EN FR
EN FR
EPIMETHEE - 2024
EPIMETHEE - 2024

2024Activity reportProject-TeamEPIMETHEE

RNSR: 202324452H
  • Research center Inria Paris Centre
  • In partnership with:Institut Pasteur, CNRS
  • Experimental and computational approaches to probe the mind of insects
  • Domain:Digital Health, Biology and Earth
  • Theme:Computational Neuroscience and Medicine

Keywords

Computer Science and Digital Science

  • A5.2. Data visualization
  • A5.3. Image processing and analysis
  • A6.1. Methods in mathematical modeling
  • A6.2. Scientific computing, Numerical Analysis & Optimization
  • A6.5. Mathematical modeling for physical sciences
  • A9.7. AI algorithmics
  • A9.10. Hybrid approaches for AI

Other Research Topics and Application Domains

  • B1.2. Neuroscience and cognitive science
  • B2.6. Biological and medical imaging
  • B2.7. Medical devices
  • B3.5. Agronomy

1 Team members, visitors, external collaborators

Research Scientist

  • Jean-Baptiste Masson [Team leader, INSTITUT PASTEUR, Researcher]

Administrative Assistants

  • Martial Le Henaff [INRIA, from May 2024]
  • Nelly Maloisel [INRIA]

Visiting Scientists

  • Alex Barbier-Chebbah [INSTITUT PASTEUR]
  • Alexis Benichou [INSTITUT PASTEUR, from Feb 2024]
  • Alexandre Blanc [INSTITUT PASTEUR]
  • Robin Cremese [INSTITUT PASTEUR]
  • Charlotte Godard [INSTITUT PASTEUR]
  • Francois Laurent [INSTITUT PASTEUR]
  • Iwan Quemada [INSTITUT PASTEUR]

2 Overall objectives

The objective of the Épiméthée laboratory is to explore and understand the organizing principles of biological information processing, particularly focusing on how evolution shapes efficient neural networks in insects to process complex sensory signals and generate behaviors. By combining physical modeling, Bayesian inference, numerical simulations, information theory, and biological experiments, the laboratory aims to decipher the relationship between biophysical constraints and neural architectures. This interdisciplinary approach seeks to develop mathematical and software frameworks for artificial biological information processing, ultimately contributing to advancements in computational neuroscience, embodied neuroAI, and the characterization of neurodegenerative and neuroinflammatory diseases at the circuit level.

3 Research program

Our research is structured around three primary initiatives and a central application, all aimed at understanding the neural mechanisms underlying decision-making and behavior in insects, particularly Drosophila larvae. The initiatives include graph approaches to neural connectomes, physical constraints on behavior, and embodied neuroAI, with a main application focused on neurodegenerative and neuroinflammatory diseases.

Graph Approaches to Neural Connectomes

This initiative focuses on exploring generative approaches to analyze and model neural connectomes, leveraging advances in electron and light microscopy. The research aims to characterize neural connectomes at the mesoscopic scale, identifying statistically significant motifs and understanding their functional roles. By employing lossless graph compression and generative latent-space modeling, the laboratory seeks to uncover the topological features and spatial embedding of neural networks, which are crucial for understanding their function and evolution.

Physical Constraints on Behavior

The second initiative investigates how physical constraints influence the behavior and neural architecture of Drosophila larvae. By modeling larval movements at multiple scales and characterizing the physical constraints associated with motion strategies, the research aims to understand the link between neural architecture and embodiment. This involves developing a framework for modeling, estimation of local and non-local constraints, and capturing the subtle physical characteristics of the larva to understand its interaction with the environment.

Embodied NeuroAI

The embodied neuroAI initiative integrates the findings from the previous two initiatives to explore the role of embodiment in structuring the larval nervous system. The research seeks to simulate and characterize identified neural circuits, infer evolutionary changes in decision circuits, and evolve neural networks in simulated physics environments. By leveraging physical models of larvae and simulations of sensory environments, the laboratory aims to optimize circuit architecture and explore the diversity of neural architectures that can implement biological decisions.

The methodological core of the research program includes amortized inferences and simulations, statistical properties of graphs, and approximating decision-making. These methodologies are essential for developing the simulation-based inference models, graph analysis techniques, and decision-making frameworks that underpin the laboratory's research initiatives.

Software Development

The laboratory is also committed to developing software platforms, such as Nyx and Eurynomé, to support data processing, simulation, and analysis of larval experiments. These platforms will facilitate the integration of experimental data with neural activity models, enabling the study of the relationship between physical constraints and neural architecture.

4 Application domains

The lab's research has broad application domains, spanning computational neuroscience, artificial intelligence, and biomedical research. By focusing on the neural mechanisms of insects, particularly Drosophila larvae, the laboratory aims to uncover fundamental principles of neural information processing that can inform the development of advanced AI systems and neurotechnologies. The laboratory's work in graph approaches to neural connectomes provides insights into the structural and functional organization of neural networks, which can be applied to improve machine learning algorithms and neural network designs. Additionally, the study of physical constraints on behavior offers valuable knowledge for robotics and autonomous systems, where understanding the interplay between physical embodiment and neural control is crucial for developing efficient and adaptive machines. The embodied neuroAI initiative further explores these concepts by integrating biological principles into AI, potentially leading to more robust and biologically plausible AI models. Moreover, the laboratory's focus on neurodegenerative and neuroinflammatory diseases at the circuit level has significant implications for medical research, offering new perspectives on disease mechanisms and potential therapeutic strategies. By leveraging the genetic tools available in Drosophila, the laboratory can model and study these diseases in ways that are not feasible with traditional mammalian models, potentially accelerating the discovery of new treatments.

5 Social and environmental responsibility

5.1 Footprint of research activities

The Épiméthée laboratory is committed to minimizing the environmental impact of its research activities. By focusing on Drosophila larvae as a model organism, the laboratory reduces the need for larger, more resource-intensive animal models. This approach not only aligns with ethical considerations in animal research but also lowers the carbon footprint associated with maintaining and experimenting on larger animals. Additionally, the laboratory employs computational and simulation-based methods, which are inherently less resource-intensive compared to traditional wet-lab experiments.

5.2 Impact of research results

By advancing our understanding of neural information processing in insects, the laboratory contributes to the development of more efficient and biologically plausible AI systems. These advancements can lead to innovations in robotics and autonomous systems, enhancing their ability to operate in complex environments with minimal energy consumption.

Furthermore, the laboratory's focus on neurodegenerative and neuroinflammatory diseases offers promising avenues for medical research. By providing insights into disease mechanisms at the circuit level, the research may accelerate the development of targeted therapies, potentially reducing the societal burden of these diseases. Additionally, the laboratory's commitment to open-source software development and collaboration fosters a broader impact on the scientific community, promoting the sharing of knowledge and resources to drive collective progress in neuroscience and AI.

6 Highlights of the year

6.1 Awards

- Innovator price Le Point (2025)

6.2 Visible publication

Lautaro Gandara et al. ,Pervasive sublethal effects of agrochemicals on insects at environmentally relevant concentrations.Science386,446-453(2024).DOI:10.1126/science.ado0251

Summary

Our paper investigates the sublethal effects of agrochemicals on insects, focusing on how low doses of pesticides impact the behavior and physiology of Drosophila melanogaster larvae. Using a comprehensive library of 1024 agrochemicals, the study reveals that a significant proportion of these chemicals, even at environmentally relevant concentrations, alter larval behavior and compromise long-term survivability. The research highlights that sublethal doses induce widespread changes in the phosphoproteome and affect development and reproduction, with these effects being amplified at higher temperatures. The findings suggest that sublethal pesticide exposure may contribute to the global decline in insect populations, emphasizing the need for more comprehensive chemical safety assessments.

Our achievements include the development of a high-throughput screening platform to assess the impact of a wide range of pesticides on insect behavior and physiology. By demonstrating that even non-insecticide pesticides can have significant effects on larval survivability and behavior, the research underscores the importance of considering sublethal effects in pesticide regulation. The work also extends its findings to other insect species, such as mosquitoes and butterflies, indicating the broad relevance of these effects across different ecological contexts. This research provides valuable insights for improving the precision of agrochemical use and mitigating their unintended consequences on insect biodiversity.

7 New software, platforms, open data

7.1 New software

7.1.1 nyx

  • Name:
    Nyx
  • Keywords:
    3D, Finite element modelling, Machine learning, Ethology, Unsupervised learning, Statistical inference, Bayesian estimation
  • Functional Description:
    Nyx uses : -> Larva Tagger to generate classifiers of larva behaviour and annotate data -> a SOFA based finite element simulation framework of the body including muscle -> a CNS simulation framework based on the larva neural connectome
  • Contact:
    Francois Laurent

8 New results

8.1 Insects decline

Participants: JB Masson, Francois Laurent.

Lautaro Gandara et al. ,Pervasive sublethal effects of agrochemicals on insects at environmentally relevant concentrations.Science386,446-453(2024).DOI:10.1126/science.ado0251

Summary

Our paper investigates the sublethal effects of agrochemicals on insects, focusing on how low doses of pesticides impact the behavior and physiology of Drosophila melanogaster larvae. Using a comprehensive library of 1024 agrochemicals, the study reveals that a significant proportion of these chemicals, even at environmentally relevant concentrations, alter larval behavior and compromise long-term survivability. The research highlights that sublethal doses induce widespread changes in the phosphoproteome and affect development and reproduction, with these effects being amplified at higher temperatures. The findings suggest that sublethal pesticide exposure may contribute to the global decline in insect populations, emphasizing the need for more comprehensive chemical safety assessments.

Our achievements include the development of a high-throughput screening platform to assess the impact of a wide range of pesticides on insect behavior and physiology. By demonstrating that even non-insecticide pesticides can have significant effects on larval survivability and behavior, the research underscores the importance of considering sublethal effects in pesticide regulation. The work also extends its findings to other insect species, such as mosquitoes and butterflies, indicating the broad relevance of these effects across different ecological contexts. This research provides valuable insights for improving the precision of agrochemical use and mitigating their unintended consequences on insect biodiversity.

8.2 statistical testing

Participants: JB Masson, Francois Laurent, Alexandre Blanc, Chloe Barre, Alexis benichou, Christian Vestergaard.

We have developed a suite of statistical methods to detect subtle behavioral changes in Drosophila melanogaster larvae in response to neural manipulation. Understanding how the nervous system generates behavior remains a fundamental challenge in neuroscience, but existing approaches often miss subtle behavioral modulations that may have significant biological implications. We addressed this gap by analyzing an unprecedented dataset of over 280,000 larvae across 569 genetic lines, focusing specifically on responses to air-puff stimuli. Our approach integrates multiple timescales of behavioral dynamics with advanced statistical techniques to identify neurons that induce subtle but significant behavioral changes.

Our methodological innovations include four major components: a physics-informed Bayesian approach that regularizes larval shape inference across the entire dataset; an unsupervised kernel-based method for statistical testing in learned behavioral spaces to detect subtle deviations in behavior; a generative model for larval behavioral sequences that serves as a benchmark for identifying higher-order behavioral changes; and a comprehensive analysis technique using suffix trees to categorize genetic lines into clusters based on common action sequences. These approaches have significantly expanded our catalog of "hit" neurons beyond those previously identified through conventional methods. Notably, we discovered neurons that modulate behavioral responses to different stimulus intensities and others that influence higher-order sequence structure without changing individual action probabilities. These findings advance our understanding of the neural mechanisms underlying behavior generation and decision-making, demonstrating the power of sophisticated statistical approaches in revealing subtle but important biological phenomena

8.3 Behavioural choices

Participants: JB Masson, Francois Laurent, Alexandre Blanc, Chloe Barre, Alexis benichou, Christian Vestergaard.

In our recent study, we delved into the intricate neural mechanisms that underlie the adaptive defensive behaviors in Drosophila larvae, focusing on how these organisms respond to various threatening stimuli in their environment. Our approach combined advanced techniques such as neuronal manipulations, machine learning-based behavioral detection, electron microscopy (EM) connectomics, and calcium imaging to map the specific neural circuits involved in these behaviors. By leveraging the genetic tractability of Drosophila and its well-characterized nervous system, we aimed to uncover how second-order interneurons differentially influence the competition between startle and escape behaviors in response to aversive cues. This research not only sheds light on the fundamental principles of neural circuit function but also provides insights into how context-dependent modulation of behavioral responses is achieved at the neuronal level.

Our findings revealed that specific second-order interneurons play a crucial role in modulating the balance between startle and escape behaviors in Drosophila larvae. Through detailed EM connectomics and calcium imaging, we identified key interneurons, such as A19c and early-born ELs, that are differentially involved in these defensive actions. We demonstrated that mechanosensory stimulation can inhibit escape behaviors in favor of startle responses by influencing the activity of these interneurons. Furthermore, our study highlighted the role of descending neurons in promoting startle behaviors and potentially modulating the escape sequence. These results collectively underscore the complexity and flexibility of neural circuits in orchestrating context-appropriate defensive behaviors, offering a deeper understanding of the neural basis of adaptive behavior in response to environmental threats.

9 Bilateral contracts and grants with industry

9.1 Bilateral contracts with industry

- NVIDIA - support the project on foetus of the lab - will sign an MOU with pasteur (reflexion on signing something with INRIA)

-

9.2 Bilateral Grants with Industry

Participants: JB Masson, Francois Laurent, Christian Vestergaard.

- ANR with Orange - microbrain on Edge - Cifre with Biomerieux for september 2025

10 Partnerships and cooperations

10.1 International initiatives

10.1.1 Associate Teams in the framework of an Inria International Lab or in the framework of an Inria International Program

N/A

10.1.2 Inria associate team not involved in an IIL or an international program

N/A

10.1.3 STIC/MATH/CLIMAT AmSud projects

N/A

10.1.4 Participation in other International Programs

Participants: JB Masson, Christian Vestergaard, Francois Laurent.

- NeuroAI network spearheaded by Adrienne Fairhall

10.2 International research visitors

10.2.1 Visits of international scientists

N/A

Inria International Chair

N/A

Other international visits to the team
Tomoko Ohyama
  • Status
    Principal investigator
  • Institution of origin:
    McGill
  • Country:
    Canada
  • Dates:
    December 2024 - september 2025
  • Context of the visit:
    collaboration on larva research
  • Mobility program/type of mobility:
    Post tenure visiting funded by Mcgill

10.2.2 Visits to international teams

Sabbatical programme

N/A

Research stays abroad
persJBMasson
  • Visited institution:
    Janelia Research Campus
  • Country:
    USA
  • Dates:
    ongoing since 2013 (multiple weeks to multiple months depending on the years)
  • Context of the visit:
    multiple collaborations and visiting scientist status
  • Mobility program/type of mobility:
    Janelia visiting projects
persFrancoisLaurent
  • Visited institution:
    LMC cambridge
  • Country:
    UK
  • Dates:
    2014-2015
  • Context of the visit:
    Automating pipelines of larva experiments analysis
  • Mobility program/type of mobility:
    Cambridge visiting projects

10.3 European initiatives

10.3.1 Horizon Europe

N/A

10.3.2 H2020 projects

N/A

10.3.3 Digital Europe

N/A

10.3.4 Other european programs/initiatives

- EIC with the spinoff of the lAB, Avatar Medical

10.4 National initiatives

-> JBMasson handles the interdisciplinary work packages for the IHU ReConnect and IHU ICE -> JBMasson Co-leads the RHU ReBones and lead the WP5

10.5 Regional initiatives

N/A

10.6 Public policy support

-> IHDEN classes

11 Dissemination

11.1 Promoting scientific activities

11.1.1 Scientific events: organisation

-> QBIO : simulated bodies (mars 2025) -> FADEX : NeuroAI (20 janvier 2025) -> Pr[Ai]rie seminars and Mini Symposium(contribution)

General chair, scientific chair

N/A

Member of the organizing committees

->

11.1.2 Scientific events: selection

-> Neuro-pediatrie-UPC-2-decembre-2024 -> PINTS (Paris île de France Neural Theory Symposium) - 06/12/24 -> DGA -02 -12- 2024 -> JIFF-29-11-2024 -> Cours IA -senologie - 21 - november 2024 -> IHU-ICE 22 novembre -> 13-15/11/2024 : GDR neural net annual meeting -> The Guild - Paris Cite, PSL event on "AI in Universities" event - 06 November - Paris -> jfXR 2024 : Journées Françaises de la Réalité Etendue - 29 octobre -> MICCAI 2024 ocotbre 2024 -> Séminaires polytechniques Santés 3 octobre -> 16-17 sep 2024 : Physics of Living Matter 18 (Marseille) – François, Alexandre, Christian -> 28-30 of October: EWRL 2024 (Toulouse) Alex et Etienne boursier

Chair of conference program committees

N/A

Member of the conference program committees

N/A

Reviewer

-> regular reviewer of PRL, PRX, PRE, nature, Plos Comp Biology

11.1.3 Journal

N/A

Member of the editorial boards

-> Scientific report, Frontiers in bioinformatics

Reviewer - reviewing activities

-> regular reviewer of PRL, PRX, PRE, nature, Plos Comp Biology

11.1.4 Invited talks

-> 10 per year

11.1.5 Leadership within the scientific community

N/A

11.1.6 Scientific expertise

-> chief scientifif officer of Avatar Medical -> Scientific advisory board of startup Robeauté

11.1.7 Research administration

N/A

11.2 Teaching - Supervision - Juries

11.2.1 Supervision

-> PhDs: Anqi Zhou, Alexis Benichou, Charlotte Godard -> Postdoc: CHloe Barre -> HDR: Mohamed el Beheiry

11.2.2 Juries

-> 4 PhD defense -> 2 HDR defense

11.3 Popularization

AI in Healthcare Keys to Innovation While Addressing Digital Sovereignty Challenges whitepaper https://www.legifrance.gouv.fr/jorf/id/JORFTEXT000049780607 https://www.maddyness.com/uk/2025/01/31/ais-power-to-transform-healthcare-under-the-microscope-with-la-french-tech-london/ https://www.polytechnique-insights.com/tribunes/sante-et-biotech/vers-une-psychiatrie-augmentee-par-le-numerique/

11.3.1 Specific official responsibilities in science outreach structures

N/A

11.3.2 Productions (articles, videos, podcasts, serious games, ...)

N/A

11.3.3 Participation in Live events

N/A

11.3.4 Others science outreach relevant activities

N/A

12 Scientific production

12.1 Major publications

12.2 Publications of the year

International journals

Reports & preprints