Statistical tools for studying genetic architecture

Advances in molecular biology have revealed that most phenotypes are controlled by highly complex and interactive gene networks. This complex genetic basis is, however, not reflected in the rather simple genotype-phenotype maps postulated in population an d quantitative genetics. These models are strongly focused on statistically defined additive effects, and treat gene interactions as featureless and non-consequential noise. Recent research in population genetics has, however, made it clear that systemati c gene interactions can have profound effects on evolutionary dynamics. This has produced a need for empirical investigation of specific patterns of dynamically relevant epistasis and pleiotropy. Although there is much data that can be used for this purpo se, there is a lack of appropriate statistical methodology. In this project we seek to develop and apply statistical methods to study genetic architecture. In particular, we focus on estimating general patterns of epistasis and pleiotropy from Quantitativ e Trait Loci (QTL) analysis. Existing QTL methodology is designed to find the positions of single genes with major effects, and is not well suited to study genetic architectures with many weak interactions. We will develop statistical methods for QTL anal ysis that are aimed at estimating composite effects over many genes and gene interactions rather than the effects and positions of individual genes. We will also develop methods to utilize artificial-selection data. A time series of selection responses co ntains a lot of information about genetic architecture, but there is little formal statistical methodology available to extract this information. We plan to develop likelihood methods to test hypotheses about genetic architecture from such time series. Th e methods developed in this project will be useful for evolutionary geneticists and agricultural scientists, and may lead to new insights into the complex relationship between genotype and phenotype.