Résumé
Docteur en Biologie des Systèmes.
Spécialités:
Approches quantitatives pour la Biologie des Systèmes: modèles mathématique (systèmes dynamiques, inférence et identifiabilité paramétriques, notions connexes) et statistiques (sélection de modèles, modèles à effets mixte)
Bioinformatique: analyse de données et simulations (python)
Expérience professionnelle
Doctorant en biologie
ECOLE NORMALE SUPERIEURE DE LYON , Lyon - CDD
De Septembre 2016 à Décembre 2019
Titre: Erythroid differentiation in vitro under the lens of mathematical modelling
Résumé: During the past several years, several independent studies have showed that cell differentiation is accompanied by an increase of the level of variability of gene expression. Hypothesizing that gene expression variability is a driver of cell differentiation, our group has identified three chemical drugs (called artemisinin, indomethacin and MB3) which influence both the extent of this variability and the number of differentiated cells in a culture of avian erythroid progenitors.
This thesis follows these observations, and is divided in two parts. First, we define a mathematical model of avian in vitro erythroid differentiation and confront it to our experimental data, in order to disentangle the effect of the drugs on cell proliferation, differentiation and death. Since the comparison of parameter values between the treated and untreated conditions requires precise parameter estimates, an important part of the design of our model is the identifiability of its parameters. We prove that the drugs which decrease gene expression variability (artemisinin and indomethacin) also decrease the differentiation rate of the cells, and that the drugs which increase variability (MB3) also increases the differentiation rate.
Then, observing significant variability in the outcome of our in vitro differentiation experiment, we motivate the design of a mixed effect model of erythropoiesis, in which each replicate of the experiment is an individual characterized by its own parameter values. We demonstrate the unidentifiability of this mixed effect model, and we finish by exploring ways of rendering it identifiable, using experimental design and model reduction.
website