Description du sujet de thèse

Machine Learning for the Volatilome in the diagnosis of respiratory infections The analysis of Volatilome data (exhaled air measured by mass spectrometer) for medical purposes poses complex questions for statistical analysis and machine learning methods. The aim of the thesis project is to propose new methods that are suitable for providing robust, relevant and explainable profiles and markers of the pathogens detected or the associated immune response, in a wide range of measurement conditions and also capable of detecting emerging pathogens. This thesis project will therefore develop statistical analysis and machine learning methods adapted to the processing of Volatilome data from the VORTEX project, with the following objectives i) to obtain robust efficiency with a limited number of data (of sizes compatible with the VORTEX medical study); ii) to extract relevant elements, on a useful and explainable scale (profiles or markers) to characterise a group or a class; iii) to be able to detect and describe new, emerging and abnormal groups, and more generally to assign a confidence score in the classification for a given profile. It is also a question of being able to adapt the method to the various conditions for obtaining data. The subject involves both the methodological study, as well as the theory and practice, of approaches that can achieve such objectives, and their application to the data that will be obtained in the VORTEX project.

Contexte de travail

The PhD student will work within the Sisyphe team (machine learning and signal and image processing team) of the Physics Laboratory at ENS Lyon, France (UMR CNRS 5672), in collaboration with the LBBE (Biostat team) and participants in the VORTEX project of the PEPR MIE (infectious and emerging diseases). The student will benefit from a stimulating environment with experts in signal processing, machine learning, graph-based data processing and data analysis in biology and medicine. The SISYPHE team (http://www.ens-lyon.fr/PHYSIQUE/teams/signaux-systemes-physique ) specialises in the development of new data processing and machine learning methods, adapted both theoretically and practically to extract relevant information from complex data sets (heterogeneous, structural, non-stationary, multi-scale, etc.). It covers several applications, some of which are specifically adapted to the analysis of medical data (for example, for the robust estimation of the replication rate during Covid-19 pandemics, in neuroscience for the study of epileptic patients, for fECG during foetal delivery, for research into infectious diseases) or biological data (for example, for the modelling of DNA replication, for the analysis of gene expression).

Le poste se situe dans un secteur relevant de la protection du potentiel scientifique et technique (PPST), et nécessite donc, conformément à la réglementation, que votre arrivée soit autorisée par l'autorité compétente du MESR.