Lauren Borg is an undergraduate student in the Faculty of Engineering and Information Sciences at the University of Wollongong. She’s recently finished her third and final year of a Bachelor of Medical Mathematics and intends on undertaking the Honours program next year. Throughout her undergraduate degree she has been a part of the NIASRA (National Institute for Applied Statistics Research Australia) Winter (2013) and Summer (2014) Scholarship program as well as a Dean’s Scholar student. Lauren was also a member of the Dean’s Merit list in 2013. She is interested in pursuing research in the field of biostatistics.
Issues in QTL Analysis and Association Studies in Plants with High Dimensional Marker Platforms
High-throughput, high-dimensional marker platforms are now available for many species. The Affymetrix Genome-Wide Human SNP Array 6.0 has been available for some time and has almost 1 million single nucleotide polymorphisms (SNPs) in a 1.8 million marker array. The rapid advancement of these technologies has provided a range of exciting opportunities in plant breeding. Traditional plant breeding approaches have been largely based on the well established theory of quantitative genetics which produce pedigree based predictions of additive genetic value. Although there has been much progress for hundreds of years using these approaches, the use of high density molecular markers data opens possibilities to better understand the genetic architecture of complex quantitative traits of commercial interest as well as deriving accurate prediction of genetic merit without the need to phenotype large numbers of individuals in the early stages of selection. There is a need to examine and improve the statistical methods that make use of high-dimensional marker data for QTL analysis and association studies in plants. There are many methods availablefor the analysis of QTL; it is not possible to review all the methods so a selection of key papers are discussed. Verbyla et al. (2007) developed a whole genome approach called WGAIM, which consisted of a forward selection approach for determining putative QTLs, which was embedded within the framework of a linear mixed model (lmm). This lmm included terms to account for the design of the experiment, as well as terms for both full additive and polygenic effects while undertaking the search. Verbyla et al. (2012) extended the WGAIM approach to make it more suitable for high-dimensional marker data. An important feature of both approaches is the useof a fully-efficient one-stage analysis. Many other widely used approaches for QTL analysis and association studies use a two-stage approach. In the first stage summarised or meta-data is formed from an analysis (often simply averaging of raw values for each line) ignoring the marker data. This meta-data is then subjected to a QTL analysis or association study in the second stage with incorporation of measures of statistical uncertainty. In this project we will undertake a comprehensive simulation study to examine the loss in efficiency resulting from a two-stage approach. Our study will be based on a real data-set taken from experiments involving a mapping population in wheat.