| Popis: |
The subject area for this thesis is detection of chromosomal regions or QTL causing complex variation at the phenotypic level. In particular, the differentiation of sources of additive and non additive variation. Unlike QTL mapping using divergent or inbred lines, this study aims to explore methods within populations, facilitating direct application of techniques such as marker assisted selection. Specifically, objectives were to evaluate a linear model or variance components (VC) approach to explore the existence and magnitude of variation caused by additive, dominant and imprinted QTL segregating in general pedigrees. This has been achieved by combining extensive simulation and analysis of real commercial poultry data. Linear models were constructed to simultaneously estimate fixed, polygenic and QTL effects. Different genetic models were compared by hierarchical extension to incorporate more variance components, and likelihood ratio test statistics derived from the comparison of full with reduced or null models. A range of additive, dominant and imprinted QTL effects were simulated within two-generation poultry, pig and human type pedigrees. Effects of family size and structure on power, accuracy of variance component estimation, and distribution of the test statistic, were evaluated. Empirical thresholds were derived by simulating populations under the null hypotheses for each type of simulated pedigree and permutation analysis in real data. In the commercial poultry data, dominant and imprinted QTL effects were found for bodyweight and conformation score. Under simulation, although power to detect QTL effects was high in two-generation livestock pedigrees, considerable variation was found in power and behaviour of test statistics. Power to detect dominance was greater in pig and poultry than human type pedigrees with theoretical thresholds increasingly conservative as the number of dams per sire decreased, highlighting the need for empirical derivation of the critical test statistic. The detection of variance caused by imprinted genes and in particular estimates of variance components were also heavily dependent upon the number of sire and dam families used to estimate them. Results showed that VC analysis can be used to routinely detect genetic effects including imprinting and dominance in complex pedigrees. The work presented is the most extensive evaluation of the detection of non additive QTL using VC methods to date. Results challenge standard assumptions made about power and null distributions and show that optimal use of methodology is dependent on pedigree structure. |
