• Citer

Quantitative methods in immuno-oncology: deconvolution tool, anti-PD1 therapy, model of chronic myeloid leukemia

par Vande Velde, Sylvie
Président du jury Loris, Ignace
Promoteur Lenaerts, Tom
Co-Promoteur Leo, Oberdan
Publication Non publié, 2024-04-26

Thèse de doctorat

• Accès en ligne
• Détails
• Contenu
• Statistiques

Résumé :

During my PhD thesis, I mainly worked on three projects involving mathematical and bioinformatics methods in immuno-oncology. 1) Project 1: development of the SmartFACS deconvolution tool. I developed a deconvolution tool called SmartFACS in collaboration with the Mechanics and Applied Mathematics Department (ULB) and the Immunobiology Laboratory (ULB). The role of this computer program is to estimate the cellular composition of a sample from RNA sequencing data. This tool has been developed in the context of immunology, where it is often of interest to identify immune cells present in the blood or in tumors. At the time this project began, no deconvolution tool for mouse data was yet available, despite the fact that much biomedical research is carried out in mouse models. During my thesis, I therefore decided to develop a tool adapted to mouse data, while trying to improve existing tools for human data by detecting a wider range of immune populations. 2) Project 2: research of clinical biomarkers for anti-PD1 immunotherapy treatment. In collaboration with the Immunobiology group (ULB), I worked on anti-PD1 therapy, which is already used clinically for certain cancers. Unfortunately, the success rate of this treatment is still relatively moderate. This is why the aim of my project was to determine predictive markers of response to anti-PD1 in order to better understand the mechanisms of resistance and find avenues of improvement for this treatment, as well as to better identify responder patients. To achieve these objectives, a mouse model was developed in the Immunobiology Laboratory. This model enabled us to reproduce the dichotomous response observed in the clinic. My work focused on analyzing pre-treatment tumor sequencing data obtained using the mouse model, in order to compare the immune infiltrate of responder and non-responder mice. My data analyses, together with experiments conducted by Jelena Gabrilo, revealed two immune populations important for treatment response: a subpopulation of Ly6c+ MHCII+ PDL1+ macrophages and a subtype of CD8+ T cells. My work focused on characterizing the macrophage subpopulation and the signaling pathways that regulate its activity, as well as characterizing the immune infiltrate using bulk RNA-seq data generated by Dr. Coralie Henin. A comparative study was also carried out with other similar mouse models described in the literature. 3) Project 3: Mathematical modeling of chronic myeloid leukemia. The third major project of my thesis was carried out in collaboration with the University of Melbourne (Australia) and the ICHEC Brussels Management School. We have developed a stochastic model to recreate the hierarchical organization of the blood system. This system is based on stem cells located in the bone marrow. These cells differentiate into different types of progenitor cells until they give rise to mature blood cells. To recreate this dynamic, we have developed a 27-compartment model in which each compartment represents a different stage of cell differentiation. Specifically, this is a model of continuous-time multi-type branching process. We then adapted our model to the case of chronic myeloid leukemia (CML) by adding a new population of cells, the cancer cells, which are characterized by their own parameter values determined from patient data. By combining our CML model with a Bayesian approach called ABC-PMC (Approximate Bayesian Computation - Population Monte Carlo) and adapted for discrete case, we addressed the question of the origin of CML and showed that this disease starts from the stem cell compartment. This result was confirmed by an independent study (Ilhan et al, 2021). We then tackled the question of treatment discontinuation. We succeeded in showing that it is the number of cancer stem cells that is decisive, and we defined new criteria to better identify patients who can stop their treatment without risk of relapse.