Inference of natural selection on quantitative traits

The concept of evolution, which was introduced by Charles Darwin in 1859, and also its mathematical description by the theory of population genetics are well-established. Population genetics describes the development of a population under the influence of mutations, creating new genetic variants, and natural selection, increasing the frequency of favorable phenotypes. Yet, the experimental verification of selective forces acting on species has proven difficult. With new experimental techniques that have been established in the field of quantitative genetics, like the sequencing of DNA or measurements of gene expression levels, it has become possible to find signs of natural selection on the level of the genome. In this thesis, I develop a statistical test based on population genetics theory that can infer lineage-specific differences in selection between multiple lines of a species. The test employs data from quantitative trait experiments and uses a log-likelihood scoring to quantify the evidence for different selective scenarios. I show that the use of multiple lines increases both the power and the scope of selection inference. Extensive numerical simulations demonstrate that the test can distinguish selection from neutral evolution as well as different scenarios of lineage-specific evolution. The principle of maximum entropy is used to derive a modified version of the selection test that accounts for the multiple testing problem arising when many traits are tested for selection at the same time. The developed test is applied to two published plant datasets and a published dataset of gene expression levels in three yeast lines. In all cases, I find signs of selection not seen with a two-line test. For the yeast dataset I find pervasive adaptation linked to stress resistance both on the level of individual genes as well as for larger gene modules consisting of several genes, like protein complexes and pathways. This adaptation signal is also reflected on the protein levels.