Ajar Ayijel

Title: Insight into the physics of DNA nanostar hydrogels

Abstract: DNA nanostars are self-assembled structures formed from multiple single-stranded DNA strands. To form a hydrogel, the nanostars arms have overhanging sticky-ends, that facilitate percolation across the gel network. Through precise sequence design, the rheological and thermodynamic properties of the resulting gel can be tuned and even predicted qualitatively. While previous studies have demonstrated the modulation of rheological properties, the underlying mechanisms governing these effects remain less understood. In this study, we aim to understand the physical properties of DNA hydrogels through a multi-technique approach. We present our current understanding of the thermodynamics and rheology of these gels, supported by experimental results.

Guillaume Faye-Bédrin

Title: Generative models describe spontaneous brain dynamics across timescales.

Abstract: Spontaneous neural activity is not stochastic noise, but a highly structured dynamical exploration of a set of recurrent motifs. Comparing these dynamics across individuals requires models capable of disentangling shared spatial patterns from individual-specific temporal trajectories. Here, we develop a generative modeling framework to decode spontaneous whole-brain calcium imaging in larval zebrafish. We first use latent-aligned Restricted Boltzmann Machines (LaRBMs) to coarse-grain neuronal activity into a shared latent space of cell assemblies. To capture the temporal grammar, we integrate this spatial representation with Hidden Markov Models (RBM-HMM), creating a predictive model of brain state transitions.

This unsupervised framework reveals a common structure of spontaneous dynamics across individuals and provides a quantitative basis for how brain assemblies organize into discrete, metastable states. It also offers a foundation for predicting how sensory inputs or targeted perturbations may reshape the brain's transition landscape. 

Séminaire réserervé aux membres du LJP.