Search

International audience

International audience

International audience

International audience

International audience

International audience

Additional file 1 Metafounder’s theory as applied to the S matrix, rules and code for this matrix and its inverse.

International audience

Additional file 2: Table S1. Comparison of realized (marker-based) and expected (pedigree-based) additive genetic relationships on the X-chromosome (specific part) for different categories of animals. Table S2. Comparison of realized (marker-based) and expected (pedigree-based) additive genetic relationships on the autosomes (all together) for different categories of animals. Table S3. Comparison of realized (marker-based) and expected (pedigree-based) additive genetic relationships on the BTA2 for different categories of animals.

International audience; The recently presented method Linear Regression (LR), was proposed to benchmark models in genetic selection schemes. The principle of the method is to estimate biases and accuracies comparing estimated breeding values (EBVs) obtained from a partial data set (with old records, EBVp) and from a whole data set (with the addition of new records, EBVw). In this study, we estimate μwp (the bias or difference between average EBVp and TBV, with an expected value of 0), bwp (the slope of regression of TBV on EBVp, with an expected value of 1) and ρpw (inversely proportional to increase in accuracy from ‘partial’ to ‘whole’) for Manech Tete Rousse dairy sheep. Pedigree contains 540,986 individuals born from 1950 to 2017 and the phenotypic data file includes milk production of 1,842,295 lactations. No marker data was used. The model used to estimate breeding values includes five fixed effects as well as the permanent effect for milk production. The heritability used was 0.30. The model takes account of heterogeneous residual variances and genetic groups. The comparisons involved parent average EBV of rams born from 2005 to 2015 (as EBVp) and results from first daughters (as EBVw). Finally, we obtained statistic from 11 pairs of EBVp-EBVw. The mean (mean ± standard error) of μwp, expressed in genetic standard deviation of trait, was 0.26±0.03 and the mean of bwp was 0.80±0.09, indicating a larger bias. On other hand, the mean of ρpw was 0.51±0.04; in other words, accuracy doubles with first progeny performance. As conclusion, Manech Tete Rousse parent average EBVs are over dispersed for unknown reasons. Work financed by the ARDI project of Poctefa program (European Union FEDER funds), Metaprogram GENSEL (INRA) and La Region Occitanie.

International audience; Indirect predictions (IP) can be obtained for young animals that are not yet included or will never make up to an official genomic evaluation. Additionally, if lots of animals are genotyped but they do not contribute to the evaluation, having IP for them would reduce computing cost. This is beneficial if accuracy levels are maintained. If single-step GBLUP (ssGBLUP) is the official evaluation method, SNP effects can be obtained by backsolving genomic EBV (GEBV). Subsequently, IP are obtained by the sum of SNP effects weighted by gene content. When IP are released, a measure of accuracy that reflects the standard error of IP is needed. Our objective was to implement formulas to calculate accuracy for IP and to apply the method in a cattle population. The American Angus data used in this study consisted of 35k records for post-weaning gain and pedigree information for 202k animals, of which 60k were genotyped for 38.4k SNP (after quality control). From the genotyped animals, 2k were in the validation set. A complete dataset had phenotypes and pedigree up to 2013, genotypes up to 2014, and was used to calculate the benchmark accuracy by the inverse of the LHS. This accuracy is referred to as GEBV accuracy. A reduced dataset had phenotypes, pedigree and genotypes up to 2013, and was used to obtain the prediction error covariance (PEC) for SNP effects. This PEC was used to calculate IP accuracy. The process of obtaining IP accuracy involved several steps: factorization and inversion of the LHS of ssGBLUP mixed model equations; extraction of the inverse of the LHS for genotyped animals; calculation of SNP effects from GEBV; computation of PEC for SNP effects; calculation of IP; computation of accuracy of IP as a function of PEC for SNP effect and SNP content. Correlations between GEBV and IP accuracies were over 0.997, indicating the possibility of getting a good measure of IP accuracy based on ssGBLUP, without directly switching to a SNP-BLUP model. Obtaining IP accuracies from large-scale ssGBLUP evaluations, which use the algorithm for proven and young, will be the subject of future investigations.

International audience; Inbreeding depression is caused by increased homozygosity of individuals and it is involved in the decrease of performance and fitness of the animals. The goal of this study is to estimate inbreeding depression for semen traits in Basco-Béarnaise rams. Genomic (e.g. using SNP-by-SNP approach) or pedigree-based inbreeding can be used to obtain inbreeding coefficients (f). It is possible to detect inbreeding depression (b) from the regression of the phenotype (y) on f, as y=Xβ+fb+u+e, where u is the vector of breeding values. In sheep, inbreeding from pedigree tends to be underestimated due to missing pedigrees. Inbreeding coefficients, f, were estimated using four methods. The first method used the traditional pedigree-based inbreeding. The second used pedigree-based inbreeding but accounting for non-zero relationships for unknown parents. The third method used the metafounder relationships in pedigree-based inbreeding. The fourth method calculates inbreeding from a combined relationship matrix using pedigree and genotypes with metafounder relationships. Breeding values, u, for the first three methods were estimated with the classical BLUP using phenotypes and pedigree information and the fourth method was Single Step GBLUP with metafounders.

International audience; Our objective was to identify genomic regions associated with individual growth (ADGf), and cage feed intake (FIf), feed conversion ratio (FCRf) and residual feed intake (RFIf) in ad libitum-fed rabbits; as well as with individual growth (ADGr) in rabbits fed under restriction. Three genome-wide association methods were used to analyze the aforementioned traits on 438 rabbits genotyped for 114,604 SNPs. The methods were: i) regression on each SNP allele counting (Reg), ii) bivariate animal model between the performance trait and each SNP allele counting (Bi) and iii) random QTL model combining LD and linkage information (LDLA). LDLA returned the largest number of significant associations. Bi and Reg yielded significant associations only at chromosome (Chr) level. At this level, LDLA showed 22 chromosomal regions on Chr2, 3, 5, 6, 7, 9, 12, 14, 15, 18, 19 and 21 associated with ADGf. Reg yielded significant associations between ADGf and regions on Chr3, 5 and 21. Lastly, Bi revealed significant associations on Chr3, 5 and 16. All methods identified the region 102–113 Mb on Chr3 as significantly associated with ADGf. At Chr5, the 19–20 Mb region was associated with ADGf by Reg and Bi, while the 34–35 Mb region was associated with ADGf by Bi and LDLA. One region (0–2 Mb) on Chr13 and 2 regions (29–32 Mb and 38–39 Mb) on Chr9 were significantly associated with ADGr by Reg and Bi methods, respectively. Bi, the only method used to analyze traits recorded at cage level, revealed significant associations between FIf and region 3–4 Mb on Chr5, and between RFIf and regions Chr8:108–110 Mb and Chr21:7–8 Mb. No significant associations were identified for FCRv. A total of 36 chromosome regions were significantly associated with the traits under study, one of them (Chr21:7–8 Mb) having pleiotropic effect on ADGf and RFIf. After gene and functional annotation, 29 candidate genes on 22 out of these 36 significant regions were identified. It is remarkable that one of the identified candidate genes (FTO) was previously associated with growth in rabbits. In addition, candidate genes FEZF2 and PTPRG for ADGr, are involved in the control of feeding behavior traits.

International audience; Traditionally, inbreeding estimates have been estimated based on pedigree information. However, in sheep there is a considerable proportion of unknown pedigree due to natural mating and limited use of paternity analysis. Therefore, there is an under estimation of inbreeding coefficients based on pedigree. In the genomics era, genomic information can be used to estimate inbreeding. In this study, three different inbreeding estimation methods were assessed (a pedigree-based methodology, a single SNPbased approach and a method based on runs of homozygosity, ROH) to analyse the genetic diversity of three populations of Latxa dairy sheep: Latxa Cara Rubia (LCR) and Latxa Cara Negra from Euskadi (LCNEUS) and from Navarre (LCNNAF). A total of 981 animals were genotyped with the Illumina OvineSNP50 BeadChip, bringing around 41200 SNPs and 4468 animals in pedigree. The results found for LCNEUS and LCNNAF showed an effective population size (Ne) below 100 when inbreeding coefficients were estimated based on pedigree or ROH: Ne PED=64, Ne ROH=86 for LCNEUS, and Ne PED=53, Ne ROH=66 for LCNNAF. Nevertheless, SNP based estimations yielded higher values: Ne SNP= 282 and 153 for LCNEUS and LCNNAF, respectively. LCNEUS showed a higher genetic diversity than LCNNAF in any of the evaluated methods. There is an evident difference between pedigree and ROH based results and the SNP based ones, possibly due to the reduced number of genotyped animals. In the case of LCR, which historical importation of semen from the French Manech Tête Rousse (MTR) has avoided the increase of inbreeding per generation, the estimation of effective population size is meaningless. For this breed a study of coancestry between the two breeds has been done, based on pedigree and genomic data, to analyse the evolution of genetic variability. Both methods have reflected, as expected, the important effect on inbreeding and genetic variability of introducing animals from another close breed.

International audience

International audience

International audience; Runs of homozygosity (ROH) are contiguous homozygous segments of the genome where the haplotypes inherited from each parent are identical. Currently, inbreeding estimated from ROH is considered a powerful approach to distinguish between recent,potentially harmful, from ancient inbreeding, potentially beneficial. Accordingly, inbreeding based on ROH can help to improve the understanding of inbreeding depression. The occurrence of ROH is not randomly distributed across the genome, and islands of ROH across a large number of animals may be the result of selective pressure. The objective of this study is to evaluate whether ROH can be used to explore signatures of selection in French dairy sheep. The data set available included animals from various breeds and subpopulations: Basco-Béarnaise breed (BB); Manech Tête Noire breed (MTN); Manech Tête Rousse breed (MTR); Lacaune Confederation subpopulation (LACCon); and Lacaune Ovitest subpopulation (LACOvi). Animals were genotyped with the Illumina OvineSNP50 BeadChip. After applying filtering criteria, the genomic data included 38,287 autosomal SNPs distributed across 26 chromosomes and 8,700 individuals. One island of ROH was detected on autosome 6 in the same genomic position across animals (between 30 – 40 Mb). Wright’s differentiation coefficients for two SNPs within this island of ROH were high (0.67 – 0.68). The linkage disequilibrium between both SNPs was also elevated (0.98). The divergence in allele frequencies in those SNPs grouped BB, MTN and MTR breeds in one cluster, and LACCon and LACOvi subpopulations in another cluster. The closest candidate gene is NCAPG-LCORL, which has been reported to be under positive selection and suggested to control height and stature in sheep, as well as growth and feed efficiency in cattle. These findings contribute to the understanding of the effects of selection in shaping the sheep genome.

Traditionally, inbreeding estimates have been estimated based on pedigree information. However, in sheep there is a considerable proportion of unknown pedigree due to natural mating and lack of filiation analysis. Therefore, there could be an under estimation of inbreeding coefficients based on pedigree. In the genomics era, genomic information can be used to estimate inbreeding. In this study, three different inbreeding estimation methods were assessed (a pedigree-based methodology, a single SNP-based approach and a ROH based method) to analyse the genetic diversity of two populations of Latxa Cara Negra breed: Latxa Cara Negra from Euskadi (LCNEUS) and from Navarre (LCNNAF). A total of 533 animals were genotyped with the Illumina OvineSNP50 BeadChip, bringing around 41700 SNPs and 2505 animals in pedigree. The results showed an effective population size below 100 in most of the evaluated cases. In addition, LCNEUS showed a higher genetic diversity than LCNNAF for each evaluated method. These results suggest that it could be interesting to consider these estimates in the management of both populations.

National audience

National audience

International audience; Traditionally, inbreeding has been estimated from pedigree-based information. Nevertheless, in the last years the interest for estimating inbreeding using genomic information has considerably grown. The objective of this study is to evaluate whether inbreeding estimated from ROHs or calculated on a SNP-by-SNP basis provide an accurate measure of inbreeding. The data set available included individuals from French dairy sheep breeds and subpopulations (Basco-Béarnaise, BB; Manech Tête Noire, MTN; Manech Tête Rousse, MTR; Lacaune Confederation, LACCon; and Lacaune Ovitest, LACOvi). Those animals were genotyped with the Illumina OvineSNP50 BeadChip. After filtering, the genomic data included 38,287 autosomal SNPs and 8,700 individuals. Pedigree from those genotyped animals comprised 72,803 animals. Genome-based inbreeding was higher than pedigree-based inbreeding. This can be explained considering that pedigree-based inbreeding assumes that loci are neutral and it does not consider the potential bias resulting from selection. The highest rates of increase in inbreeding per generation were observed when using pedigree (0.0022 – 0.0099; BB, MTN and LACOvi) and ROHs (0.0022 – 0.0046; MTR and LACCon). The lowest rates of increase in inbreeding per generation were observed when calculating inbreeding on a SNP-by-SNP basis (0.0016 – 0.0044), except for LACOvi where the lowest rates were also observed with ROHs (0.0014). Both genome-based measures rank breeds and subpopulations in the same order regarding the effective population size: LACOvi (357), LACCon (227 – 313), MTR (109 – 200), MTN (81 – 179) and BB (59 – 114). Effective population size obtained from pedigree information provided similar estimates. ROHs yield a correct estimate of inbreeding at the genomic level and make it possible to identify specific IBD regions. These regions can be used to improve mating decisions and to minimise the unfavourable effects of inbreeding. Work financed by the ARDI project of Poctefa program (European Union FEDER funds).

International audience; Cross validation is the most extended method to estimate the prediction ability in animal selection schemes; however it present problems on the quality of results. The Linear Regression (LR) method compares EBVs obtained with old (‘partial’) and old+new (‘whole’) data to infer biases and accuracies. In this work, we present preliminary results on the behaviour of LR method using simulated data. Based on heritabilities of 0.5 and 0.1, 20 populations were simulated with the QMSim software. In the simulation individuals were selected by BLUP evaluation and mating system were performed to reduce the average inbreeding and kinship of the population. Only fathers born in generation 5 with at least 5 daughters in generation 6 were used. BLUP pedigree evaluations were performed using a partial data set (without daughter’s information) and a whole data set (with daughter’s information). Statistics were obtained between the estimated breeding values of the partial (EBVp) and whole data sets (EBVw) and between the EBVp and true breeding values (TBV) obtained from the simulation. Five statistics were considered: (bias) difference between average EBVp and EBVw, with an expected value of 0, (slope) regression of EBVw on EBVp with an expected value of 1, (accuracies) correlation between EBVp and EBVw (ρpw) (proportional to increase in accuracy), covariance between EBVp and EBVw (proportional to accuracy on partial) and regression of EBVp in EBVw (proportional to increase in reliabilities). All metrics were also calculated by substituting EBVw for TBV to ascertain if metrics using EBVw predict results for TBV. The statistics comparing EBVp-EBVw and EBVp- TBV are almost identical for both values of h2. The most important differences were observed in ρpw (in h2 of 0.1: EBVp-EBVw=0.36 and EBVp-TBV=0.27; h2 of 0.5: EBVp-EBVw=0.32 and EBVp-TBV=0.27) and it was always overestimated on EBVp-EBVw. In short, biases and accuracies were correctly estimated using statistics from ‘whole’ and ‘partial’ genetic evaluations. The similarity of the results obtained using EBVw or TBV suggest that statistics proposed in method LR could be useful for measure of bias and accuracies in breeding schemes. Further work will include genomic information.

International audience

International audience; In this study the genetic diversity of five selected French dairy sheep subpopulations and breeds was quantified. Three different inbreeding estimation methods were assessed: a pedigree-based methodology, a single SNP-based approach and a ROHbased method. The data set available included individuals from the Basco-Béarnaise (BB) breed, Manech Tête Noire (MTN) breed, Manech Tête Rousse (MTR) breed, Lacaune Confederation (LACCon) subpopulation and Lacaune Ovitest (LACOvi) subpopulation. Animals were genotyped with the Illumina OvineSNP50 BeadChip. After filtering, the genomic data included 38,287 autosomal SNPs and 8,700 individuals. Pedigree from the genotyped animals comprised 72,803 animals. The results indicate that no significant differences were observed in e N estimates between the three inbreeding estimation methods. In general, estimates of e N were above 200 in LACCon and LACOvi subpopulations and below 200 in BB, MTN and MTR breeds.

International audience

International audience

International audience; he Four Horsemen Of Genomicalypse: Fuzzy notions in genomic selection I describe four notions that I found ill-defined in genomic selection. Linkage Disequilibrium plays a vital role in genomic selection, yet we lack of conceptual tools to measure, describe, and predict multilocus linkage equilibrium. There is no consensus on what “relationship” means – we suggest to base it on observed similarity at markers. Effective number of markers should ideally be a populational parameter with a clear estimation method and interpretation. We should not conclude from simulations with oversimplified genetic architecture which assumes additive genetic architecture and point locations of QTL loci

International audience; The breeding organizations of the Lacaune dairy sheep introduced genomic selection in 2015. Results from the first crop of daughters from genomically selected rams confirm its efficiency: after a genomic selection proportion of 1/3 at 3 month-old, less than 20% of AI rams born in 2015 have been culled at 2.5 year-old in 2017, versus 50% previously in classical selection, according to the total merit index threshold used to select the progeny-tested rams. This paper also highlights practical considerations that the breeding societies had to handle regarding genomic selection

International audience; Much effort has been recently put into identifying causative quantitative trait nucleotides (QTN) in animal breeding, aiming more accurate genomic prediction. Among the genomic methods available, single-step GBLUP (ssGBLUP) became the choice because of its simplicity and potentially higher accuracy. When QTN are known, they need to be properly weighted, so the accuracy can be maximized. The weighted ssGBLUP is still under development, and a proper weighing algorithm is needed. The objectives of this study were to investigate whether ssGBLUP is useful for genomic prediction when causative variants are known and to verify the impact of different SNP weighting in ssGBLUP compared to GBLUP. Analyses involved about 4M records for stature of 3M cows. Genotypes were available for 27k sires for a regular 54k chip (BovineSNP50; Illumina), and imputed with extra 17k sequence variants having largest effects, including causative variants affecting stature and 32 other traits. Direct genomic value (DGV) and genomic EBV (GEBV) were calculated using GBLUP and ssGBLUP with regular genomic relationship matrices (G). Later, G was weighted based on the squared value of SNP effects or with nonlinear A equations, which limits the changes in SNP weights. In GBLUP, the residuals were either homogeneous or heterogeneous. Reliability (R2) was assessed from forward prediction of DGV or GEBV for young sires with at least 10 daughters. The lowest to highest R2 for DGV were by GBLUP with homogeneous residuals, GBLUP with heterogeneous residuals, and extracted from ssGBLUP. SNP weighting by nonlinear A increased R2 by up to 1.7% with homogeneous residuals and by up to 0.2% with heterogeneous residuals, compared to unweighted GBLUP. Linear weighting reduced accuracy in GBLUP and had no effect in ssGBLUP. Overall, adding 17k causative variants increased accuracy up to 0.6% in GBLUP, but had no impact in ssGBLUP. Reliability for DGV extracted from ssGBLUP was at least 0.6% more accurate than any DGV from GBLUP. Linear weighting is not helpful with causative variants with small effect. Gains with SNP weighting in multistep (GBLUP or SNP BLUP) may be partly due to corrections in modeling issues associated with pseudo-observations.

International audience; The objective of this study was to implement single-step genomic BLUP (ssGBLUP) for national Angus cattle evaluation in the US. National evaluations include a variety of models with several linear and categorical traits, maternal effects, multibreed data, and a large number of genotyped animals. For the initial investigation, we used a dataset from 2014 that comprised over 8 million animals, 6 million birth weight (BW) and weaning weight (WW) records, 3.4 million post-weaning gain (PWG) records, and genotypes for 52k animals. A dataset from 2017 was later used that included 335k genotyped animals. The ability to predict future performance of young animals was investigated when using regular BLUP and ssGBLUP. Because of the increasing number of genotyped animals and the high computing cost to invert the genomic relationship matrix (G), the algorithm for proven and young (APY) was used to approximate the inverse of G. The APY uses recursions on a small subset of genotyped animals, called core. We further tested the feasibility of having daily interim genomic predictions for newly-genotyped animals based on SNP effects derived from the previous official ssGBLUP evaluation. In addition, we extended all models used in traditional evaluations to ssGBLUP, and compared genetic trends from traditional BLUP, ssGBLUP, and a multistep method that was implemented for the American Angus genomic evaluation in 2009. A new algorithm to approximate accuracy of GEBV for large genomic data was also developed. On average, the increase in ability to predict future performance, for BW, WW, and PWG, with ssGBLUP was 25% in the 2014 data and 36% in the 2017 data, compared to the traditional BLUP. The ssGBLUP with APY was as accurate as the regular ssGBLUP when the number of core animals was at least 10,000, independently of which animals were in the core group. Interim predictions derived from ssGBLUP provided accurate genomic values for newly-genotyped animals. Genetic trends for ssGBLUP and BLUP were similar, revealing overestimation in multistep evaluations, especially for traits with less phenotypes. Single-step GBLUP became a reality for American Angus evaluation and its implementation process resulted in successful updates in methodology, making this approach mature for national beef cattle evaluation.

International audience; Past studies suggested limited dimensionality of the genomic SNP information was related to effective population size (Ne). The objective of this study was to estimate that dimensionality with simulated and with livestock (Holstein, Jersey, Angus, pigs, and chicken) data sets. That dimensionality can be defined as the number of non-negligible singular values of gene content, or the number of non-negligible eigenvalues of genomic relationship matrix (GRM). In this study, eigenvalue analysis determined the numbers of largest eigenvalues corresponding to 90, 95, 98, and 99% of the variation of the GRM for each population. With many genotyped animals and SNP markers, the numbers corresponding to 90, 95, and 98% approached NeL, 2NeL and 4NeL, respectively, where L is genome length in Morgans. Realized accuracies were calculated for single-step GBLUP (ssGBLUP) with an algorithm for inversion of GRM that takes into account the limited dimensionality of the genomic information. Realized accuracies peaked with the dimensionality corresponding to 98 to 99% of variation depending on population, indicating that 1 to 2% of variation in the GRM was due to noise. However, the accuracies were only slightly reduced at half the optimum dimensionality. Subsequently, the dimensionality of the genomic information was estimated at about 14,000 for Holstein and Angus cattle, 12,000 for Jersey cattle, and 6000 for pigs and chickens, which corresponds approximately to 3NeL. Based on interpolation of simulated and real data with L of 30 Morgans, approximate Ne was 149 for Holsteins, 101 for Jerseys, 113 for Angus, and 44 for chickens; for pigs and L of 20 Morgans, approximate Ne was 48. Limited dimensionality of the genomic information has serious implications for genomic prediction and possibly GWAS.

International audience; Socially affected traits must be recorded in group-housed animals. In these conditions sometimes it is not possible to obtain individual records of certain traits and in some species. Previous studies have addressed the issue on how to use group (i.e. pooled) data to estimate genetic parameters and to predict breeding values, but the value of this pooled data to conduct QTL mapping studies has not been assessed yet. . Our objective was to present a method, based on a Bayesian single step genomic evaluation, in which the SNPs effects were assumed to follow the prior of Bayesian model, allowing thus a variable selection approach to pinpointing the genome regions most likely harbouring QTLs. The method was applied to a multi-trait simulated data set, in which for one of the traits pooled records were generated summing groups of 10 individual records. Our results show that an important loss of power was observed when pooled data were used, but even though one of the true QTLs can be detected with a probability of being associated to the trait of 0.83. This QTL was associated to a mutation explaining 16% of the genetic variance and with a frequency of 0.46. Other mutation with an even greater effect (22%) but with lower frequency (0.38) could not be detected. It can be concluded that although the proposed model can be used for QTL mapping when grouped data are available its power is limited and only strongly associated regions are likely to be declared as QTLs.

International audience

International audience; Epistatic genetic variances for litter size in pigs using genomic models Z.G. Vitezica1, H. Crichan1, W. Herring2, A. Reverter3 & A. Legarra1 1 INRA/INPT, UMR 1388 GenPhySE, F-31326 Castanet-Tolosan, France zulma.vitezica@ensat.fr (Corresponding Author) 2 PIC North America, Hendersonville, USA 3CSIRO Agriculture & Food, 306 carmody Rd., St Lucia, Queensland 4067, Australia Epistatic genomic relationship matrices for interactions of any-order can be constructed using Hadamard products of additive and dominant genomic orthogonal matrices plus a standardization based on the trace of these matrices. With a pig genomics dataset, the broad-sense heritability for litter size was almost twice the narrow-sense heritability. Estimates of variance components were stable across models. Including epistasis reduced bias without changing prediction accuracy in the genetic evaluation. Epistasis can be exploited in mating allocation to maximize the overall genetic merit.

International audience; Metafounders are representations of inbred and related base populations. Dairy sheep breeding schemes have steady genetic progress but also unrecorded parentships of 10 to 80% of the individuals depending on the breed. Thus,use of genetic groups is essential to obtain unbiased breeding values and correct genetic trends. However, fitting genetic groups as fixed effects is not tenable in view of genomic information and metafounders are a theoretically appealing solution. This involves estimation of relationships within and across metafounders (Gamma: Γ) using genotyped of individuals that may be several generations removed from them. Here, we present experiences in estimating Gamma in Manech Tete Rousse by a pseudo-EM maximum likelihood method. The number of genotyped individuals is 2,111, and a pedigree with their ancestors included 15,270 individuals and 11 metafounders (Gamma is a 11×11 matrix) defined every 3-4 years from 1975 to 2009. These metafounders only model missing pedigrees, there is no introduction of foreign animals in this breed. The pseudo-EM algorithm always maximized the likelihood and different initial points ended up in the same estimates of Gamma. Results show a self-relationship of 0.47 to 0.54, increasing with time, and high relationships (correlations ranging from 0.89 to 0.98) across metafounders.Closer metafounders are more related, as expected. Results are coherent with a priori knowledge of the dynamics of the breed.

National audience

National audience

Additional file 1: Tables S1 and S2. Correlation among the 16 statistics employed in the cross-validation study of the beef cattle dataset using the pedigree-based NRM or the SNP-based GRM (NB: These are the values used to generate the left and right panels of Fig. 2

International audience; Single-step genomic BLUP (ssGBLUP) is the method of choice for genomic evaluation because of simplicity and ability to combine pedigree, genotypes, and phenotypes. As the availability of sequence data is increasing, we investigated whether ssGBLUP can be useful for genomic analyses when causative Quantitative Trait Nucleotide (QTN) are known. Simulations included 180k animals from 11 generations. Phenotypes (h2=0.3) were available for all animals in generations 6-10. A total of 24k parents in generations 6-10 and 5k validation animals in generation 11 were genotyped for 60k SNP. The genetic variance was fully accounted for by 100 or 1000 biallelic QTN. Genomic relationship matrices were computed based on: unweighted 60k SNP; unweighted 60k SNP and QTN; 60k SNP and QTN with variance from GWAS; unweighted 60k SNP and QTN with known variance; only QTN. Accuracy of EBV for validation animals were computed by BLUP and ssGBLUP. To ensure full rank, the genomic relationship matrix (G) was blended with 1% or 5% of the numerator relationship matrix. The inverse of G was computed directly or using APY, the algorithm that exploits the limited dimensionality of G for sparse computations. This dimensionality was calculated as the number of largest eigenvalues explaining 98% of the variance of G. Rank of G with 100 QTN was 16,980 for G based on unweighted 60k SNP; 19,112 after 5% blending; 5,093 for G with 60k SNP and QTN with known variance; 98 when only QTN were used to create G. With 1000 QTN, trends were similar, but the rank with only QTN was 930. BLUP accuracy for validation animals was 0.32. Accuracies increased to 0.49, 0.53, 0.63, and 0.89 for ssGBLUP with unweighted 60k SNP, unweighted 60k SNP and QTN, 60k SNP and QTN with variance from GWAS, and unweighted 60k SNP and QTN with known variance, respectively. When G was constructed based only on QTN, the accuracy was 0.95 with 5% blending and 0.99 with 1% blending. Accuracies with 1000 QTN were lower, with a similar trend. Accuracies using APY G-1 were equal or higher than those with direct inverse. Single-step genomic BLUP can account for causative variants with nearly optimum accuracy when QTN and their variances are known.

International audience; This study aims at assessing the contribution of the additive and dominance genomic variances to the phenotypic expression of several purebred Piétrain and crossbred (Piétrain × Large White) pig performances. A total of 636 purebred and 720 crossbred male piglets were phenotyped for 22 traits that can be classified into the trait groups growth rate and feed efficiency, carcass composition, meat quality, behaviour, boar taint and puberty. Additive and dominance variances estimated in univariate genotypic models including additive and dominance genotypic effects and a genomic inbreeding covariate allowed us to retrieve the additive and dominance SNP variances for purebred and crossbred performances. These estimated variances were used, together with the allelic frequencies of the parental populations, to obtain additive and dominance variances in terms of genomic breeding values and dominance deviations. Estimates of additive genetic variances across traits were consistent with previous results without dominance indicating that additive and dominance genetic effects were non-confounded. Some traits showed a relevant amount of dominance genetic variance in both populations (i.e. growth rate 8%, feed conversion ratio 9-12%, backfat thickness 14-12%, lean meat 10-8%, carcass lesions 9%, in purebreds and crossbreds, respectively) or increased amounts in crossbreds (i.e. ham cut 8-13%, loin 7-16%, pH semimembranosus 13-18%, pH Longissimus dorsi 9-14%, dressing yield 5-15%, androstenone 5-13% and estradiol 6-11%). Results suggest that accounting for dominance in the models of these traits could lead to an increased GEBV accuracy and that using crossbred information can be beneficial to evaluate purebred candidates to selection for crossbred performance. Further research will compare additive and dominance marker effects between crossbred and purebred performances.

International audience; Dominance genetic effect is rarely included in pedigree-based genetic evaluation, because large-scale datasets including a high proportion of full sibs are needed to estimate dominance effects accurately. With the availability of SNP markers and the development of genomic evaluation, estimates of dominance effects have become feasible. Usually, studies involving additive and dominance effects have ignored the relationships between them. However, mating in an inbred population will produce deviations from Hardy-Weinberg equilibrium, and these deviations generate correlations between breeding values and dominant deviations. Also, it has been often suggested that the magnitude of additive and dominance effects at the QTLs are related, but there is no existing applicable approach accounting for such correlation. Wellmann and Bennewitz showed that magnitudes of additive and dominant genetic effects in quantitative trait loci are related by the dominance coefficients δ=d/|a|, and they put forwarded two ways of directional relationships between additive and dominance effects, such as cor(|a|,δ)=0 (BayesD2) and cor(|a|,δ)>0 (BayesD3). Nevertheless, these relationships cannot be fitted in individual scales in the animal model and they are not compatible with standard animal breeding software. In this study, we present a simple way of fitting the correlation between genotypic additive and dominant effects in individual scales by using a combined additive and dominance relationship matrix computed from marker genotypes. Then through a simulation study, we show that such correlations can easily be estimated by animal breeding software and accuracy and unbiasedness of prediction for genetic values would be significantly improved if such correlations are used in GBLUP.

XVII Jornadas sobre Producción AnimalISBN: 978-84-697-3065-2; This study aims at assessing the contribution of the additive and dominance genomic variances to the phenotype expression of several purebred Piétrain and crossbred (Piétrain x large White) pig performances. A total of 636 purebred and 720 crossbred male piglets were phenotyped for 22 traits that can be classified into growth rate and feed efficiency, carcass composition, meat quality, behavior, boar taint and puberty groups of traits. Additive and dominance variances estimated in univariate genotypic models including additive and dominance genotypic effects allowed us to retrieve the additive and dominance SNP variances. These ones were used, together with the allelic frequencies of the parental populations, to obtain additive and dominance variances in terms of genetic breeding values and dominance deviations. Some traits showed relevant amount of dominance genetic variance in both populations (i.e. backfat thickness, pH) or increased amount in crossbreds (i.e. ham cut, loin and dressing yield) suggesting that accounting for dominance in the models of these traits could lead to an increased GEBV accuracy and that using crossbred information can be beneficial to evaluate purebred candidates to selection for crossbred performance.