Statistical methods for estimating intra- and inter-population variation in genetic diversity
MetadataShow full item record
- Institutt for biologi 
Knowledge of the level of (additive) genetic variation within and among populations is important for our understanding of the future direction and rate of evolutionary change. In this thesis I first investigate how evolutionary and population genetic processes influence within population levels of genetic variation and levels of genetic and phenotypic differentiation between populations. Secondly, I quantify the relative contribution of additive genetic and environmental effects to phenotypic variation in phenotypic traits. Statistical methods are developed for answering these important questions in evolutionary biology. Existing statistical methods were modified to analyse a natural system of island and mainland populations of house sparrow (Passer domesticus) along the coast of Norway. The genetic differentiation between populations was estimated by FST analysed using statistical tools to examine estimate the influence of different factors such as population type and geographic distance on the genetic population structure. The results suggested that within population level of genetic variation tended to be lower in island populations and that the level of population differentiation was higher, especially at shorter interpopulation distances. The differentiation did however increase faster with geographical distance for mainland populations than island populations. These results suggest that the genetic composition of island populations is more influenced by population bottleneck events and genetic drift than mainland populations. To identify the relative influence of genetic drift and selection in creating the observed difference in phenotypic trait means between the house sparrow populations we developed a statistical tool for comparing the genetic differentiation of a quantitative trait (QST ) and the genetic differentiation at neutral molecular markers (FST ) using phenotypic measurements. This creates a null hypothesis for identifying selection, where QST equals FST if the quantitative trait has not been under selection. However, as QST is based on additive genetic variance the estimation of QST based on phenotypic measurements is problematic. We proposed to explore the difference between QST and FST by combining two recently developed statistical methods, one that allows estimating QST based on phenotypic measurements by allowing the additive genetic variance of the phenotypic between population to vary and one that compares observed QST with neutral expectations to test the significance of the difference in QST and FST . Our results suggest that directional selection may have favoured different optimal phenotypes for body mass for both males and females and for male ornamental phenotypes. A generalized linear mixed model called animal model was used to estimate the within population genetic part of traits based a pedigree and trait data. The non-sampling based Bayesian methodology Integrated Nested Laplace Approximation (INLA) was investigated as a inference method for different versions of the animal models, with Gaussian, binomial and Poisson likelihoods. The analyses were performed for both simulation studies and house sparrow case studies, showing that the INLA methodology can be used as a fast and accurate inference method for many versions of the Bayesian animal models. We present a framework for how to evaluate the difference in Deviance Information Criteria (DIC) as a model selection criteria using simulation studies. An R package, AnimalINLA was developed for easy and fast inference for Bayesian animal models using INLA. An extended animal model was investigated that modelled both autosomal and sex-linked inheritance using INLA as inference method. The analyses were performed for both simulation studies and for a phenotypic plumage trait (diameter of spots on the ventral side) diameter trait in a Swiss barn owl (Tyto alba) population. The simulation study showed that when including both autosomal and sex-chromosome linked inheritance these two effects are identifiable. Furthermore, it was shown that difference in DIC is suitable as a model selection criteria for these models. The spot diameter was found to be substantially influenced by genes on the largest sex chromosome, but that changes in autosomal genes appeared to be underlying the phenotypic change across cohorts in this trait. The work of this thesis contributes to the ongoing research on the interaction between evolutionary processes and the level of genetic variation in natural populations. It introduces a statistical framework for performing analyses for estimating and exploring genetic variation in natural populations, which will be of value for future studies.