Abstract
Controlling infectious diseases is a major issue for human, animal, and plant public health. In addition to traditional epidemiological data such as incidence or prevalence, genetic data resulting from advances in sequencing techniques are now available. The abundance of these data associated with the rapid evolution of pathogens, in particular RNA viruses, has led to the development of a disciplinary field called phylodynamics. One of the underlying hypotheses of this field is that the way viruses spread leaves footprints in their genomes. Phylodynamic methods take advantage of this genomic information to estimate epidemiological parameters such as the growth rates of an epidemic, the number of secondary infections caused, or the average infection duration.Bayesian phylodynamic inference methods are the most commonly used and are generally based on likelihood functions. However, they can be unsuitable for models with many parameters, e.g. those that capture transmission heterogeneity. Other inference methods are based on Approximate Bayesian Computation (ABC) and do not require a likelihood function. These approaches involve simulations from epidemiological models, summary statistics capturing epidemiological information from phylogenies, and regression techniques. So far, phylodynamic ABC approaches have focused on simple epidemiological models.This thesis work comprises two main parts. The first part consisted in developing a tool for rapid simulations of time series and phylogenies, which can be used in ABC approaches. This simulator, TiPS, has the advantage to include a large diversity of epidemiological models while being faster than many existing tools. The second part of the thesis consisted of studying epidemics in different contexts using ABC or likelihood-based approaches.The first application consisted of a phylodynamic analysis of the spread of hepatitis C virus (HCV) in Lyon. The epidemiological context led us to extend the application of the ABC method to a structured model with two types of hosts. For this purpose, we developed new summary statistics adapted to labeled phylogenies, where leaves are associated with a risk group. This ABC approach was then also used in another context to study the transmission advantage that the presence of a certain mutation would confer to HIV-1/O. The last study is devoted to epidemiological and phylodynamic analyses of SARS-CoV-2 from French data using Bayesian likelihood-based inference methods.Although in full expansion, the field of phylodynamics is poorly known and studied in France. In addition to its methodological developments and its analyses on current epidemics, this work illustrates, at a national scale, the potential of genomic sequence data analysis to help public health organisms to establish more efficient control measures, especially in times of crisis.