Statistical quantification and modelling of changes in gene expression and biological processes in stem cell differentiation

• Applicant: Dr. Birte Hellwig
• Subject Area: Epidemiology and Medical Biometry/Statistics
• Term: since 2021
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 491064493

In this project, statistical methods are developed to model and biologically characterize the differentiation process from stem cells to hepatocyte-like cells. Primary human hepatocytes are an established tool in pharmacology and toxicology and have the potential to be used for transplantation in metabolic liver diseases. However, they can only be obtained from surgically resected livers. An interesting alternative is the possibility to differentiate pluripotent stem cells into hepatocyte-like cells. However, the differentiation protocols available so far are not yet able to generate HLCs that have all the characteristics of primary hepatocytes.The statistical methods for modelling the differentiation processes that are developed and evaluated in this project contribute to the understanding of the processes at the gene level with respect to different biological properties. This understanding is important to improve differentiation protocols. For method development, we have access to gene expression data of the different cell types in the differentiation process (stem cells – definitive endoderm cells - hepatocyte-like cells), both from bulk and single cell analyses.In the project, statistical solutions will be developed for four questions. For the evaluation of a differentiation protocol, it is essential to quantify the distance between two cell types in the differentiation process in the best possible way. For this purpose, new measures of distance are developed that take into account both mean differences and differential co-expression. Genes with a similar course of expression during the differentiation process are likely to be controlled by similar control mechanisms. Therefore, clustering methods will be identified that divide genes into biologically coherent groups based on the course of these genes.In bulk analyses of in vitro differentiated cells, properties of different cell types can be observed. Single cell sequencing offers the possibility to use cluster analyses to investigate whether the differentiation protocol actually leads to cell groups with different phenotype. For this purpose, we will analyze which clustering techniques find subgroups of cells that are both statistically clearly separated from each other and that can be assigned to different tissue types. The differentiation path from stem cells to hepatocytes is of great importance both from a biological point of view for the interpretation of the intermediate stages and from a statistical point of view for the quantification of progress. In contrast to modern methods of pseudotime analysis, progression models directly model the genetic events that characterize the course of differentiation. The selection of events and the choice of the model are optimized with statistical methods in this project.

• DFG Programme: Research Grants