Thesis topics

Recurrent selection is a strategy in plant breeding programs, aimed at improving populations through repeated cycles of selection and recombination. In such programs, multiple crosses between parental lines generate families that differ in their expected mean performance and segregation variance. How breeding resources are allocated across these families, in terms of family size could influence long-term genetic gain and the maintenance of genetic diversity.

This thesis aims to investigate the long-term impact of different family size optimization in recurrent selection using simulation-based approaches. The study will assess how alternative allocation strategies affect genetic gain, genetic variance, and population structure over multiple selection cycles.

You will be provided with relevant datasets and will implement simulations of recurrent selection schemes under different family size allocation strategies.

Target group: Master students of Nutzpflanzenwissenschaften (Crop Science), Informationstechnologie, or Agrobiotechnology.

Supervisor: Joshua Okoye (uche.okoye[at]agrar.uni-giessen.de).

Requirements: MK-002 or MK-002-EN and either MK-119-EN, MP-236-EN (would be ideal), or MP-240-EN completed successfully.

Focus: R programming, breeding theory.

Genome-wide association studies (GWAS) have become a central tool for dissecting the genetic basis of complex traits. However, the large majority of methods developed to implement GWAS have been designed with only diploid organisms in mind. The dissection of complex traits in polyploids is equally relevant, as they account for a substantial proportion of agronomic and ornamental crop production. The aim of this thesis is to compare GWAS methods specifically developed for polyploid species alongside relatively novel diploid-based models, using a tetraploid carnation dataset. Tetraploid carnation represents a particularly interesting system for this comparison, as its mode of inheritance remains poorly characterised, making the choice of an appropriate GWAS model non-trivial. Performance will be evaluated based on the ability of each method to detect marker-trait associations for key ornamental traits.

Target group: Bachelor students of Agricultural Science, and master students of Crop Sciences (Nutzpflanzenwissenschaften), Informationstechnologie, or Agrobiotechnology.

Supervisor: Hugo H. Tavera (hugo.tavera[at]agrar.uni-giessen.de).

Requirements: Bachelor: BP-041 Biostatistics, Master: MK-002 or MK-002-EN and either MK-119-EN, or MP-240-EN completed successfully.

Focus: Data management, R programming.

Literature to start with: Chen et al. 2025 Nature Plants 11.9: 1714-1728

In plant breeding, we often use statistical models to predict the traits of interest of certain genotypes based on their genetic composition, rather than growing them in the field. One possible approach is to combine several statistical models to create a larger model, known as a stacked ensembles or model stacking.

The aim of this thesis is to establish a model stacking procedure and test the hypothesis that a more diverse composition of base models results in better ensemble models, as some authors claim. Thesis can be in English or German.

Target group: Bachelor students of any programme.

Supervisor: Philipp Heilmann (philipp.g.heilmann[at]agrar.uni-giessen.de).

Requirements: BK-005 and BP-041 completed successfully. Ideally, you find programming intuitive and pick up new concepts quickly.

Focus: Programming in R or Python, Machine Learning.

Literature to start with: Heilmann et al. 2023

The effect of including epistatic relationship matrices into genomic prediction models with regard to prediction accuracy has been mixed: Some studies report increases, others do not. If epistasis is not modeled by epistatic relationship matrices but with explicit epistatic effects, it is hypothesized that due to the typically rather small populations, the effect estimation is relatively imprecise. To date, there are no studies that systematically explore what population size would be needed for accurate estimation of epistatic effects. The master thesis will consist mainly of a simulation study with SelectionTools that investigates which population sizes are needed under which conditions (e.g. heritability of the trait, LD patterns, and population structure) for precise effect estimation assuming epistasis.

Target group: Master students of Crop Sciences (Nutzpflanzenwissenschaften), Informationstechnologie, Agrobiotechnology, or Agrobioinformatics.

Supervisor: Carola Zenke-Philippi (carola.zenke-philippi[at]uni-giessen.de).

Requirements: For master students: MK-002 or MK-002-EN and either MK-119-EN (would be ideal), MP-236-EN, or MP-240-EN completed successfully.

Focus: R programming, theory of population genetics.

Literature to start with: Jiang and Reif: Modeling epistasis in genomic selection. Genetics (2015) 201:759-768.

Plant breeders are interested in a wide range of questions regarding their breeding material. Very often, answering these questions requires knowing the marker data of the genotypes in question. In order to save money, it would be helpful if "synthetic genotypes" could be created with simulations which are similar but not equal to the genotypes that breeders already have information on. Very eloborate models have been developed for this task, e.g. HAPGEN2 and HAPNEST. However, for some questions to be answered by quantitative genetics, it might be sufficient to use a relatively easy algorithm that derives marker genotypes from a combination of a genetic map and known LD patterns and relationships in a population. The focus of this master thesis will be to develop such an algorithm and test it on several available wheat, barley, and oat populations.

Target group: Master students of Crop Sciences (Nutzpflanzenwissenschaften), Informationstechnologie, Agrobiotechnology, or Agrobioinformatics.

Supervisor: Carola Zenke-Philippi (carola.zenke-philippi[at]uni-giessen.de).

Requirements: For master students: MK-002 or MK-002-EN and either MK-119-EN (would be ideal), MP-236-EN, or MP-240-EN completed successfully.

Focus: R programming, theory of population genetics.

The R package StageWise (Endelman 2023) provides different methods for the two-stage analysis of field trials: First, each environment (year/location combination) is analyses separately. In the second stage, the resulting adjusted entry means are taken as the observations of a model with the environments as an independent variable in the model. The objective of the thesis is to conduct a simulation study that uses pre-defined marker effects. Based on these, field trials are simulated. The goal is then to investigate which of the two-stage evaluation methods for the field trials results in estimates of marker effects that resemble the original values as closely as possible.

Target group: Bachelor students of Agrarwissenschaften, Master students of Crop Sciences (Nutzpflanzenwissenschaften), Informationstechnologie, or Agrobiotechnology.

Supervisor: Carola Zenke-Philippi (carola.zenke-philippi[at]uni-giessen.de).

Requirements: For Bachelor students: BK-005 and BP-041 completed successfully. For master students: MK-002 or MK-002-EN and either MK-119-EN (would be ideal), MP-236-EN, or MP-240-EN completed successfully.

Focus: R programming, theory of population genetics.

Literature to start with: Endelman 2023 Theoretical and Applied Genetics (2023) 136:65