Here is a list of possible topics for Bachelor thesis in 2025 in our group. These are examples, and topics might slightly evolve depending on your precise interests, and the status of these projects at the beginning of the thesis!
Upon infection, the infected cell triggers an innate immune response, by secreting cytokines, in particular interferons. This cascade is activated in the infected cell and leads to a first internal immune response. Additionally, produced interferon moleculaes are secreted out of the cell, and are detected through the interferon receptors, leading to a second wave of immune response, both in the infected cell (autocrine signalling) as well in the neighboring cells (paracrine). The kinetic of the immune respones is still unclear: how long does the first wave last? How fast is the activation of the paracrine signalling? and, when do cells return to their ground state after interferon signalling? To answer these questions, we have generated a fine-grained time-lapse genomic dataset, consisting of bulk RNA-seq data, as well as multiome single-cell data. We want to use these datasets to define an interferon clock, which will allow us to time the immune status of cells upon infection. In particular, we want to compare the transcriptomics clock to the epigenomic clock, to evaluate how much immune memory is capture by the epigenome. Finally, we want to apply this clock to patient data, e.g. from COVID19 patients.
Brain imaging is widely used as a diagnostic tool. In particular, brain MRIs are used to study changes in the brain structure or function in disorders such as mental disorders. Both structural MRI as well as functional MRI is used. In a collaboration with ZI Mannheim, we have collected a large dataset of publicly available brain MRI data covering disorders such as schizophrenia, depressive disorders and autism spectrum disorders, as well as a cohort of patients collected from ZI Mannheim. Our attention focuses on catatonia, a complex neuropsychiatric behavioral syndrome. We want to understand how the catatonia patients relate to other patients, using a deep-learning (DL) model to obtain a lower dimensional representation of patient data. Using this model, we want to understand which regions of the brain distinguish catatonic patients from non-catatonic patients using methods that allow to interpret the results of the DL models.
Single-cell genomics has uncovered a new level of insights regarding tissue heterogeneity. While scRNA-seq is now widely used, other single-cell omics data is increasingly used, to shed light on new aspects of gene regulation, such as scATAC-seq. Finally, multiomics assay allow to capture multiple signals from the very same cell (for example scRNA-seq and scATAC-seq). Using these multiomics datasets, we have built a deep-learning model, PELICAN, that allows to translate one modality into another. Hence, given scRNA-seq, the model can predict the possible scATAC-seq profiles. In this project, the student would collect a large set of available multiome data, in order to train a library of models, that could be used by researchers to translate their own scRNA-seq datasets and generate synthetic scATAC-seq. Using gene knock-outs, we want to evaluate the impact on the chromatin landscape.
Zhang, R., Meng-Papaxanthos, L., Vert, J.-P. & Noble, W. S. Semi-supervised single-cell cross-modality translation using Polarbear. 2021.11.18.467517 https://www.biorxiv.org/content/10.1101/2021.11.18.467517v1 (2021) doi:10.1101/2021.11.18.467517.
Wu, K. E., Yost, K. E., Chang, H. Y. & Zou, J. BABEL enables cross-modality translation between multiomic profiles at single-cell resolution. PNAS 118, (2021).
In cancer, malignant cells can modify their phenotype to adapt to the environment and develop resistance to drugs and treatments. This plasticity is regulated at the genetic and epigenetic level. Cells that acquire genetic mutations (e.g. gene mutations) will modify their phenotypes. Hence, it is important to anticipate these changes, and be able to predict the impact of genetic mutations on the cell. Several machine-learning approaches have been developed to make predictions of gene knock-out effects using single-cell datasets. In this project, we want to compare several of these tools and apply them on a dataset of glioblastoma cells. We want to understand how the predicted modified gene expression profile impacts the phenotype, e.g. the metabolic state of cancer cells.