Prediction of haplotypes for ungenotyped animals and its effect on marker-assisted breeding value estimation

Genomic integration to identify molecular biomarkers associated with indicator traits of gastrointestinal nematode resistance in sheep

This study aimed to integrate GWAS and structural variants to propose possible molecular biomarkers related to gastrointestinal nematode resistance traits in Santa Inês sheep. The phenotypic records FAMACHA, haematocrit, white blood cell count, red blood cell count, haemoglobin, platelets and egg counts per gram of faeces were collected from 700 naturally infected animals, belonging to four Brazilian flocks. A total of 576 animals were genotyped using the Ovine SNP12k BeadChip and were imputed using a reference population with Ovine SNP50 BeadChip. The GWAS approaches were based on SNPs, haplotypes, CNVs and ROH. The overlapping between the significant genomic regions detected from all approaches was investigated, and the results were integrated using a network analysis. Genes related to the immune system were found, such as ABCB1, IL6, WNT5A and IRF5. Genomic regions containing candidate genes and metabolic pathways involved in immune responses, inflammatory processes and immune cells affecting parasite resistance traits were identified. The genomic regions, biological processes and candidate genes uncovered could lead to biomarkers for selecting more resilient sheep and improving herd welfare and productivity. The results obtained are the start point to identify molecular biomarkers related to indicator traits of gastrointestinal nematode resistance in Santa Inês sheep.

Genome-Wide Association Studies of Somatic Cell Count in the Assaf Breed

Simple Summary

Mastitis causes economic loss due to discarded milk and reduced milk production and quality, increased medical care costs and somatic cell count (SCC) penalties. The use of genetic markers associated with the variability of this trait through marker-assisted selection (MAS) could help traditional methods. Our objectives were to identify new single nucleotide polymorphisms (SNPs) and genes associated with mastitis resistance in Assaf sheep by using the Illumina Ovine Infinium® HD SNP BeadChip (680K). Firstly, corrected phenotype estimates for somatic cell score (SCS) were calculated using 6173 records from 1894 multiparous Assaf ewes, and were used to select 192 extreme animals (low SCS group: n = 96; and high SCS group: n = 96) for the genome-wide association study (GWAS). Four SNPs (rs419096188, rs415580501, rs410336647, and rs424642424), three of them totally linked, were found to be significant at the chromosome level (FDR 10%) in two different regions of OAR19 close to genes related to the immune system response. Validation studies of two SNPs (rs419096188 and rs424642424) by Kompetitive Allele-Specific PCR (KASP) genotyping in the total population (n = 1894) confirmed previous GWAS association results for the SCS trait. Finally, the SNP rs419096188 was also associated with lactose content trait.

Abstract

A genome-wide association study (GWAS) was performed to identify new single nucleotide polymorphisms (SNPs) and genes associated with mastitis resistance in Assaf sheep by using the Illumina Ovine Infinium^® HD SNP BeadChip (680K). In total, 6173 records from 1894 multiparous Assaf ewes with at least three test day records and aged between 2 and 7 years old were used to estimate a corrected phenotype for somatic cell score (SCS). Then, 192 ewes were selected from the top (n = 96) and bottom (n = 96) tails of the corrected SCS phenotype distribution to be used in a GWAS. Although no significant SNPs were found at the genome level, four SNPs (rs419096188, rs415580501, rs410336647, and rs424642424) were significant at the chromosome level (FDR 10%) in two different regions of OAR19. The SNP rs419096188 was located in intron 1 of the NUP210 and close to the HDAC11 genes (61 kb apart), while the other three SNPs were totally linked and located 171 kb apart from the ARPP21 gene. These three genes were related to the immune system response. These results were validated in two SNPs (rs419096188 and rs424642424) in the total population (n = 1894) by Kompetitive Allele-Specific PCR (KASP) genotyping. Furthermore, rs419096188 was also associated with lactose content.

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.

Assessing haplotype-based models for genomic evaluation in Holstein cattle

A single-nucleotide polymorphisms-based genomic relationship matrix (GSNP) discriminate less identity by state from identity by descent (IBD) alleles compared with a multi-locus haplotype-based relationship matrix (GHAP), which can better capture IBD alleles and recent relationships. We aimed to compare the prediction reliability and prediction bias of genomic best linear unbiased prediction (GBLUP) using either GSNP or GHAP in Holstein cattle. Therefore, a total of 57 traits with a wide range of heritability values were analyzed. Classical validation tests were done using a validation dataset comprised of 50k genotype records of 561–669 proven bulls born in 2010–2011 with an official estimated breeding value (EBV) in 2016 and a training set of 5314–19 678 bulls born before 2010, depending on the trait. The method for building the genomic relationship matrix (G) had significant, but small effect on observed reliability (r2GEBV) (p < 0.0001) and bias (p < 0.0001). A significant interaction between G and the level of trait heritability on r2GEBV and bias was also observed (p < 0.0001). The small gains in r2GEBV and small reductions in the bias by using GHAPBLUP were increased when predicting moderate to high-heritability traits compared with low-heritability traits.

Genetic evaluation of Atlantic salmon for growth traits incorporating SNP markers

Genetic evaluations of individual fish were calculated for growth traits in North American Atlantic salmon with and without inclusion of genetic markers. The number of SNP markers was reduced to 6,000 and further to 270 in order to reduce the problem of overparameterization. SNP genotypes were predicted for all ungenotyped animals in the pedigree. Analysis of traits used a model with polygenic effects and SNP markers together. Polygenic effects refer to the additive genetic effects that remain after accounting for SNP genotypes. SNP marker genotypes were included as covariates to evaluate fish for growth traits (weight and length) in different environments (freshwater and seawater) with genders separated. Including regressions on SNP marker genotypes reduced the sum of squares of residuals by 2.7%-12.5% and increased the variability of Mendelian sampling effects (i.e., within-family variation) compared to traditional animal model evaluations. Genetic evaluations may be carried out with a few hundred markers which may be more affordable for genotyping large numbers of fish.

Using markers with large effect in genetic and genomic predictions

The first attempts of applying marker-assisted selection (MAS) in animal breeding were not very successful because the identification of markers closely linked to QTL using low-density microsatellite panels was difficult. More recently, the use of high-density SNP panels in genome-wide association studies (GWAS) have increased the power and precision of identifying markers linked to QTL, which offer new possibilities for MAS. However, when GWAS started to be performed, the focus of many breeders had already shifted from the use of MAS to the application of genomic selection (using all available markers without any preselection of markers linked to QTL). In this study, we aimed to evaluate the prediction accuracy of a MAS approach that accounts for GWAS findings in the prediction models by including the most significant SNP from GWAS as a fixed effect in the marker-assisted BLUP (MA-BLUP) and marker-assisted genomic BLUP (MA-GBLUP) prediction models. A second aim was to compare the prediction accuracies from the marker-assisted models with those obtained from a Bayesian variable selection (BVS) model. To compare the prediction accuracies of traditional BLUP, MA-BLUP, genomic BLUP (GBLUP), MA-GBLUP, and BVS, we applied these models to the trait "number of teats" in 4 distinct pig populations, for validation of the results. The most significant SNP in each population was located at approximately 103.50 Mb on chromosome 7. Applying MAS by accounting for the most significant SNP in the prediction models resulted in improved prediction accuracy for number of teats in all evaluated populations compared with BLUP and GBLUP. Using MA-BLUP instead of BLUP, the increase in prediction accuracy ranged from 0.021 to 0.124, whereas using MA-GBLUP instead of GBLUP, the increase in prediction accuracy ranged from 0.003 to 0.043. The BVS model resulted in similar or higher prediction accuracies than MA-GBLUP. For the trait number of teats, BLUP resulted in the lowest prediction accuracies whereas the highest were observed when applying MA-GBLUP or BVS. In the same data set, MA-BLUP can yield similar or superior accuracies compared with GBLUP. The superiority of MA-GBLUP over traditional GBLUP is more pronounced when training populations are smaller and when relationships between training and validation populations are smaller. Marker-assisted GBLUP did not outperform BVS but does have implementation advantages in large-scale evaluations.

Multi-allelic haplotype model based on genetic partition for genomic prediction and variance component estimation using SNP markers

Background

The amount of functional genomic information has been growing rapidly but remains largely unused in genomic selection. Genomic prediction and estimation using haplotypes in genome regions with functional elements such as all genes of the genome can be an approach to integrate functional and structural genomic information for genomic selection. Towards this goal, this article develops a new haplotype approach for genomic prediction and estimation.

Results

A multi-allelic haplotype model treating each haplotype as an ‘allele’ was developed for genomic prediction and estimation based on the partition of a multi-allelic genotypic value into additive and dominance values. Each additive value is expressed as a function of h − 1 additive effects, where h = number of alleles or haplotypes, and each dominance value is expressed as a function of h(h − 1)/2 dominance effects. For a sample of q individuals, the limit number of effects is 2q − 1 for additive effects and is the number of heterozygous genotypes for dominance effects. Additive values are factorized as a product between the additive model matrix and the h − 1 additive effects, and dominance values are factorized as a product between the dominance model matrix and the h(h − 1)/2 dominance effects. Genomic additive relationship matrix is defined as a function of the haplotype model matrix for additive effects, and genomic dominance relationship matrix is defined as a function of the haplotype model matrix for dominance effects. Based on these results, a mixed model implementation for genomic prediction and variance component estimation that jointly use haplotypes and single markers is established, including two computing strategies for genomic prediction and variance component estimation with identical results.

Conclusion

The multi-allelic genetic partition fills a theoretical gap in genetic partition by providing general formulations for partitioning multi-allelic genotypic values and provides a haplotype method based on the quantitative genetics model towards the utilization of functional and structural genomic information for genomic prediction and estimation.

Genetic evaluation with major genes and polygenic inheritance when some animals are not genotyped using gene content multiple-trait BLUP

Background

In pedigreed populations with a major gene segregating for a quantitative trait, it is not clear how to use pedigree, genotype and phenotype information when some individuals are not genotyped. We propose to consider gene content at the major gene as a second trait correlated to the quantitative trait, in a gene content multiple-trait best linear unbiased prediction (GCMTBLUP) method.

Results

The genetic covariance between the trait and gene content at the major gene is a function of the substitution effect of the gene. This genetic covariance can be written in a multiple-trait form that accommodates any pattern of missing values for either genotype or phenotype data. Effects of major gene alleles and the genetic covariance between genotype at the major gene and the phenotype can be estimated using standard EM-REML or Gibbs sampling. Prediction of breeding values with genotypes at the major gene can use multiple-trait BLUP software. Major genes with more than two alleles can be considered by including negative covariances between gene contents at each different allele. We simulated two scenarios: a selected and an unselected trait with heritabilities of 0.05 and 0.5, respectively. In both cases, the major gene explained half the genetic variation. Competing methods used imputed gene contents derived by the method of Gengler et al. or by iterative peeling. Imputed gene contents, in contrast to GCMTBLUP, do not consider information on the quantitative trait for genotype prediction. GCMTBLUP gave unbiased estimates of the gene effect, in contrast to the other methods, with less bias and better or equal accuracy of prediction. GCMTBLUP improved estimation of genotypes in non-genotyped individuals, in particular if these individuals had own phenotype records and the trait had a high heritability. Ignoring the major gene in genetic evaluation led to serious biases and decreased prediction accuracy.

Conclusions

CGMTBLUP is the best linear predictor of additive genetic merit including pedigree, phenotype, and genotype information at major genes, since it considers missing genotypes. Simulations confirm that it is a simple, efficient and theoretically sound method for genetic evaluation of traits influenced by polygenic inheritance and one or several major genes.

Haplotype phasing after joint estimation of recombination and linkage disequilibrium in breeding populations

A novel method for haplotype phasing in families after joint estimation of recombination fraction and linkage disequilibrium is developed. Results from Monte Carlo computer simulations show that the newly developed E.M. algorithm is accurate if true recombination fraction is 0 even for single families of relatively small sizes. Estimates of recombination fraction and linkage disequilibrium were 0.00 (SD 0.00) and 0.19 (SD 0.03) for simulated recombination fraction and linkage disequilibrium of 0.00 and 0.20, respectively. A genome fragmentation phasing strategy was developed and used for phasing haplotypes in a sire and 36 progeny using the 50 k Illumina BeadChip by: a) estimation of the recombination fraction and LD in consecutive SNPs using family information, b) linkage analyses between fragments, c) phasing of haplotypes in parents and progeny and in following generations. Homozygous SNPs in progeny allowed determination of paternal fragment inheritance, and deduction of SNP sequence information of haplotypes from dams. The strategy also allowed detection of genotyping errors. A total of 613 recombination events were detected after linkage analysis was carried out between fragments. Hot and cold spots were identified at the individual (sire level). SNPs for which the sire and calf were heterozygotes became informative (over 90%) after the phasing of haplotypes. Average of regions of identity between half-sibs when comparing its maternal inherited haplotypes (with at least 20 SNP) in common was 0.11 with a maximum of 0.29 and a minimum of 0.05. A Monte-Carlo simulation of BTA1 with the same linkage disequilibrium structure and genetic linkage as the cattle family yielded a 99.98 and 99.94% of correct phases for informative SNPs in sire and calves, respectively.

Comparison of analyses of the QTLMAS XIV common dataset. I: genomic selection

BACKGROUND

For the XIV QTLMAS workshop, a dataset for traits with complex genetic architecture has been simulated and released for analyses by participants. One of the tasks was to estimate direct genomic values for individuals without phenotypes. The aim of this paper was to compare results of different approaches used by the participants to calculate direct genomic values for quantitative trait (QT) and binary trait (BT).

RESULTS

Participants applied 26 approaches for QT and 15 approaches for BT. Accuracy for QT was between 0.26 and 0.89 for males and between 0.31 and 0.89 for females, and for BT ranged from 0.27 to 0.85. For QT, percentage of lost response to selection varied from 8% to 83%, whereas for BT the loss was between 15% and 71%.

CONCLUSIONS

Bayesian model averaging methods predicted breeding values slightly better than GBLUP in a simulated data set. The methods utilizing genomic information performed better than traditional pedigree based BLUP analyses. Bivariate analyses was slightly advantageous over single trait for the same method. None of the methods estimated the non-additivity of QTL affecting the QT, which may be one of the constrains in accuracy observed in real data.

Imputation of missing single nucleotide polymorphism genotypes using a multivariate mixed model framework

The objective of this paper was to investigate, for various scenarios at low and high marker density, the accuracy of imputing genotypes when using a multivariate mixed model framework using information from 2, 4, or 10 surrounding markers. This model predicts genotypes at a locus, using genotypes at nearby loci as correlated traits, and the additive genetic relationship matrix to use information from genotyped relatives. For 2 scenarios this method was compared with the population-based imputation algorithms FastPHASE and Beagle. Accuracies of imputation were obtained with Monte Carlo simulation and predicted with selection index theory, using input from the simulated data. Five different scenarios of missing genotypes were considered: 1) genotypes of some loci are missing due to genotyping errors, 2) juvenile selection candidates are genotyped using a smaller SNP panel, 3) some animals in the pedigree of a breeding population are not genotyped, 4) juvenile selection candidates are not genotyped, and 5) 1 generation of animals in the top of the pedigree are not genotyped. Surrounding marker information did not improve accuracy of imputation when animals whose genotypes were imputed were not genotyped for those surrounding markers. When those animals were genotyped for surrounding markers, results indicated a limited gain when linkage disequilibrium (LD) between SNP was low, but a substantial increase in accuracy when LD between SNP was high. For scenario 1, using 1 vs. 11 SNP, accuracy was respectively 0.75 and 0.81 at low, and 0.75 and 0.93 at high density. For scenario 2, using 1 vs. 11 SNP, accuracy was, respectively, 0.70 and 0.73 at low, and 0.71 and 0.84 at high density. Beagle outperformed the other methods at high SNP density, whereas the multivariate mixed model was clearly superior when SNP density was low and animals where genotyped with a reduced SNP panel. The results showed that extending the univariate gene content method to a multivariate BLUP model with inclusion of surrounding marker information only yields greater imputation accuracy when the animals with imputed loci are at least genotyped for some SNP that are in LD with the SNP to be imputed. The equation derived from selection index theory accurately predicted the accuracy of imputation using the multivariate mixed model framework.