparation includes utility programs for haplotype imputing; defining haplotype blocks by a fixed number of SNPs, a fixed distance in base pairs per block, or user defined block lengths based on structural or functional genomic information or a mixture of both types of information; and defining haplotype genotypes within each haplotype block. GVCHAP is the main program for genomic prediction and estimation, calculates GREML (genomic restricted maximum likelihood) estimates of variance components and heritabilities, and calculates GBLUP (genomic best linear unbiased prediction) for additive and dominance values of single SNPs as well as additive values of haplotypes with reliability estimates for training and validation populations. A two-step strategy and a method of multi-node processing are implemented to remove the computing bottleneck due to the creation of genomic relationship matrices for large samples. The analysis of GVCHAP results includes calculation of observed prediction accuracies from validation studies and preparation of input files for graphical visualization of heritability estimates of haplotype blocks as well as estimates of SNP effects and heritabilities. The entire pipeline provides an efficient and versatile computing tool for identifying the most accurate haplotype model among many candidate haplotype models utilizing structural and functional genomic information for genomic selection.

Genomic partitioning of growth traits using a high-density single nucleotide polymorphism array in Hanwoo (Korean cattle).

ObjectiveThe objective of this study was to characterize the number of loci affecting growth traits and the distribution of single nucleotide polymorphism (SNP) effects on growth traits, and to understand the genetic architecture for growth traits in Hanwoo (Korean cattle) using genome-wide association study (GWAS), genomic partitioning, and hierarchical Bayesian mixture models.MethodsGWAS: A single-marker regression-based mixed model was used to test the association between SNPs and causal variants. A genotype relationship matrix was fitted as a random effect in this linear mixed model to correct the genetic structure of a sire family. Genomic restricted maximum likelihood and BayesR: A priori information included setting the fixed additive genetic variance to a pre-specified value; the first mixture component was set to zero, the second to , the third , and the fourth to . BayesR fixed a priori information was not more than 1% of the genetic variance for each of the SNPs affecting the mixed distribution.ResultsThe GWAS revealed common genomic regions of 2 Mb on bovine chromosome 14 (BTA14) and 3 had a moderate effect that may contain causal variants for body weight at 6, 12, 18, and 24 months. This genomic region explained approximately 10% of the variance against total additive genetic variance and body weight heritability at 12, 18, and 24 months. BayesR identified the exact genomic region containing causal SNPs on BTA14, 3, and 22. However, the genetic variance explained by each chromosome or SNP was estimated to be very small compared to the total additive genetic variance. Causal SNPs for growth trait on BTA14 explained only 0.04% to 0.5% of the genetic varianceConclusionSegregating mutations have a moderate effect on BTA14, 3, and 19; many other loci with small effects on growth traits at different ages were also identified.

Bias in heritability estimates from genomic restricted maximum likelihood methods under different genotyping strategies.

We investigated the effects of different strategies for genotyping populations on variance components and heritabilities estimated with an animal model under restricted maximum likelihood (REML), genomic REML (GREML), and single-step GREML (ssGREML). A population with 10 generations was simulated. Animals from the last one, two or three generations were genotyped with 45,116 SNP evenly distributed on 27 chromosomes. Animals to be genotyped were chosen randomly or based on EBV. Each scenario was replicated five times. A single trait was simulated with three heritability levels (low, moderate, high). Phenotypes were simulated for only females to mimic dairy sheep and also for both sexes to mimic meat sheep. Variance component estimates from genomic data and phenotypes for one or two generations were more biased than from three generations. Estimates in the scenario without selection were the most accurate across heritability levels and methods. When selection was present in the simulations, the best option was to use genotypes of randomly selected animals. For selective genotyping, heritabilities from GREML were more biased compared to those estimated by ssGREML, because ssGREML was less affected by selective or limited genotyping.

Estimation of Genetic Correlation via Linkage Disequilibrium Score Regression and Genomic Restricted Maximum Likelihood

Genetic correlation is a key population parameter that describes the shared genetic architecture of complex traits and diseases. It can beestimated by current state-of-art methods, i.e., linkage disequilibrium score regression (LDSC) and genomic restricted maximum likelihood (GREML). The massively reduced computing burden of LDSC compared to GREML makes it an attractive tool, although the accuracy (i.e., magnitude of standard errors) of LDSC estimates has not been thoroughly studied. In simulation, we show that the accuracy of GREML is generally higher than that of LDSC. When there is genetic heterogeneity between the actual sample and reference data from which LD scores are estimated, the accuracy of LDSC decreases further. In real data analyses estimating the genetic correlation between schizophrenia (SCZ) and body mass index, we show that GREML estimates based on ∼150,000 individuals give a higher accuracy than LDSC estimates based on ∼400,000 individuals (from combined meta-data). A GREML genomic partitioning analysis reveals that the genetic correlation between SCZ and height is significantly negative for regulatory regions, which whole genome or LDSC approach has less power to detect. We conclude that LDSC estimates should be carefully interpreted as there can be uncertainty about homogeneity among combined meta-datasets. We suggest that any interesting findings from massive LDSC analysis for a large number of complex traits should be followed up, where possible, with more detailed analyses with GREML methods, even if sample sizes are lesser.

Genome-wide pleiotropy and shared biological pathways for resistance to bovine pathogens

Host genetic architecture is a major factor in resistance to pathogens and parasites. The collection and analysis of sufficient data on both disease resistance and host genetics has, however, been a major obstacle to dissection the genetics of resistance to single or multiple pathogens. A severe challenge in the estimation of heritabilities and genetic correlations from pedigree-based studies has been the confounding effects of the common environment shared among relatives which are difficult to model in pedigree analyses, especially for health traits with low incidence rates. To circumvent this problem we used genome-wide single-nucleotide polymorphism data and implemented the Genomic-Restricted Maximum Likelihood (G-REML) method to estimate the heritabilities and genetic correlations for resistance to 23 different infectious pathogens in calves and cows in populations undergoing natural pathogen challenge. Furthermore, we conducted gene-based analysis and generalized gene-set analysis to understand the biological background of resistance to infectious diseases. The results showed relatively higher heritabilities of resistance in calves than in cows and significant pleiotropy (both positive and negative) among some calf and cow resistance traits. We also found significant pleiotropy between resistance and performance in both calves and cows. Finally, we confirmed the role of the B-lymphocyte pathway as one of the most important biological pathways associated with resistance to all pathogens. These results both illustrate the potential power of these approaches to illuminate the genetics of pathogen resistance in cattle and provide foundational information for future genomic selection aimed at improving the overall production fitness of cattle.

The low single nucleotide polymorphism heritability of plasma and saliva cortisol levels

Cortisol is an important stress hormone affected by a variety of biological and environmental factors, such as the circadian rhythm, exercise and psychological stress. Cortisol is mostly measured using blood or saliva samples. A number of genetic variants have been found to contribute to cortisol levels with these methods. While the effects of several specific single genetic variants is known, the joint genome-wide contribution to cortisol levels is unclear. Our aim was to estimate the amount of cortisol variance explained by common single nucleotide polymorphisms, i.e. the SNP heritability, using a variety of cortisol measures, cohorts and analysis approaches. We analyzed morning plasma (n=5705) and saliva levels (n=1717), as well as diurnal saliva levels (n=1541), in the Rotterdam Study using genomic restricted maximum likelihood estimation. Additionally, linkage disequilibrium score regression was fitted on the results of genome-wide association studies (GWAS) performed by the CORNET consortium on morning plasma cortisol (n=12,597) and saliva cortisol (n=7703). No significant SNP heritability was detected for any cortisol measure, sample or analysis approach. Point estimates ranged from 0% to 9%. Morning plasma cortisol in the CORNET cohorts, the sample with the most power, had a 6% [95%CI: 0–13%] SNP heritability. The results consistently suggest a low SNP heritability of these acute and short-term measures of cortisol. The low SNP heritability may reflect the substantial environmental and, in particular, situational component of these cortisol measures. Future GWAS will require very large sample sizes. Alternatively, more long-term cortisol measures such as hair cortisol samples are needed to discover further genetic pathways regulating cortisol concentrations.

Genome-wide association study and accuracy of genomic prediction for teat number in Duroc pigs using genotyping-by-sequencing

BackgroundThe number of teats in pigs is related to a sow’s ability to rear piglets to weaning age. Several studies have identified genes and genomic regions that affect teat number in swine but few common results were reported. The objective of this study was to identify genetic factors that affect teat number in pigs, evaluate the accuracy of genomic prediction, and evaluate the contribution of significant genes and genomic regions to genomic broad-sense heritability and prediction accuracy using 41,108 autosomal single nucleotide polymorphisms (SNPs) from genotyping-by-sequencing on 2936 Duroc boars.ResultsNarrow-sense heritability and dominance heritability of teat number estimated by genomic restricted maximum likelihood were 0.365 ± 0.030 and 0.035 ± 0.019, respectively. The accuracy of genomic predictions, calculated as the average correlation between the genomic best linear unbiased prediction and phenotype in a tenfold validation study, was 0.437 ± 0.064 for the model with additive and dominance effects and 0.435 ± 0.064 for the model with additive effects only. Genome-wide association studies (GWAS) using three methods of analysis identified 85 significant SNP effects for teat number on chromosomes 1, 6, 7, 10, 11, 12 and 14. The region between 102.9 and 106.0 Mb on chromosome 7, which was reported in several studies, had the most significant SNP effects in or near the PTGR2, FAM161B, LIN52, VRTN, FCF1, AREL1 and LRRC74A genes. This region accounted for 10.0% of the genomic additive heritability and 8.0% of the accuracy of prediction. The second most significant chromosome region not reported by previous GWAS was the region between 77.7 and 79.7 Mb on chromosome 11, where SNPs in the FGF14 gene had the most significant effect and accounted for 5.1% of the genomic additive heritability and 5.2% of the accuracy of prediction. The 85 significant SNPs accounted for 28.5 to 28.8% of the genomic additive heritability and 35.8 to 36.8% of the accuracy of prediction.ConclusionsThe three methods used for the GWAS identified 85 significant SNPs with additive effects on teat number, including SNPs in a previously reported chromosomal region and SNPs in novel chromosomal regions. Most significant SNPs with larger estimated effects also had larger contributions to the total genomic heritability and accuracy of prediction than other SNPs.

Mixed model methods for genomic prediction and variance component estimation of additive and dominance effects using SNP markers.

We established a genomic model of quantitative trait with genomic additive and dominance relationships that parallels the traditional quantitative genetics model, which partitions a genotypic value as breeding value plus dominance deviation and calculates additive and dominance relationships using pedigree information. Based on this genomic model, two sets of computationally complementary but mathematically identical mixed model methods were developed for genomic best linear unbiased prediction (GBLUP) and genomic restricted maximum likelihood estimation (GREML) of additive and dominance effects using SNP markers. These two sets are referred to as the CE and QM sets, where the CE set was designed for large numbers of markers and the QM set was designed for large numbers of individuals. GBLUP and associated accuracy formulations for individuals in training and validation data sets were derived for breeding values, dominance deviations and genotypic values. Simulation study showed that GREML and GBLUP generally were able to capture small additive and dominance effects that each accounted for 0.00005–0.0003 of the phenotypic variance and GREML was able to differentiate true additive and dominance heritability levels. GBLUP of the total genetic value as the summation of additive and dominance effects had higher prediction accuracy than either additive or dominance GBLUP, causal variants had the highest accuracy of GREML and GBLUP, and predicted accuracies were in agreement with observed accuracies. Genomic additive and dominance relationship matrices using SNP markers were consistent with theoretical expectations. The GREML and GBLUP methods can be an effective tool for assessing the type and magnitude of genetic effects affecting a phenotype and for predicting the total genetic value at the whole genome level.