T.H.E. Meuwissen

Effective sizes of livestock populations to prevent a decline in fitness

In livestock populations, fitness may decrease due to inbreeding depression or as a negatively correlated response to artificial selection. On the other hand, fitness may increase due to natural selection. In the absence of a correlated response due to artificial selection, the critical population size at which the increase due to natural selection and the decrease due to inbreeding depression balance each other is approximately D/2σwa2, where D=the inbreeding depression of fitness with complete inbreeding, and σwa2=the additive genetic variance of fitness. This simple expression agrees well with results from transmission probability matrix methods. If fitness declines as a correlated negative response to artificial selection, then a large increase in the critical effective population size is needed. However, if the negative response is larger than the response to natural selection, a reduction in fitness cannot be prevented. From these results it is concluded that a negative correlation between artificial and natural selection should be avoided. Effective sizes to prevent a decline in fitness are usually larger than those which maximize genetic gain of overall efficiency, i.e., the former is a more stringent restriction on effective size. In the examples presented, effective sizes ranged from 31 to 250 animals per generation.

Using the genomic relationship matrix to predict the accuracy of genomic selection

Estimated breeding values (EBVs) using data from genetic markers can be predicted using a genomic relationship matrix, derived from animals genotypes, and best linear unbiased prediction. However, if the accuracy of the EBVs is calculated in the usual manner (from the inverse element of the coefficient matrix), it is likely to be overestimated owing to sampling errors in elements of the genomic relationship matrix. We show here that the correct accuracy can be obtained by regressing the relationship matrix towards the pedigree relationship matrix so that it is an unbiased estimate of the relationships at the QTL controlling the trait. This method shows how the accuracy increases as the number of markers used increases because the regression coefficient (of genomic relationship towards pedigree relationship) increases. We also present a deterministic method for predicting the accuracy of such genomic EBVs before data on individual animals are collected. This method estimates the proportion of genetic variance explained by the markers, which is equal to the regression coefficient described above, and the accuracy with which marker effects are estimated. The latter depends on the variance in relationship between pairs of animals, which equals the mean linkage disequilibrium over all pairs of loci. The theory was validated using simulated data and data on fat concentration in the milk of Holstein cattle.

Genome-wide marker-assisted selection combining all pedigree phenotypic information with genotypic data in one step: An example using broiler chickens

Data of broiler chickens for 2 pure lines across 3 generations were used for genomic evaluation. A complete population (full data set; FDS) consisted of 183,784 and 164,246 broilers for the 2 lines. The genotyped subsets (SUB) consisted of 3,284 and 3,098 broilers with 57,636 SNP. Genotyped animals were preselected based on more than 20 traits with different index applied to each line. Three traits were analyzed: BW at 6 wk (BW6), ultrasound measurement of breast meat (BM), and leg score (LS) coded 1 = no and 2 = yes for leg defect. Some phenotypes were missing for BM. The training population consisted of the first 2 generations including all animals in FDS or only genotyped animals in SUB. The validation data set contained only genotyped animals in the third generation. Genetic evaluations were performed using 3 approaches: 1) phenotypic BLUP, 2) extending BLUP methodologies to utilize pedigree and genomic information in a single step (ssGBLUP), and 3) Bayes A. Whereas BLUP and ssGBLUP utilized all phenotypic data, Bayes A could use only those of the genotyped subset. Heritabilities were 0.17 to 0.20 for BW6, 0.30 to 0.35 for BM, and 0.09 to 0.11 for LS. The average accuracies of the validation population with BLUP for BW6, BM, and LS were 0.46, 0.30, and <0 with SUB and 0.51, 0.34, and 0.28 with FDS. With ssGBLUP, those accuracies were 0.60, 0.34, and 0.06 with SUB and 0.61, 0.40, and 0.37 with FDS, respectively. With Bayes A, the accuracies were 0.60, 0.36, and 0.09 with SUB. With SUB, Bayes A and ssGBLUP had similar accuracies. For traits of high heritability, the accuracy of Bayes A/SUB and ssGBLUP/FDS were similar, and up to 50% better than BLUP/FDS. However, with low heritability, ssGBLUP/FDS was 4 to 6 times more accurate than Bayes A/SUB and 50% better than BLUP/FDS. An optimal genomic evaluation would be multi-trait and involve all traits and records on which selection is based.

The origin of selection signatures on bovine chromosome 6

The extent and pattern of linkage disequilibrium (LD) between closely spaced markers contain information about population history, including past population size and selection history. Selection signatures can be identified by comparing the LD surrounding a putative selected allele at a locus to the putative non-selected allele. In livestock populations, locations of selection signatures identified in this way should be correlated with QTL affecting production traits, as the populations have been under strong artificial selection for these traits. We used a dense SNP map of bovine chromosome 6 to characterize the pattern of LD on this chromosome in Norwegian Red cattle, a breed which has been strongly selected for milk production. The pattern of LD was generally consistent with strong selection in regions containing QTL affecting milk production traits, including a strong selection signature in a region containing a mutation known to affect milk production. The results demonstrate that in livestock populations, the origin of selection signatures will often be QTL for livestock production traits, and illustrate the value of selection signatures in uncovering new mutations with potential effects on quantitative traits.

The unified approach to the use of genomic and pedigree information in genomic evaluations revisited.

Previous proposals for a unified approach for amalgamating information from animals with or without genotypes have combined the numerator relationship matrix A with the genomic relationship G estimated from the markers. These approaches have resulted in biased genomic EBV (GEBV), and methodology was developed to overcome these problems. Firstly, a relationship matrix, G(FG) , based on linkage analysis was derived using the same base population as A, which (i) utilizes the genomic information on the same scale as the pedigree information and (ii) permits the regression coefficients used to propagate the genomic data from the genotyped to ungenotyped individuals to be calculated in the light of the genomic information, rather than ignoring it. Secondly, the elements of G were regressed back towards their expected values in the A matrix to allow for their estimation errors. These developments were combined in a methodology LDLAb and tested on simulated populations where either parents were phenotyped and offspring genotyped or vice versa. The LDLAb method was demonstrated to be a unified approach that maximized accuracy of GEBV compared to previous methodologies and removed the bias in the GEBV. Although LDLAb is computationally much more demanding than MLAC, it demonstrates how to make best use the marker information and also shows the computational problems that need to be solved in the future to make best use of the marker data.

Revealing genetic relationships between compounds affecting boar taint and reproduction in pigs

Boar taint is characterized by an unpleasant taste or odor in intact male pigs and is primarily attributed to increased concentrations of androstenone and skatole and to a lesser extent by increased indole. The boar taint compounds skatole and indole are produced by gut bacteria, metabolized in the liver, and stored in the fat tissue. Androstenone, on the other hand, is synthesized in the testis along with testosterone and estrogens, which are known to be important factors affecting fertility. The main goal of this study was to investigate the relationship between genetic factors involved in the primary boar taint compounds in an attempt to discover ways to reduce boar taint without decreasing fertility-related compounds. Heritabilities and genetic correlations between traits were estimated for compounds related to boar taint (androstenone, skatole, indole) and reproduction (testosterone, 17β-estradiol, and estrone sulfate). Heritabilities in the range of 0.47 to 0.67 were detected for androstenone concentrations in both fat and plasma, whereas those for skatole and indole were slightly less (0.27 to 0.41). The genetic correlations between androstenone in plasma and fat were extremely high (0.91 to 0.98) in Duroc and Landrace. In addition, genetic correlations between androstenone (both plasma and fat) and the other sex steroids (estrone sulfate, 17β-estradiol, and testosterone) were very high, in the range of 0.80 to 0.95. Furthermore, a genome-wide association study (GWA) and a combined linkage disequilibrium and linkage analysis (LDLA) were conducted on 1,533 purebred Landrace and 1,027 purebred Duroc to find genome regions involved in genetic control of the boar taint compounds androstenone, skatole, and indole, and sex hormones related to fertility traits. Up to 3,297 informative SNP markers were included for both breeds, including SNP from several boar taint candidate genes. From the GWA study, we found that altogether 27 regions were significant at a genome-wide level (P < 0.05) and an additional 7 regions were significant at a chromosomal level. From the LDLA study, 7 regions were significant on a genome-wide level and an additional 7 regions were significant at a chromosomal level. The most convincing associations were obtained in 6 regions affecting skatole and indole in fat on chromosomes 1, 2, 3, 7, 13, and 14, 1 region on chromosome 6 affecting androstenone in plasma only, and 5 regions on chromosomes 3, 4, 13, and 15 affecting androstenone, testosterone, and estrogens.

Genomic selection for maternal traits in pigs

The aim of this study was to compare alternative designs for implementation of genomic selection to improve maternal traits in pigs, with a conventional breeding scheme and a progeny testing scheme. The comparison was done through stochastic simulation of a pig population. It was assumed that selection was performed based on a trait that could be measured on females after the first litter, with a heritability of 0.1. Genomic selection increased genetic gain and reduced the rate of inbreeding, compared with conventional selection without progeny testing. Progeny testing could also increase genetic gain and decrease the rate of inbreeding, but because of the increased generation interval, the increase in annual genetic gain was only 7%. When genomic selection was applied, genetic gain was increased by 23 to 91%, depending on which and how many animals were genotyped. Genotyping dams in addition to the male selection candidates gave increased accuracy of the genomic breeding values, increased genetic gain, and decreased rate of inbreeding. To genotype 2 or 3 males from each litter, in order to perform within-litter selection, increased genetic gain 8 to 12%, compared with schemes with the same number of genotyped females but only 1 male candidate per litter. Comparing schemes with the same total number of genotyped animals revealed that genotyping more females caused a greater increase in genetic gain than genotyping more males because greater accuracy of selection was more advantageous than increasing the number of male selection candidates. When more than 1 male per litter was genotyped, and thereby included as selection candidates, rate of inbreeding increased because of coselection of full sibs. The conclusion is that genomic selection can increase genetic gain for traits that are measured on females, which includes several traits with economic importance in maternal pig breeds. Genotyping females is essential to obtain a high accuracy of selection.

Gene by environment interactions for production traits in Australian dairy cattle

Dairy farming is carried out under a wide range of production environments, including large variations in the level of feeding. Although reranking of dairy sires based on the level of feeding of their daughters has been reported, detecting the genetic mutations that cause this genotype by environment interaction has not been previously attempted. In our experiment to find genetic markers for such mutations, we selected 388 Holstein bulls from the Australian dairy bull population and genotyped them for 9,919 single nucleotide polymorphism (SNP) markers. Production data, consisting of first-lactation test-day records for milk yield, fat yield, protein yield, protein percentage, and fat percentage, from the daughters of the genotyped bulls were used to estimate the effect of each SNP, which was modeled as a regression on herd mean test-day yield, where herd mean test-day yield is a descriptor of the environment. Data were analyzed with 4 models; in 2 models, daughter records were analyzed directly, with and without taking sire relationships into account. With the other 2 models, sire reaction norms for each trait were calculated and marker effects on the sire reaction norms were estimated with and without taking sire relationships into account. The results showed that using daughter records directly and accounting for sire relationships was necessary to obtain high power and to limit the number of false positives. With this approach, SNP with significant effects were found for all traits. Log transformation of the data did not affect the power of gene detection. The significant markers were categorized according to their joint effects on production and environmental sensitivity. Potential gene candidates and application of the markers is discussed. About one-third of the significant markers affect intercept and slope in opposite directions, and some of these facilitate marker-assisted selection for robustness.

Incorporating Desirable Genetic Characteristics From an Inferior Into a Superior Population Using Genomic Selection

Resistance to specific diseases may be improved by crossing a recipient line with a donor line (a distantly related strain) that is characterized by the desirable trait. However, considerable losses in the total merit index are expected when crossing recipient and donor lines. Repeated backcrossing with the recipient line will improve total merit index, but usually at the expense of the newly introgressed disease resistance, especially if this is due to polygenic effects rather than to a known single major QTL. This study investigates the possibilities for a more detailed introgression program based on marker-trait associations using dense marker genotyping and genomic selection. Compared with classical selection, genomic selection increased genetic gain, with the largest effect on low heritability traits and on traits not recorded on selection candidates (due to within-family selection). Further, within a wide range of economic weights and initial differences in the total merit index between donor and recipient lines, genomic selection produced backcrossed lines that were similar or better than the purebred lines within three to five generations. When using classical selection in backcrossing schemes, the long-term genetic contribution of the donor line was low. Hence, such selection schemes would usually perform similarly to simple purebreeding selection schemes.

DECISION RULES AND VARIANCE OF RESPONSE IN BREEDING SCHEMES

Selection decisions in breeding schemes can involve choices between candidates evaluated to different accuracies. A Bayesian framework is put forward for choosing among candidates, and it is shown that attaching loss functions for estimation errors makes this process different from selecting upon best linear unbiased predictions alone. Examples are given using both linear and quadratic loss to show that when estimation errors are penalized, the selection process tends to select more unrelated and more accurately evaluated individuals. In a dairy cattle breeding scheme response was only slightly lower than that from selection on expected breeding values but with a much reduced coefficient of variation. However, if prediction errors are preferred, with the hope of selecting individuals whose breeding value are higher than expected, extra genetic progress could be obtained by favouring the selection of individuals with low accuracy. This requires consideration of more than a single generation. With discrete generations and equal measurements on candidates the decision framework was shown to be equivalent to a single quadratic restriction on the selection scores of parents in the previous generation. A framework based on Bayes decision theory could be simply applied to produce a flexible means for producers to select according to their individual risk preferences.