Biological processes emerge from the coordinated actions of agents such as molecules in cellular processes, host-pathogen interactions, or individual cells in tissue organization and neuronal functions. The interactions between molecules or cells and their spatial organization can be observed using advanced fluorescence microscopy techniques. Super-resolution microscopy enables visualization of single molecules, while multiplexed imaging allows precise mapping of cell types in situ. These advancements deepen our understanding of biological processes and provide new biomarkers for disease and potential therapeutic targets. However, high-resolution multimodal observations generate large, heterogeneous datasets, such as time-lapse imaging of molecules or long-term neuronal activity.
Our research group focuses on developing innovative image analysis techniques and statistical models to characterize relationships between biological components, identify patterns in complex environments, and uncover their functional implications. Our research project has two main objectives:
Develop innovative tracking algorithms utilizing deep learning to track multiple objects in time-lapse imaging.
Create robust statistical frameworks to detect significant spatial and temporal relationships between biological objects and examine their functional implications.
Mapping molecular assemblies in synaptic buttons with 3D STORM and SODA (Image Lydia Danglot, IPNP)
To fully understand protein functions and the intricate molecular networks they form, it’s essential not only to achieve precise detection and tracking of labeled molecules but also to determine their exact localization and spatial proximity. Recent technological advancements in single-molecule labeling and super-resolution multicolor microscopy have revolutionized our ability to observe small molecular clusters that were previously undetectable with conventional microscopy. This capability raises a critical challenge: distinguishing between coincidental proximity—where molecules appear close due to random distribution—and genuine molecular associations. Accurately quantifying these spatial relationships is crucial for identifying meaningful patterns, such as the formation of heterogeneous molecular assemblies. These assemblies are key to understanding how proteins interact in cellular processes, but their detection requires careful analysis to ensure that observed proximity reflects actual functional relationships rather than mere chance.
At the tissue level, recent advancements in digital pathology, such as multi-color immunohistochemistry and multiplex imaging, have significantly enhanced our ability to map the intricate organization of cells within tissues. When coupled with automated image analysis, these techniques offer valuable insights into key areas of research, including cancer biology, immunology, and neuroscience. The integration of spatiotemporal cell mapping enables a more detailed examination of cellular dynamics within complex tissue environments. Understanding the spatial relationships between cells and tissue regions has proven to be a crucial factor in predicting patient outcomes. However, a comprehensive method that effectively links localized cell interactions with the broader tissue context remains an area of ongoing development.
Leveraging level sets to map the spatial relationships between different types of microglial cells in fetal human brain sections.
Main Publications
Lagache, T., Grassart, A., Dallongeville, S., Faklaris, O., Sauvonnet, N., Dufour, A., … & Olivo-Marin, J. C. (2018). Mapping molecular assemblies with fluorescence microscopy and object-based spatial statistics. Nature communications, 9(1), 698.
Perochon, T., Krsnik, Z., Massimo, M., Ruchiy, Y.,Lagache ,T*,Manassa D.* & Holcman, D.* (2025). Unraveling microglial spatial organization in the developing human brain with DeepCellMap, a deep learning approach coupled with spatial statistics. Nature Communications, 16(1), 1577. (* co-corresponding authors)
Links to videos:
A general introduction to Spatial Statistics (Course @ ENS Paris March 2025 :
To understand the origins of disease, we must explore host physiology as a metaorganism—a dynamic system shaped by the interaction and coevolution of host biology, microbes, and the environment. Specifically, my goal is to understand how molecular factors and microbiota influence the homeostasis of neuronal networks and related behaviors in Hydra.
Central to this is decoding the neural code—mapping the neural substrates that drive behavior. In my team, we focus on the small cnidarian Hydra vulgaris, which has a primitive and relatively simple nervous system that could potentially be fully understood. The nervous system consists of 200-2,000 neurons (depending on the size of the animal) that belong to only eleven cell types, organized into two nerve nets without cephalization or ganglia. Hydra exhibits a well-characterized behavioral repertoire, which has been categorized and quantified using machine learning techniques. This repertoire includes simple movements like contractions, twists, and elongations, as well as more complex fixed-action patterns, such as feeding, locomotion through somersaulting and inch-worming, and even some learning paradigms [6]. As a polyp, the Hydra possesses remarkable regenerative capabilities, making it an ideal subject for studying neurodevelopment and the impact of molecular or genetic perturbations on neural connectivity.
From an experimental standpoint, Hydra’s transparency and small size make it an ideal model for microscopy, with nearly all of its neurons suitable for high-speed confocal imaging using calcium indicators. We developed a GCaMP-expressing Hydra colony, and calcium imaging is performed at the institute’s imaging platform.
Although primitive, the molecular toolkit and functional organization of the Hydra nervous system mirror those of more evolved bilaterian organisms. In particular, Hydra’s nerve net activity is organized into coactive neuronal ensembles that can be considered as functional units, similar to the mammalian cerebral cortex. They represent fundamental building blocks, or the “alphabet” of the neural code. The main neuronal ensembles in Hydra are rhythmically activated, providing a useful model for studying central pattern generators and their modulation by neuropeptides and the animal microbiome. This rhythmic activity is particularly relevant in the context of human physiology, as recent studies have shown that disturbances in the gut microbiota can disrupt pacemaker rhythmicity and gut motility, leading to gastrointestinal conditions like irritable bowel syndrome.
To develop a comprehensive framework that would link neuronal activity to behavior, we focus on three main research axis:
Robust, long-term monitoring of single-neuron activity in behaving Hydra
Development of a statistical framework to relate single-neuron activity to the neural substrates underlying behavior.
Creation of an integrated mathematical model and simulation tools that combine imaging and behavioral data to explore how environmental factors (such as light, salinity, etc.) and molecular factors, such as neuropeptides, modulate the functional organization of Hydra’s neural network and behavior.
This approach will not only deepen our understanding of Hydra’s neural function but also contribute to the broader field of integrating and elucidating the influence of various factors on neural development and homeostasis.
Main Publications
Hanson, A., Reme, R., Telerman, N., Yamamoto, W., Olivo-Marin, J. C., Lagache, T.*, & Yuste, R.* (2024). Automatic monitoring of neural activity with single-cell resolution in behaving Hydra. Scientific Reports, 14(1), 5083. (*: co-corresponding authors)
Lagache, T., Hanson, A., Pérez-Ortega, J. E., Fairhall, A., & Yuste, R. (2021). Tracking calcium dynamics from individual neurons in behaving animals. PLoS computational biology, 17(10), e1009432.
