Philippe Roudot (I2M)

Dynamic point cloud modeling for automated mining of molecular processes in 3D live cell microscopy


2021 - present | CENTURI group leader

2014 - 2020 | Human Frontier Postdoctoral fellowship, Unsupervised discovery of intracellular processes through quantitative 3D microscopy - UT Southwestern Medical Center

2010 - 2014 | PhD Lifetime estimation of moving vesicles in frequency-domain fluorescence lifetime imaging - Inria Rennes

2008 - 2010 | Master’s degree in Computer Science - INSA de Lyon


Luminy campus, Marseille (France)

About his research

Our research focuses on the automatic identification of collective molecular processes in biomolecular clouds, establishing the acquisition parameters required to detect them and their exploration in bioimaging datasets of ever-growing size. This is a challenging problem because molecular clouds are dense and composed of a mixture of multiple concurrent processes that must be detected and tracked independently (e.g. membrane motions, signal events, chromosome capture…). The large number of possible associations makes their identification a NP-hard problem. At the same time, the local dynamics are represented by the combination of a stochastic component (e.g. tethered diffusion), a functional component (e.g. nucleation, turnover) and a structural component (e.g. membrane or embryo deformation,). To tackle these challenges, we are developing new approaches dor i) multiscale dynamic modeling for biomolecular clouds using both conventional stochastic and deep neural network modeling, ii) efficient optimization approaches to identify them into collective processes and iii) new software for efficient rendering and data access.

Open position