Description du sujet de thèse

While the molecular mechanisms of transcription are relatively well understood today, how polymerases physically find genes to transcribe among the 20,000 genes present in the nucleus remains an enigma. However, it is becoming increasingly clear that RNAP II dynamics plays a key role in the regulation of transcription, and understanding this could pave the way for new therapeutic strategies.
However, since the spatial and temporal dynamics of these highly non-stationary molecules are difficult to capture with a single instrument, it is necessary to develop a multimodal microscope combining single particle tracking (SPT) and fluorescence correlation spectroscopy (FCS). Analyzing these multimodal data is complex, as each individual technique involves sophisticated physical models that depend on numerous parameters, typically extracted using fitting methods. Furthermore, the diffusion model parameters depend on the identification of the type of diffusion process observed.

The main objective of this thesis is to exploit deep neural networks to analyze combined FCS and SPT measurements corresponding to different types of molecular displacements (free, constrained, 2D, or 3D). Our strategies will rely on recent deep learning methodologies, such as non-autoregressive transformers. We will also address the problem of signal reconstruction from the obtained trajectories, as a pretext task to improve the generalization of model training. The training of these models will initially be based on simulated data. We plan to classify the different preprocessed trajectories in correlation with the FCS maps to discriminate between the different diffusion models. We will evaluate the precision and accuracy of the spatio-temporal deep learning algorithm designed to classify the different diffusion models (Brownian, CTRW, fBm, Lw, etc.).

Contexte de travail

The Interdisciplinary Carnot Laboratory of Burgundy (ICB) is a Joint Research Unit between the CNRS, the University of Burgundy Europe, and the University of Technology of Belfort Montbéliard.

This work will be carried out at the Dijon site, in collaboration with two departments of the ICB laboratory: the Nanosciences Department and the CO2M Department, which have expertise in biophysics, microscopy techniques, analytical methods for biological imaging, computer vision, robotic vision, multimodal data analysis, and deep learning.

This thesis is part of the CAMoMill project (“Computer Assisted Multimodal Microscopy for Quantifying Molecular Diffusion in Cells”), funded by the French National Research Agency (ANR), whose objective is to analyze the dynamics of molecules within the cell nucleus.

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.