Deep-Learning Epithelial Tissue Mechanics (DELETE project)

Host laboratory and collaborators

Jean-François Rupprecht / CPT /


The majority of cancers (approximately 80%) come from epithelial tissues whose function is to protect the organs they encase. The objective of the internship is to build an algorithm to infer the mechanical properties of these epithelial tissues. Since the rigidity of a tissue is a clinical criterion for the recognition of pre-cancerous tissues, this algorithm could then be used in the analysis of biopsies.

Short description of the host team 

The OM (Out-of-equilibrium Mechanics) team within the Centre de Physique Théorique (Aix-Marseille University, Luminy campus) and the CENTURI Institute is made up of 3 theoretical physicists - JF. Rupprecht & 2 post-docs, N. Tizon-Escamilla & S. Lin - interested in the interface between statistical physics and biophysics. See

Proposed approach (experimental / theoretical / computational)

Our theoretical research group develops analytical and numerical models to understand the physics of living materials. In particular, we have implemented a numerical simulation tool for epithelial tissues called vertex model, whereby the simulated dynamics results from the mechanical balance between a set of representative cellular forces – namely, viscosity, interfacial tension and cellular pressure. The originality of our approach is to consider the effect of force fluctuations. Though out-of-equilibrium fluctuations are recognized as playing a crucial role in controlling whether a material is rigid and elastic or soft and fluid, the contribution of active force fluctuations has been significantly overlooked in tissue mechanics models so far. To fill this gap, we propose new out-of-equilibrium statistical physics tools that are predictive of the relation between local tissue fluctuations and global tissue deformations.

During this internship, we propose to use our fluctuating vertex model - with well-controlled mechanical properties and spectra of fluctuations simulating the behaviour of biological tissues. First, we will develop deep learning techniques to infer the mechanical properties of our simulated tissues. The second phase of the internship will consist in the exploitation of our trained deep-learning tool on physiological data (healthy versus pre-cancerous tissues).

Expected profile

We are expecting students with a Physics background, highly motivated in combining tools from statistical physics and hydrodynamics; some coding experience in Matlab will be appreciated. Applications from computer science/math students can also be considered.