Le jeudi 18 juin 2026, Raphaël LA ROCCA présentera l'examen en vue de l’obtention du grade académique de Docteur en Sciences (Collège de doctorat en Chimie) sous la direction de Loïc QUINTON et Gauthier EPPE.
Cette épreuve consistera en la défense publique d’une dissertation intitulée :
« High-Resolution Mass Spectrometry imaging beyond visualization: Addressing analytical challenges through computational methods ».
Le Jury sera composé de :
Mme A.-S. DUWEZ (Présidente), Mme et MM. D. BAURAIN, G. EPPE (Co-promoteur), I. FOURNIER (Université de Lille), L. QUINTON (Promoteur), S. RIGALI (Secrétaire), D. TOUBOUL (Ecole Polytechnique Paris).
Mass spectrometry imaging (MSI) uniquely links molecular composition to spatial context across tissues, microbial systems, and cell cultures. These rich, high-resolution mass signals are turned into biological insight by matching each detected mass to a molecular formula in a database.
This process, however, is limited by three bottlenecks: (i) pixel-to-pixel mass shifts that weaken identification confidence; (ii) limited database coverage, which leaves a large proportion of the spectrum unannotated and therefore uninterpretable; and (iii) a spatial-resolution gap between MSI and microscopy that confines interpretation to large structures and blunts insight at the cellular scale.
This thesis addresses these bottlenecks with an integrated informatics workflow. First, we stabilize spectra through adaptive, pixel-wise recalibration that restores mass accuracy and increases the number of FDR-controlled annotations across diverse MSI datasets. Second, we tackle signal complexity by exploiting high-resolution mass features: we begin with Kendrick-based filtering and then generalize it into a graph of mass differences that clusters peaks into coherent molecular families without any database matching. Beyond unsupervised clustering, we further show that this same graph representation supports supervised segmentation of MSI images while remaining insensitive to mass shift and signal mixing, and that it can isolate the molecular families specific to spatial regions. Third, we bridge MSI and microscopy through deep image fusion that predicts ion distributions at the microscope scale, improving interpretability at the near-cellular level.
Taken together, the contributions of this thesis form a coherent workflow for extracting biologically meaningful information from raw MSI data, moving the analysis from manual curation toward robust, scalable interpretation.
La thèse est intitulée : « Milankovitch Forcing and the Timing and Pacing of Devonian Anoxic Events ».
La thèse est intitulée : « Structural Phase Transitions and Complex Polar Orderings in Oxide Perovskites: Insights and Control from First- and Second-Principles ».
La thèse est intitulée : « Improving Representation of Land-Atmosphere Interactions Using Remote Sensing and Climate Models: Insights from Model Coupling and Satellite Data Integration ».
