Plant breeding continuously produces new cultivars (varieties) of agricultural crops, which
are independently evaluated in field trials prior to commercialisation. Variety evaluation has
been ongoing in England and Wales since the 1930s and has been a statutory requirement in
the UK since 1963. The legacies are seed banks of historic cultivars and corresponding
phenotype databases.
The hypothesis to be tested is that this resource, with some genotyping of seed, can be
exploited in association mapping experiments. The chosen species represent those with the
most complete historic phenotype databases and the least within-variety genetic variation.
The thesis examines the population structure and cryptic relatedness in the available panels
with reference to their confounding effects on association mapping. Different analysis
methods are considered and mixed effects modelling is identified as the most reliable.
Association analysis of Mendelian and quantitative traits is conducted and the results
compared with the success rate expected from simulation. Some of the potential QTL
identified are shown to be co-incident on genetic maps with QTL identified by others in biparental
mapping studies, the observed level of co-incidence is found to be significantly
greater than would be expected by chance alone.
In conclusion the potential for association mapping of quantitative traits using this
experimental model is constrained by the number of varieties for which sufficient phenotype
data exists. In wheat the number of available mapped molecular markers is also limiting.
Simply inherited traits, however, appear amenable to study in this way.
Larger genotyping arrays are becoming available but simulation shows that larger panels (2x
for barley, 5x for wheat) of individuals will be needed if the power of the experiments is to be
sufficient to discover, for example, 80% of loci contributing ≥10% of variation. The
potential to achieve this panel size using advanced inter-crossed populations is briefly
discussed.