Show simple item record

Genomic and conventional evaluations for fertility traits in pigs

dc.contributor.advisorSimianer, Henner Prof. Dr.
dc.contributor.authorFangmann, Anna Maria
dc.date.accessioned2019-01-17T10:59:13Z
dc.date.available2019-01-17T10:59:13Z
dc.date.issued2019-01-17
dc.identifier.urihttp://hdl.handle.net/11858/00-1735-0000-002E-E55D-B
dc.identifier.urihttp://dx.doi.org/10.53846/goediss-7218
dc.language.isoengde
dc.rights.urihttp://creativecommons.org/licenses/by-nc-nd/4.0/
dc.subject.ddc630de
dc.titleGenomic and conventional evaluations for fertility traits in pigsde
dc.typedoctoralThesisde
dc.contributor.refereeSimianer, Henner Prof. Dr.
dc.date.examination2018-11-06
dc.description.abstractengThe aim of genomic selection (GS) is to predict breeding values with high accuracy for young animals (without own phenotypic record) as early as possible. GS can increase the accuracy of the breeding values at the time point of selection, but often the number of available animals for the reference set within an organization (subpopulation) is the limiting factor. One possibility to overcome this problem is to enlarge the reference population by combining closely (or distantly) related subpopulations within a breed, a so called multi-subpopulation reference population. The assessment of predictive ability of genomic breeding values when using single- and multi-subpopulation references sets within a breed for the trait number of piglets born alive (NBA) was conducted in Chapter 2. Furthermore, a comprehensive comparison of different genomic relationship matrices (partly accounting for subpopulation structures) was investigated to assess their usefulness for multi-subpopulation approaches. Superiority of multi-subpopulation predictions in pigs compared to within-subpopulation predictions turned out to be rather small. Although predictions were performed within one breed (i.e. Large White), but different subpopulations, no increase or even a decrease in predictive ability was observed. Anyway, closely related subpopulation reference sets performed better than distantly related subpopulation reference sets. Despite the low differentiation of the subpopulations (low FST-values), the genetic connectedness between different subpopulations seems to be too small to improve the prediction accuracy by using multi-subpopulation reference sets, which may be caused by the separate breeding work of different German pig breeding organizations and have led to stratified subpopulations within the breed German Large White. The consideration of possible substructures through the use of different genomic relationship matrices in genomic estimations was also only partially successful. For practical application, resources of pig breeding companies should be used genotyping animals (boars and sows) within organization to create a sufficient large reference population which should be updated continuously. Since GS is considered to be state-of-the-art in animal breeding, a comprehensive comparison of different genomic models, multi- and single-step, was performed for NBA and two breeds (German Landrace and German Large White) in Chapter 3. Multi-step methods consist of mainly three parts with many parameters and multiple assumptions: (i) constructing of a response variable for genotyped animals that integrate all phenotypic information, (ii) exploiting the association between response variable and marker information through genomic prediction, and (iii) blending the genomic information with parental average. If assumptions in those steps are violated, loss of information, inaccuracies and biases may arise. One possibility to overcome these issues is the single-step method. In single-step methodology, all available information (i.e. pedigree, phenotypic and genetic) is combined within a single model. Assessment of predictive abilities for young genotyped animals indicated that both genomic methods, multi- and single-step, outperformed conventional predictions, while single-step provided higher reliabilities than multi-step. Bias was assessed by regression of corrected phenotypes on the different genomic breeding values. Predictions were less biased for single-step compared to multi-step. In general, reliabilities and predictive abilities for young animals were relatively small for both breeds, which may be caused by (i) small numbers of genotyped animals in general, (ii) rather moderate reliabilities of pseudo-observations, (iii) low numbers of genotyped progenies per boar and (iv) only few parent-offspring-links between reference and validation set. In order to potentially improve prediction accuracy and reduce bias of genomic predictions, an adjustment of G through sophisticated weighting and scaling strategies was performed. However, an increase of predictive ability through adjustment was not successful for both small empirical data sets. For practice, single-step turned out to be useful and conceptually convincing approach for NBA in moderately sized German Large White and German Landrace populations. Although GS is considered to be the preferred method, accurately estimated conventional breeding values through (consequent) performance testing along with recording phenotypes still remains one of the most important steps in the animal breeding schemes. Fertility traits such as NBA are economically important and included in most breeding schemes. In order to improve efficiency of breeding programs (and efficiency of piglet producers), traits like mothering ability (MA) of a sow, piglet survival (PS) or number of piglets weaned (NOW) from a sow have become more and more important. Therefore, knowledge of genetic parameters of fertility traits is necessary to estimate conventional breeding values accurately, to combine fertility traits in selection and to optimize breeding schemes. In Chapter 4 estimates of genetic parameters for e.g. heritability, repeatability, genetic and phenotypic variances and correlations between traits were calculated in order to evaluate an appropriate model for the routine breeding value estimation for a German pig breeding organization. The analyzed traits were: NBA, NOW, MA, PS and farrowing interval (FI). Variable selection for fixed effects was performed and different models (bivariate animal or repeatability model) were used to estimate genetic components. Genetic components were generally close to literature means, although estimated variance components strongly depended on the population structure and data set used, which made a direct comparison (of differences) difficult. However, estimates of additive genetic variance, heritability and genetic correlation indicated that the amount of genetic variation for selection was large enough to improve the traits studied. Trends observed from the data already showed an improvement of NBA (NOW) per sow and year. For the routine breeding value estimation, a bivariate animal model should be used for NBA and NOW, in which the first parity and subsequent parities should be considered as different traits. A repeatability animal model should be used for MA, PS and FI. With regard to animal welfare of sow and piglet, decreasing individual birth weights and biological limitations of reproduction performance of a sow, especially PS and MA are getting more and more important and thus should be further addressed and studied.de
dc.contributor.coRefereeBennewitz, Jörn Prof. Dr.
dc.contributor.thirdRefereeTetens, Jens Prof. Dr.
dc.subject.enggenomic selectionde
dc.subject.engpigde
dc.subject.engfertilityde
dc.identifier.urnurn:nbn:de:gbv:7-11858/00-1735-0000-002E-E55D-B-4
dc.affiliation.instituteFakultät für Agrarwissenschaftende
dc.subject.gokfullLand- und Forstwirtschaft (PPN621302791)de
dc.identifier.ppn1046911023


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record