Job offer 2023: Ph.D. Position in Computational Neuroscience – Position filled
Computational biologically realistic modeling of the hippocampal electrical activity and plasticity for the study of pathophysiological mechanisms of schizophrenia
Application website: https://adum.fr/as/ed/voirproposition.pl?site=adumR&matricule_prop=49787
Directrice de thèse : Laure Buhry (MCF- HDR – LORIA) – laure.buhry@loria.fr
Co-directeur de thèse : Radu Ranta (MCF – HDR – CRAN) – radu.ranta@univ-lorraine.fr
Co-encadrant : Jérémie Gaidamour (IR CNRS – IECL) – jeremie.gaidamour@univ-lorraine.fr
Mots-clefs : neurosciences computationnelles, modélisation mathématique, schizophrénie, séparation de sources, modèle animal
Contexte et état de l’art
Les troubles schizophréniques sont des troubles psychiatriques qui affectent environ 1% de la population mondiale. D’après le DSM-V [1], ils sont caractérisés par un ensemble de symptômes comportementaux et émotionnels décrits comme positifs (hallucinations, etc.), négatifs (anhédonie, etc.) et cognitifs (affectant apprentissage, mémoire, attention et prise de décision, etc.). Alors que les symptômes positifs sont généralement contrôlés par les antipsychotiques, les symptômes négatifs et les déficits cognitifs ne sont que mal pris en charge par ces traitements pharmacologiques [2]. L’un des principaux obstacles au développement de thérapies efficaces reste notre compréhension limitée des mécanismes physiopathologiques sous-jacents. Afin de pallier cette lacune, des modèles animaux de schizophrénie sont développés depuis plusieurs années, mais ils ne permettent pas de répondre seuls aux interrogations. Il n’est en effet pas encore possible de dissocier totalement les propriétés cellulaires, synaptiques des réseaux de neurones impliqués dans la pathologie.
Des études récentes ont mis en évidence des perturbations des propriétés synaptiques hippocampiques dans les modèles animaux validés de schizophrénie [3] en comparaison à des souris sauvages. Ces perturbations pourrait expliquer en partie les troubles cognitifs retrouvés chez les patients, en particulier ceux impliquant la mémoire. Les propriétés en questions impliquent notamment la balance excitation-inhibition avec un rôle crucial joué par les récepteurs NMDA et GABA. De plus, des études immunologiques et génétiques, chez l’être humain et le rongeur porteurs de troubles schizophréniques, ont pu montrer des gènes de susceptibilité et des modifications épigénétiques résultant principalement de phénomènes inflammatoires capables d’altérer la fonction des canaux ioniques, essentiellement potassiques et calciques [4,5,6]. Ces canaux sont présents à la fois sur le corps cellulaire et au niveau de la synapses, mais la plupart des traitements pharmacologiques proposés actuellement visent essentiellement la communication synaptique.
Pertinence, originalité et objectifs
Ce travail à pour objectif l’étude des mécanismes physiopathologiques de la schizophrénie grâce à des approches de modélisation mathématique, de simulation et de traitement du signal multi-échelles en s’appuyant sur un modèle animal de la pathologie. Il permettra d’identifier de nouvelles cibles thérapeutiques, promettant une meilleure personnalisation des traitements pour une médecine de précision. La modélisation mathématique associée à un traitement de signaux électrophysiologiques peuvent permettre d’extraire la contribution individuelle des dysfonctions de canaux ioniques, des perturbations synaptiques et des modifications de connectivité structurelle. Nous émettons l’hypothèse que les changements de plasticités observés dans les modèles animaux pourraient être le résultat d’un combinaison de facteurs intrinsèques à la cellules, incluant une dysfonction des canaux ioniques, et de facteurs de connectivité tels que les capacité de libération de neurotransmetteurs ou la topologie des connexions (balance excitation-inhibition, projections, etc) dont le principal perturbateur serait induit par les modifications de fonctions des canaux ioniques et d’homésotasie ionique. Si cette
hypothèse venait à être confirmée, ces canaux cellulaires pourraient être considérés comme de nouvelles cibles thérapeutiques et mener à une exploration pharmacologique visant à améliorer symptômes cognitifs et négatif de la pathologie. De plus, dans le contexte de la règle des 3R concernant l’expérimentation animale, modélisation et simulation représentent des outils alternatifs aux approches expérimentales utilisant des animaux particulièrement adaptés pour répondre à ces nouveaux enjeux.
Méthodologie et techniques mises en œuvre
Modélisation biologiquement réaliste des réseaux de neurones.Le travail s’appuiera sur un modèle mathématique d’hippocampe [12,13]déjà développé dans le cadre des thèses de F. Giovannini et d’A. Aussel, co-encadrée par L. Buhry (LORIA) et R. Ranta (CRAN). La première étape consistera à adapter ce modèle humain à un modèle murin en utilisant notamment les données de connectome du Allen Institute puis à compléter le modèle d’hippocampe en y intégrant différents types d’interneurones dont les constantes de temps synaptiques et projections topographiques diffèrent d’une variété à l’autre, pouvant jouer un rôle crucial dans la synchronisation des activités du réseau neuronal et donc, l’intégration des potentiels d’action dans les mécanismes de plasticité intervenant dans l’apprentissage. La seconde étape visera à l’implémentation des mécanismes de plasticité synaptique et fera intervenir des compétences de programmation parallèle pour l’implémentation optimisée des réseaux et la résolution d’équations d’équations différentielles non linéaires dans des graphes de très grande dimension. Dans cette optique, le.a doctorant.e interagira avec J. Gaidamour (IECL), mais aussi avec nos collaborateurs S. Contassot (Loria) et M. Stimberg (Institut de la Vision) qui est le principal développeur du logiciel avec lequel nous avions implémenté le modèle initial (Brian) [15].
Une fois le modèle conçu en conditions non pathologiques, nous explorerons parla simulation différents scenarios physiopathologiques incluant le rôle joué par les dysfonctions de canaux ioniques et des propriétés purement synaptiques, ainsi que l’influence de la connexion structurelle, souvent différente (considérée comme cause ou conséquence de mécanismes de compensation), chez le sujet souffrant de troubles du spectre de la schizophrénie, de celle du sujet sain. Certains de ces mécanismes sont explorés actuellement dans le cadre de la thèse de L. Raison-Aubry et du travail de postdoctorat de L. Naudin, sous la direction de L. Buhry. Ces résultats préliminaires obtenus dans un modèle de rétine [11, 14] serviront de point de départ. L’exploration des différents régimes de fonctionnement et des paramètres du modèle demandera également des compétences HPC (High Performance Computing) pour lesquels J. Gaidamour sera sollicité.
Signaux électrophysiologiques.Afin de confronter la modélisation précédente aux signaux réels, il est nécessaire de rajouter une dernière étape de modélisation biophysique qui permettra de générer des champs électriques dépendants de la morphologie neuronale et de l’anatomie. Dans nos travaux précédents, cette étape comprenait essentiellement des contributeurs synaptiques dipolaires [13], et nous souhaitons enrichir le modèle en intégrant des potentiels d’action. Les premiers résultats [7] indiquent que la contribution de ces derniers dans les hautes fréquences peut être significative (voir aussi [8]), et nous souhaitons confronter ces modèles à des enregistrements électrophysiologiques strictement contrôles. Ces enregistrements, effectués sur des tranches d’hippocampe après stimulation électrique par nos collaborateurs de l’équipe COMETE (UMR 1075 INSERM Université de Caen,sur un modèle animal de schizophrénie [3]) serviront de référence pour le modèle mathématique pathologique. Ils nécessiteront un traitement préalable, afin de séparer les différents contributeurs qui les génèrent. Les développements récents (thèse de P Jurczinsky et travaux associés) sur la séparation spikes-LFP (potentiels d’action-courants synaptiques) [9] et sur l’activité proches vs. propagée [10] seront adaptées au contexte des enregistrements sur tranche (multi-electrodes arrays MEA).
Encadrement et collaborations
La thèse se déroulera au LORIA (Neurorhythms), et au CRAN (BioSIS) avec un co-encadrement à l’IECL pour ce qui concerne la mise en œuvre de la simulation numérique et le HPC. Les expertises du LORIA seront particulièrement sollicitées pour les aspects de modélisation et de simulation, en particulier pour la modélisation biologiquement réaliste des réseaux hippocampo-corticaux. La contribution du CRAN concernera davantage la modélisation des phénomènes électrophysiologiques (problème direct des modèles de sources de courant jusqu’aux capteurs), ainsi que l’analyse des données expérimentales. Celle de l’IECL concernera les aspects liée à l’implémentations logiciels des modèles de neurones et de plasticité cérébrale, à la simulation de système complexes et à l’efficacité des expérimentations envisagées. La validation des modèles développés sera faite par confrontation à des données expérimentales in vitro chez le rongeur dans l’équipe COMETE. Les développements informatiques seront intégrés dans le logiciel de neurosciences computationnelles Brian, développé par l’équipe de Neurosciences Computationnelles des Systèmes Sensoriels, Institut de la Vision, Paris [15].
Ce sujet est complémentaire d’une collaboration que nous établissons entre les différents laboratoires dans le cadre d’un dépôt de pré-proposition ANR.
Références Bibliographiques
1- DSM-V (Diagnostic and Statistical Manual of Mental Disorders). American Psychiatric Association. 2015.
2- Krogmann, A. et al. Keeping up with the therapeutic advances in SCZ: A review of novel and emerging pharmacological entities. CNS Spectr. 2019.
3- Percelay S, et al. Functional Dysregulations in CA1 Hippocampal Networks of a 3-Hit Mouse Model of SCZ. Int J Mol Sci. 2021.
4-Lam J, et al. The therapeutic potential of small-cond. KCa2 channels in neurodegenerative and psychiatric diseases. Expert Opin Ther Targets. 2013.
5- Andrade, A. et al. Genetic associations between voltage-gated calcium channels and psychiatric disorders. Int. J. of Mol. Sci. 2019.
6- Dietz, A.G. et al. Glial cells in schizophrenia: a unified hypothesis, The Lancet Psychiatry, 2020.
7 – Aussel, A. et al. Extracellular synaptic and action potential signatures in the hippocampal formation: a modelling study. CNS*2019, Barcelona.
8- Scheffer-Teixeira R et al. On high-frequency field oscillations (>100 Hz) and the spectralleakage of spiking activity. J Neurosci. 2013.
9- Le Cam S. et al. A Bayesian approach for simultaneous spike/LFP separation and spike sorting, J. of Neural Eng., submitted 2022
10- Juczinsky P et al Separating local and propagated contributors to the Behnke-Fried microelectrode recordings, BioSignals, 2021
11- Naudin, L. et al.. A general pattern of non-spiking neuron dynamics under the effect of K and Ca channel modifications. J. Comp. Neurosci.11.2022.
12-Giovannini, F. et al. The CAN-In network : a biologically inspired model for self-sustained theta-oscillations and memory maintenance in the hippocampus. Hippocampus, 2017.
13- Aussel, A.et al. A detailed anatomical and mathematical model of the hippocampal formation for the generation of SWR and theta-nested gamma oscillations. J. Comp. Neurosci. 2018
14- Raison-Aubry, L. et al.. Modélisation des mécanismes sous-jacents de l’ERG pathologique dans la schizophrénie. L’Encéphale, Paris, 2022.
15- Stimberg, M. et al. Brian 2, an Intuitive and Efficient Neural Simulator. eLife 8, 2019: e47314.
Workshop on Clinical Computational Neuroscience:
How computational neuroscience can solve problems from the clinic
July 7th 2021
at OCNS*2021 Online
To register to the OCNS meeting: https://www.cnsorg.org/cns-2021
Organizers: Xenia Kobeleva and Laure Buhry
Date: July 7th — 9:30 am to 5:45 pm (Paris Time)
Abstract and Programme:
Despite great advances in understanding of brain function using computational neuroscience, complex neural activity analyses and models are rarely used for clinical purposes. This workshop aims to bridge this gap and to bring computational neuroscientists and clinicians together to discuss current challenges and solutions for bringing more computational neuroscience to patient-oriented questions. Topics will include translation from animal models to patients, exploration of virtual therapies and personalized medicine. Apart from several success stories of clinical computational neuroscience, we will also have discussion rounds to deep dive into current challenges of the computational neuroscience community and their potential solutions. The outcome of the workshop will be to:
– inform interested researchers on clinically relevant questions
– provide an overview of computational tools that can be used to answer these questions
– establish a working group on clinical computational neuroscience for all interested clinicians and researchers and discuss practical steps to work on the roadblocks that were identified during the workshop
The day will be organized around three « challenges »:
Challenge 1: Translation from animal models to patients: bridging species and scales
Challenge 2: Virtual therapies: exploring new options and refining existing treatments
Challenge 3: Personalized medicine: how to make models fit the patient
|
Times in CET |
Agenda |
| 09:30-09:45 | Introduction to clinical computational neuroscience and the workshop |
| Challenge 1: Translation from animal models to patients: bridging species and scales | |
| 9:45-10:15 |
Talk 1: Kirk Leech (confirmed) |
| 10:15-10:45 |
Talk 2: Matthieu Gilson (confirmed) |
| 10:45-11:15 |
Panel discussion: What are the current challenges? What tools and methods do we need to solve the challenges? with Kirk Leech, Matthieu Gilson and Julien Modolo |
|
11:15-11:30 |
Break |
|
Challenge 2: Virtual therapies: exploring new options and refining existing treatments |
|
|
11:30-12:00 |
Talk 1: Yujiang Wang (confirmed) |
|
12:00-12:30 |
Talk 2: Jil Meyer (confirmed) |
|
12:30-13:00 |
Discussion and brainstorming: What are the current challenges? What tools and methods do we need to solve the challenges? with Yujiang Wang and Jil Meyer |
|
13:00-14:00 |
Break |
| Challenge 3: Personalized medicine: how to make models fit the patient | |
|
14:00-14:30 |
Talk 1: Christian Meisel (confirmed) |
|
14:30-15:00 |
Talk 2: Vesna Vuksanovic (confirmed) |
|
15:00-15:30 |
Talk 3: Peter Hitchcock (confirmed) |
| 15:30-16:00 |
Discussion and brainstorming: What are the current challenges? What tools and methods do we need to solve the challenges? with Christian Meisel, Peter Hitchcock, Vesna Vuksanovic |
| 16:00-16:30 | Final break |
|
16:30-17:30 |
Final discussion Clinical computational neuroscience- challenges, needs and opportunities of the OCNS community, planning the road ahead with all speakers and panel discussion members |
|
17:30… |
Get together with virtual coffee and drinks (small Zoom chat rooms?) |
Speakers:
– Kirk Leech, Executive Director of The European Animal Research Association (EARA)
– Matthieu Gilson, Institute of Neuroscience and Medicine (INM-6) and Institute for Advanced Simulation (IAS-6) and JARA Institute Brain Structure-Function Relationships (INM-10), Jülich Research Centre, Jülich, Germany, and Center for Brain and Cognition, Department of Information and Telecommunication technologies, Universitat Pompeu Fabra, Barcelona, Spain
– Julien Modolo (panel discussion), Research Scientist at INSERM, LTSI (Laboratory of Signal and Image processing) Rennes, France
– Yujiang Wang, CNNP Lab, Interdisciplinary Computing and ComplexBioSystems Group, School of Computing, Newcastle University, Newcastle uponTyne, United Kingdom
– Jil Meier, Berlin Institute of Health at Charité – Universitätsmedizin Berlin, Berlin, Germany, and Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt-Universität zu Berlin, Department of Neurology with Experimental Neurology, Brain Simulation Section, Berlin, Germany
– Christian Meisel, Department of Neurology, Charité – Universitätsmedizin Berlin and Berlin Institute of Health, Berlin, Germany
– Vesna Vuksanovic, Swansea University Medical School and Health Data Research United Kingdom
– Peter Hitchcock, Cognitive, Linguistic, and Psychological Sciences, Brown University, Providence, RI
Talk abstracts:
Challenge 1: Translation from animal models to patients: bridging species and scales
Kirk Leech
Executive Director of The European Animal Research Association (EARA)
Title: Update on the use of animals in research
Abstract: The goal of this talk would be, for computational neuroscientists, to gain insight into the place of computational approaches in the translation from animal models to patients.
The talk will focus on the need for the research community to more pro-actively engage with regulators, media and public on the need for animals in research. I will use the recent example of EURL ECVAM proposals to immediately end the use of animal derived anti-bodies in research as an example of the dangers of the research community ‘falling asleep at the wheel’ whilst those totally opposed to animal use are increasing their engagement with regulators. I will also explain how European transparency agreements work – there are now 5 in operation, involving close to 340 institutions (with three more agreements close) – to show how they can play a role in improving public understanding and acceptance of animal research, including where we can and currently can’t move away from animal use.
Matthieu Gilson
Institute of Neuroscience and Medicine (INM-6) and Institute for Advanced Simulation (IAS-6) and JARA Institute Brain Structure-Function Relationships (INM-10), Jülich Research Centre, Jülich, Germany
Center for Brain and Cognition, Department of Information and Telecommunication technologies, Universitat Pompeu Fabra, Barcelona, Spain
Title: Bridging spatial scales in biophysical models for translational clinical applications
Abstract: A great deal of dynamic models have been developed to reproduce signals obtained from neuroimaging techniques in order to formalize the relationship between structure and function. Such models typically relate empirical data like anatomical connectivity with mechanisms that describe the neuronal dynamics, which may be modelled at the microscopic level using spiking neurons or at a more mesoscopic level using neuronal population nodes. For neurology or neuropsychiatry, the computational models have been extended to incorporate data like neuromodulator or gene expression maps, which are of importance for neuropathologies. The goal is then to uncover their influence in shaping the activity patterns at the whole brain level. Such a model construction implies a number of choices in the design for the mechanisms, as well as parameters to estimate or set.
In this talk I will focus on models for functional magnetic resonance imaging (fMRI) and review recent advances in the direction of clinical research. Meanwhile pointing at strengths and limitations of those studies, I will argue for a modelling that keeps a balance between predictability and interpretability. The former is to be understood in the sense of enabling a robust decoding of the fMRI activity patterns for distinct conditions (e.g. patients versus controls, or different pathological subtypes) using the estimated model parameters to derive biomarkers. The latter is necessary to relate the model parameters back to biological variables, and uncover the mechanisms involved in the neuropathological activity, beyond a « simple » phenomenological or statistical analysis of the empirical data. This perspective has been recently developed in an opinion article [1], which proposes to combine decoding and biophysical models to achieve the desired modelling requirements.
Julien Modolo (panel discussion)
Research Scientist at INSERM, LTSI (Laboratory of Signal and Image processing) Rennes, France
Expertise abstract: The translation of neuromodulation strategies from in vivo to in clinico is notoriously challenging. Among those challenges, emerging approaches attempt at combining morphological and computational approaches of electric field dosimetry combined with realistic models of neuronal activity to predict the response of brain tissue to specific neuromodulation sequences. In this discussion panel, I will discuss existing strategies on the topic, along with their assumptions and capabilities, and how this could bring neuromodulation faster from animal to human studies.
Challenge 2: Virtual therapies: exploring new options and refining existing treatments
Yujiang Wang
CNNP Lab, Interdisciplinary Computing and ComplexBioSystems Group, School of Computing, Newcastle University, Newcastle uponTyne, United Kingdom
Title: Towards time-adaptive treatments in epilepsy using data-driven subject-specific models
Abstract: Epilepsy is increasingly being recognised as a dynamic disease, where seizure characteristics and severity change over time in each patient. Understanding why, in the same patient, one seizure is severe and debilitating, while the next seizure is relatively mild in symptoms is crucial for developing time-adaptive treatments that can render severe seizure into more benign forms.
To this end, we recently investigated over 500 seizure EEGs and found variable electrographic seizure dynamics within individual patients [1]. A detailed analysis of the nature of this within-subject variability revealed that specific recurrent patterns of brain dynamics during seizures can be of variable duration in each seizure or be completely absent [2]. Importantly, the variability appeared to follow subject-specific circadian or longer timescale modulations.
We hypothesise that markers of these modulations on different timescales can be captured on continuously recorded EEG data over days to years. In this talk I will show that several features in the continuously recorded EEG display fluctuations on ultradian, circadian, and infradian timescales, and several subject-specific timescales are strongly associated with the presence/absence or duration of recurrent patterns in each patient. Finally, I will also demonstrate that these fluctuations on different timescale provide a good explanation (above chance level) for how seizures change over time in each patient [3].
Our results indicate that slow (ultradian, circadian, and infradian) fluctuations captured from continuously recorded EEG may serve as a biomarker to track and predict how seizures change over time. This is an important step in interacting and controlling the seizure-modulating fluctuation. Ultimately, time-adaptive treatments (e.g. via drug-delivery systems) could not only render severe seizure into more benign forms, but may be able to suppress symptomatic seizures altogether, whilst reducing side-effects.
Jil Meier
Berlin Institute of Health at Charité – Universitätsmedizin Berlin, Berlin, Germany
Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt-Universität zu Berlin, Department of Neurology with Experimental Neurology, Brain Simulation Section, Berlin, Germany
Title: Multiscale co-simulation of deep brain stimulation
Deep brain stimulation (DBS) has been successfully applied in various neurodegenerative diseases as an effective symptomatic treatment. However, its mechanisms of action within the brain network are still poorly understood. Many virtual DBS models analyze a subnetwork around the basal ganglia and its dynamics as a spiking network with their details validated by experimental data. However, connectomic evidence shows widespread effects of DBS affecting many different cortical and subcortical areas. From a clinical perspective, various effects of DBS besides the motoric impact have been demonstrated. The neuroinformatics platform The Virtual Brain (TVB) offers a modeling framework allowing us to virtually perform stimulation, including DBS, and forecast the outcome from a dynamic systems perspective prior to invasive surgery with DBS lead placement. For an accurate prediction of the effects of DBS, we implement a detailed spiking model of the basal ganglia, which we combine with TVB via our previously developed co-simulation environment. In the first demonstration of our model, we show that virtual DBS can move the basal ganglia firing rates of a Parkinson’s disease patient’s network towards the healthy regime. This study represents a proof of concept to perform virtual DBS in a co-simulation environment with TVB. The developed modeling approach has the potential to optimize DBS lead placement and configuration and forecast the success of DBS treatment for individual patients.
Challenge 3: Personalized medicine: how to make models fit the patient
Vesna Vuksanovic
Swansea University Medical School and Health Data Research United Kingdom
Title: Data-driven approach to neuroimaging analysis to identify dementia subtypes
Abstract: Understanding variations in neurodegeneration across brain disorders that cause dementia represents one of the main challenges in clinical neuroscience. Combining advanced neuroimaging with computational models in network neuroscience has already contributed to the understanding of biological bases of such variations. Here, I will present data on the analysis of the cortical networks, derived from magnetic resonance brain scans, to assess which organisational patterns of these networks are characteristics of dementia subtypes. Statistical methods of graph theory combined with hierarchical clustering was used to analyse topology of the cortical interactions across two types of dementia, frontotemporal dementia (its behavioural variant) and Alzheimer’s disease. The aim was to identify patients with more homogenous patterns of neurodegeneration.
Christian Meisel
Department of Neurology, Charité – Universitätsmedizin Berlin and Berlin Institute of Health, Berlin, Germany
Title: Resilience in neural systems: from an understanding based on dynamical principles towards clinical diagnostics
Abstract: In the recent years it has become apparent that dynamical system frameworks and bifurcation theory may apply to biomedical systems, beyond the systems classically considered in physics. Here, we will provide a critical survey of recent research in this field. We will discuss approaches building on dynamical systems theory to develop a better understanding of resilience and risk to rapid state changes, such as the onset of epileptic seizures, and how these theory-driven insights may provide clinical diagnostic markers.
Peter Hitchcock
Cognitive, Linguistic, and Psychological Sciences, Brown University, Providence, RI
Title: On the Importance of Incorporating Time and Context in Computational Psychiatry Models
Abstract: Why has computational psychiatry been slow to influence routine clinical practice? I will argue that one reason is the field has had difficulty recognizing the variability among mental health problems—and, consequently, the need to model context and temporal dynamics for many problems. I will propose three heuristics for deciding when modeling time and context is important. And as a case study of when such modeling is important, I will highlight contextual factors and temporal dynamics relevant to rumination and worry and thus to depression and anxiety disorders.
