Projects
Our central hypothesis that the need for visual information to be captured with the required resolution and sensitivity must be counterbalanced by the metabolic capacity of the visual system.
To address this, limits2vision aims (i) to generate an integrated view on how the retina and the early visual system overcome the many challenges they face, and (ii) to extract general rules that help neural systems to adapt to rapidly changing conditions. The key questions are:
- To what extent do genetics define health or pathology of the (aging) retina and, especially, how do genetics impact metabolism in the different cell types (i.e. in the outer retina)?
- What characterises healthy retinal metabolism and how do its limits shape visual function and information processing?
- What are the general principles that enable effective and dynamic information processing in the (early) visual system in view of those metabolic limitations and challenges?
To address these goals, limits2vision will combine the broad spectrum of approaches (from cell and tissue cultures to the whole organism) and techniques (from single-cell genetics to in vivo imaging and computational modelling).
The outcomes of limits2vision are expected to greatly improve our understanding of how the retina can fulfil its unique functions. Knowing the mechanisms that allow the tissue to “live on the edge” may also guide the rational design of future therapeutic interventions for retinal diseases and visual disorders.
The following projects from the French (FR) and German (DE) partners cover a range of topics that can be grouped into three main scientific sectors, namely genetics, metabolism, and information processing.
DE-9 The role of HCN1 channels in shaping of retinal signal output.
DE-9 – The role of HCN1 channels in shaping of retinal signal output.
IOR, Tübingen University
iRTG PhD Student
Tübingen University
Seeliger lab – Ocular Neurodegeneration
Mathias Seeliger on Web of Science
This project aims to provide novel insights in HCN1 ion channel activity for the comprehensive understanding of outer retinal signalling and ultimately disclose new targets for therapy. Hyperpolarization-activated cyclic nucleotide-gated (HCN) channels modulate the timing of electrical signals in many cell types including neurons. In the retina, type 1 HCN channels (HCN1) enable short-term adaptation (via their voltage sensing capability with delayed channel action) and increase temporal resolution. The aim of this project is to elucidate the differential roles of Hyperpolarization-activated Cyclic Nucleotide-gated-1 (HCN1) channels in retinal signalling of photoreceptors in health and disease. To this end, we are establishing an informative set of functional in vivo data in respective mouse lines using Electroretinography (ERG) supported by a histological workup in normal and mutant retina.
FR-1 Shedding Light on Myopia Through Inherited Retinal Disorders.
FR-1 – Shedding Light on Myopia Through Inherited Retinal Disorders
IDV, Sorbonne University
Zeitz lab – Identifying the molecular pathways involved in eye diseases, from gene to therapy
Starting in October 2026 / PhD position available -> apply here
Myopia is the most prevalent eye disorder worldwide, affecting 30% of the population and projected to reach 50% by 2050. In its severe form, it can lead to irreversible blindness through complications such as retinal degeneration. Despite its global impact, the biological mechanisms driving myopia remain poorly understood.
Emerging evidence highlights the retina as a central regulator of eye growth, translating visual signals into molecular cues that shape ocular development. Our research leverages inherited retinal disorders—particularly complete congenital stationary night blindness (cCSNB)—as powerful models to uncover these mechanisms. cCSNB disrupts retinal signaling and is consistently associated with high myopia, offering a unique window into disease pathways.
Using RNA sequencing in mouse models, we have identified key differentially expressed genes linked to myopia and uncovered novel pathways, including MAPK signaling, eye development, and synaptic function. Yet, the roles of many of these genes remain unexplored.
At the same time, current treatments (e.g., lenses, atropine) show limited efficacy and notable side effects, especially in severe or syndromic cases. There is a critical need for new, targeted therapeutic strategies.
FR-2 Metabolic dysregulation in disease models for novel IRD gene defects.
FR-2 – Metabolic dysregulation in disease models for novel IRD gene defects.
IDV, Sorbonne University
Starting later.
DE-1 Krebs cycle aconitase deficiency in inherited retinal ganglion cell dysfunction: Functional validation of ACO2 variants in patients with optic atrophy and their metabolic impact.
DE-1 – Krebs cycle aconitase deficiency in inherited retinal ganglion cell dysfunction: Functional validation of ACO2 variants in patients with optic atrophy and their metabolic impact
IOR, Tübingen University
IOR, Tübingen University
iRTG PhD Student
Tübingen University
Wissinger lab – Molecular Genetics Laboratory
Mutations in the ACO2 gene, encoding aconitase-2, a key enzyme of the mitochondrial tricarboxylic acid cycle (Krebs cycle), cause a spectrum of rare inherited neurometabolic diseases including syndromic but also isolated optic atrophy, a progressive blinding disease due to the degeneration of retinal ganglion cells and their axons which form the optic nerve. Interpreting ACO2 gene variants detected in patient by genetic testing as either pathogenic mutations or benign polymorphisms is particularly challenging for missense variants (which lead to single amino acid substitutions) and for intronic variants that may affect transcript splicing. These issues hamper reliable genetic diagnostics and proper counselling of patients and families.
The project aims to assess the pathogenic impact of such putative disease-associated variants in ACO2 by (i) applying functional in vitro bioassays, (ii) investigating disease-related metabolic dysfunction by means of mitochondrial function assays and (iii) exploring the presence of ACO2-deficiency specific metabolic biomarkers or biomarker profiles in biofluids. To this end, we first will compile a comprehensive ACO2 variant dataset from in-house and public genetic databases and score the evidence for the pathogenicity of each putative disease-associated variants from available functional or genetic data or from the application of a bioinformatic prediction toolset. For the analysis of potential splicing variants, ACO2 transcript analysis will be done on patient blood samples (where available) or applying minigene assays based on the expression of mutant versus wildtype ACO2 constructs in HEK293 cells. For the evaluation of missense variants, a high-throughput complementation assay in aco2-deficient yeast will be established and applied. In addition, mtDNA content (compared to nuclear DNA) will be determined in patient blood samples by dPCR and metabolomic profiles explored in patient biofluids as potential disease-associated biomarkers.
DE-2 Impact of Complement Factor H risk variant Y402H on metabolic damage asso-ciated to age-related macular degeneration (AMD).
DE-2 – Impact of Complement Factor H risk variant Y402H on metabolic damage associated to age-related macular degeneration (AMD)
IOR, Tübingen University
Director of the IOR
Tübingen University
iRTG PhD Student
Tübingen University
Leelja Rössler
iRTG PhD Student in FR-4
Sorbonne University
associated PhD Student
Tübingen University
Ueffing Lab – Molecular Biology of Retinal Degeneration
Age-related macular degeneration (AMD) is a progressive degenerative disease of the macula, which leads to blindness in the elderly population and for which therapeutic options are limited. The retinal pigment epithelium (RPE) cells and the adjacent neuroretina, mainly photoreceptors, coexist in a well-regulated metabolic balance with each other. During disease progression, this balance is altered by several risk factors contributing to AMD, such as ageing, genetic risks, unhealthy life-style habits, such as smoking or poor diet. Our group and others have shown that the AMD-associated genetic variants alter the metabolism of RPE cells, causing mitochondria damage, autophagy dysfunction, and oxidative stress. In this project, we aim to identify genetic risk-dependent metabolic changes and the different metabolic response of the RPE cells and photoreceptors. These findings may help to identify novel therapeutic targets to treat AMD.
In this project, we will investigate the impact of AMD-genetic risk and life-style factors on the metabolism of RPE cells differentiated from induced pluripotent stem cells (iPSCs). Moreover, we will employ organotypic retinal explants derived from porcine eyes and co-culture them with iPSC-RPE cells to study the metabolic changes in the photoreceptors. RPE and retina metabolism will be studied by a combination of multiplexed in vitro assays, immunofluorescence, CO-Detection by indEXing (CODEX), high-resolution microscopy (2-photons) and functional tests. This will create a technical platform for pharmacological testing.
FR-3 Inflammation-metabolism crosstalk as a driver of retinal degeneration.
FR-3 – Inflammation-metabolism crosstalk as a driver of retinal degeneration
IDV, Sorbonne University
IDV, Sorbonne University
iRTG PhD Student
Sorbonne University
associated PhD Student
Sorbonne University
Sennlaub lab – Inflammation & Immunology of Retinal Diseases
Age-related macular degeneration (AMD) is a leading cause of vision loss, and growing evidence links it to the interplay between chronic inflammation and metabolic dysfunction in the retina. Understanding how these processes influence disease progression is essential for developing better treatments. This project will help uncover fundamental mechanisms at the intersection of inflammation, metabolism, and neurodegeneration. In this project, we will explore how modulating inflammation — genetically and pharmacologically — affects disease severity in outer retinal metabolic dysfunction relevant to AMD. The project uses a multidisciplinary approach, including in vivo retinal imaging, electron and fluorescence microscopy, transcriptomics, and biochemical assays.
FR-4 The retinal pigment epithelium as a developmental regulator of the neural retina: Insights from albinism and FHONDA syndrome.
FR-4 – The retinal pigment epithelium as a developmental regulator of the neural retina: Insights from albinism and FHONDA syndrome
IDV, Sorbonne University
iRTG PhD Student
Sorbonne University
Anastasia Chikhladze
iRTG PhD Student in
DE-2
Tübingen University
associated PhD Student
Sorbonne University
Rebsam lab – Development, evolution and function on commissural systems
Proper development of the neural retina is critically dependent on signals from the adjacent retinal pigment epithelium (RPE), a monolayer of pigmented epithelial cells essential for retinal homeostasis and function. During embryogenesis, the RPE not only contributes to the structural organization of the retina but also plays a pivotal role in modulating the proliferation and differentiation of retinal progenitor cells. Mutations in genes expressed in the RPE in congenital diseases such as albinism and FHONDA syndrome cause altered retinal development leading to visual impairment, but the precise mechanisms remain incompletely understood. Albinism is characterized by hypopigmentation and visual defects such as misrouting of optic nerve fibres, causing an abnormal stereovision and foveal hypoplasia, causing a reduced visual acuity. Interestingly, FHONDA syndrome (Foveal Hypoplasia, Optic Nerve Decussation defects, and Anterior segment Dysgenesis) involves analogous abnormalities of the retina and optic pathways, but without pigmentation deficits. We have identified candidate pathways playing a key role in albinism and FHONDA syndrome.
This project aims at studying signalling pathways in the RPE that control retinogenesis in normal and pathological development. To this end, we will use a combination of biological systems (in vitro: RPE cells and retinal organoids differentiated from human induced pluripotent stem cells (hiPSCs), in vivo: mouse models), Omics approaches (metabolomics, transcriptomics) as well as genetic modification of hiPSCs (CRISPR/Cas9). Altogether, this project project will decipher the cellular and molecular mechanisms underlying these two rare diseases and shed new light on the processes of visual system development.
FR-12 Defining metabolic interventions to protect cone photoreceptors.
FR-12 – Defining metabolic interventions to protect cone photoreceptors
Director of the IDV
Sorbonne University
IDV, Sorbonne University
iRTG PhD Student
Sorbonne University
Picaud lab – Visual information processing: neural coding and vision restoration
The loss of cone photoreceptors is the major cause of blindness in highly developed countries. Blinding diseases include hereditary retinal dystrophies and other diseases such as age-related macular degeneration or retinal detachment. In humans, cone photoreceptors are essential for colour vision and high visual acuity. Unfortunately, no treatment can prevent the loss of cone photoreceptors in any of these diseases and the mechanisms of cone degeneration are very poorly understood. However, given the very high energy demand of cones, it seems likely that many cone diseases have a bearing on energy metabolism and that vice versa a strengthening of cone metabolism can produce cone neuroprotection.
In this project, we will evaluate neuroprotection on different in vitro and in vivo models of cone degeneration. The goal will be to define how universal the neuroprotective effect of our lead compounds is for different types of cone neurodegenerative diseases. We will therefore assess whether the selected neuroprotective molecules can prevent the loss of cone photoreceptors in a model of retinal detachment and in mouse models of primary (e.g., cpfl1) and secondary (e.g., rd1) cone degeneration. Overall, we will assess to what extent cone photoreceptor survival depends on energy metabolism and whether treatments intended to improve cone metabolism can prevent cone degeneration in a gene- and mutation-independent fashion.
DE-3 Impact of damaged Bruch´s membrane on RPE metabolism: metabolic analyses of in vitro iPS-RPE models.
DE-3 – Impact of damaged Bruch´s membrane on RPE metabolism: metabolic analyses of in vitro iPS-RPE models.
IOR, Tübingen University
IOR, Tübingen University
iRTG PhD Student
Tübingen University
Clark lab – Molecular mechanisms driving age-related macular degeneration (AMD)
This project investigates how changes in Bruch’s membrane (BrM), the extracellular matrix beneath the retinal pigment epithelium (RPE), impact RPE cell metabolism and, by extension, the health of the retina. We will develop and refine in vitro models using highly characterised induced pluripotent stem cell (iPSC)-derived RPE cells cultured on BrM isolated from porcine eyes. These models will be subjected to defined extracellular matrix (ECM) alterations, mimicking early changes seen in retinal diseases such as age-related macular degeneration (AMD). Using state-of-the-art imaging (electron microscopy, RAMAN, fluorescent lifetime microscopy) and transcriptomic profiling (RNAseq), we will map the resulting metabolic and morphological responses. The ultimate aim is to produce a comprehensive “REMIX-MAP” (Retinal Epithelium and Matrix Interaction Morphological and Metabolic Activity Profile) to identify early indicators of retinal dysfunction.
BrM plays a vital role in maintaining RPE health, and its early deterioration is a hallmark of AMD. While previous research has shown that ECM stiffness and composition can drastically alter cellular metabolism in other systems, this project addresses a major knowledge gap in how such changes affect RPE cells specifically. This is crucial because the RPE’s metabolic state directly influences its ability to support photoreceptors and maintain visual function. By exploring these interactions, we will lay the foundation for future diagnostic and therapeutic approaches for retinal diseases.
DE-4 Single-cell multiplexed and targeted metabolic imaging of the retina to decipher energy flow and dependencies in cellular niches.
DE-4 – Single-cell multiplexed and targeted metabolic imaging of the retina to decipher energy flow and dependencies in cellular niches.
IOR, Tübingen University
Institute of Pathology
Tübingen University
iRTG PhD Student
Tübingen University
Schwarz lab – High-Resolution Functional Imaging and Testing
Schürch lab – Advanced Tissue Imaging & Tumor Microenvironment
The retina is one of the most metabolically active tissues in the body, and its ability to sustain vision depends on precisely coordinated energy flow between diverse cell types. Disruption of these metabolic interactions underlies vision-threatening diseases such as retinal detachment and diabetic retinopathy. Recent discoveries reveal that retinal cells have distinct metabolic pathways that can adapt – or fail – under stress. By mapping these pathways at single-cell resolution, we can uncover how cells support each other’s metabolism and how these relationships break down in disease. This knowledge has the potential to identify novel, targeted interventions for preserving or restoring vision.
This project is situated at the intersection of functional imaging and high-dimensional spatial biology. Using two-photon autofluorescence microscopy, we will capture the dynamic metabolic activity of retinal cells in real time by tracking NADH/FAD redox states under normal and challenged conditions. In parallel, we will apply CODEX high-multiplex microscopy to profile dozens of metabolic enzymes and cell-type markers simultaneously, generating a spatial atlas of retinal metabolism at subcellular resolution. These complementary approaches will be applied to healthy and diseased mouse retina models – such as retinal detachment, acute and chronic diabetic retinopathy – using both sexes to assess possible sex-dependent effects. We will also collaborate with multiple labs in the iRTG, integrate with metabolomics and intervention studies, and plan to extend the work to in vivo imaging in a second funding phase.
DE-5 The metabolism of rod and cone photoreceptors in health and disease.
DE-5 – The metabolism of rod and cone photoreceptors in health and disease.
IOR, Tübingen University
M3 Center / WSIC
Tübingen University
iRTG PhD Student
Tübingen University
associated PhD Student
Tübingen University
Paquet-Durand lab – Cell Death Mechanisms
Trautwein lab – Metabolomics & Systems Medicine.
The photoreceptors of the retina are characterized by what is likely the highest energy demand of any cell type in the human body. What fuels photoreceptors use, how these are supplied, and how they are metabolized, are largely still open questions. However, preliminary data suggests that photoreceptor energy metabolism may deviate significantly from what textbooks suggest. Moreover, we have reason to believe that the metabolism of rod and cone photoreceptors differs significantly between these two cell types, potentially explaining the selective vulnerability of these cells in certain neurodegenerative diseases. In this project, we will investigate the impact of targeted pharmacological and genetic manipulations on photoreceptor energy metabolism, under entirely controlled conditions, using organotypic retinal explants derived from wild-type and retinal degeneration mutant mice. A single-cell resolution expression atlas for retinal enzymes and metabolites related to energy metabolism will be established, using immunofluorescence, CO-Detection by indEXing (CODEX), high-resolution microscopy and ion mobility mass spectrometry MALDI-2 imaging with 5-µm resolution (microGRID). This will be supported by RNA sequencing and metabolomics/lipidomics, as well as NMR spectroscopy approaches, in situ biochemistry, and µERG functional testing to comprehensively assess the consequences of metabolic interventions in individual retinal cells.
FR-6 Glial activation and metabolic changes induced by vascular leakages in diabetic retinopathy (DR).
FR-6 – Glial activation and metabolic changes induced by vascular leakages in diabetic retinopathy (DR)
IDV, Sorbonne University
IDV, Sorbonne University
Tual Monfort – Live imaging in patients and cells
Xavier Guillonneau – Inflammation and Immunology in Retinal Diseases
Diabetic retinopathy (DR) is a leading cause of vision loss worldwide, particularly among working-age adults. Despite advances in diagnosis and treatment, understanding the underlying mechanisms remains challenging. We recently showed that lipid droplets (LDs) accumulate in regions with vascular leakage and aneurysms, which are pathological hallmarks of DR and are associated with the activation of retinal glial and immune cells. These dynamic organelles play essential roles beyond energy storage—alleviating lipotoxic stress, endoplasmic reticulum stress, and oxidative stress. While lipid-lowering drugs can reduce DR progression, our knowledge of cell-specific lipid dynamics in the retina is limited by current imaging techniques, which are invasive and fail to capture the three-dimensional interactions in physiological conditions. This knowledge gap hinders the development of targeted therapies for this increasingly prevalent condition.
In this project, we will pioneer an innovative approach by integrating two cutting-edge imaging techniques – Dynamic Full-Field Optical Coherence Tomography (D-FFOCT) and Coherent Anti-Stokes Raman Scattering (CARS) – into a novel microscope called No-D-EYE. This unique system will enable simultaneous, label-free, non-invasive 3D visualization of retinal cells and specific tracking of lipid droplets in living tissue over extended periods. Using mouse retinal explants treated or not with lipids or retinal explants from mouse models of vascular leakage, we will monitor lipid dynamics and cellular responses longitudinally. Advanced AI-based algorithms will automatically extract cell positions, types, and lipid droplet characteristics to provide a comprehensive physiological readout of lipid-induced modification of the glial and neuronal compartment of the retina. We will test promising compounds, including PPAR modulators and anti-inflammatory agents, to identify potential therapeutic interventions that normalize lipid handling and restore metabolism in diabetic retinopathy. Additionally, we will leverage the advanced facilities in Tübingen, including mass spectrometry and Raman spectroscopy, to precisely identify and characterize lipid types within the retinal tissue, complementing the spatial and temporal information provided by our No-D-EYE system.
FR-7 Impact of deregulated RPE metabolism on RPE and photoreceptor function and global retinal homeostasis.
FR-7 – Impact of deregulated RPE metabolism on RPE and photoreceptor function and global retinal homeostasis.
IDV, Sorbonne University
Starting later.
DE-6 From photoreceptor to brain: Spatiotemporal activity and metabolic mapping of the visual system in mouse models of retinal degeneration.
DE-6 – From photoreceptor to brain: Spatiotemporal activity and metabolic mapping of the visual system in mouse models of retinal degeneration.
WSIC, Tübingen University
IOR, Tübingen University
iRTG PhD Student
Tübingen University
Herfert lab – Functional and metabolic brain imaging
Haq lab – Neuroretinal Electrophysiology and Imaging
Blinding diseases are typically detected only after noticeable visual impairment has developed, limiting treatment options. Current diagnostic methods like electroretinogram (ERG) provide limited insights into early degenerative changes within specific retinal cells. We aim to identify early biomarkers of retinal degeneration by systematically mapping visual responses from photoreceptors to cortical brain areas. Understanding how early-stage retinal degeneration affects downstream signalling and metabolism could reveal new targets for treatment and vision rescue strategies.
In this project, we will combine cutting-edge retinal and brain imaging techniques to create a comprehensive map of visual system dysfunction in retinal degeneration mouse models (rd1, rd10, cpfl1, and wild-type controls). Using micro-ERG (µERG) with multi-electrode arrays and calcium imaging, we will measure cell-type-specific retinal responses ex vivo. In addition, we will assess whole-brain functional activity and glucose metabolism in vivo using functional MRI and PET imaging during visual stimulation. By correlating cellular retinal responses with brain-wide activity patterns, we will identify where abnormal signal deterioration emerges in the visual pathway and investigate energy-dependent mechanisms underlying cellular dysfunction.
DE-7 Visual representations in young adult and aged mice.
DE-7 – Visual representations in young adult and aged mice.
IOR, Tübingen University
IOR, Tübingen University
iRTG PhD Student
Tübingen University
associated PhD Student
Tübingen University
Garaschuk lab – Physiology of Neural Circuits
Euler lab – Ophthalmic Research
Aging significantly reduces visual performance, but it remains unclear when and where along the visual pathway initial deficits arise. In the primary visual cortex (V1), aging causes weaker neural responses, energy homeostasis changes, and inflammation. Age-related retinal changes mirror brain alterations, with synaptic density decreases and dendritic remodelling affecting visual output. Understanding these mechanisms is crucial for developing strategies to delay or reverse aging-induced visual dysfunction.
In this project, we will characterize visual stimulus representations in both retina and V1 of young adult, middle-aged, and old mice using two-photon calcium imaging. Visual stimuli including gratings and natural movies will be presented to probe contrast sensitivity, orientation/direction selectivity, and feature representations. Additionally, we will test whether caloric restriction can mitigate aging-related changes by comparing old mice to age-matched calorie-restricted animals. The project combines retinal ganglion cell analysis in explants with in vivo V1 layer 2-3 cell recordings. By comparing visual representations in retina and V1, we expect to pinpoint age-related deficits in early vision and provide an unprecedented view of the aging visual system. Together with iRTG partners, we shall elucidate underlying metabolic and circuit-level mechanisms, enabling new strategies
DE-8 How energy demands shape optimal coding in early vision.
DE-8 – How energy demands shape optimal coding in early vision.
Director of the Hertie AI
Tübingen University
Hertie AI
Tübingen University
Mercator Fellow
University of Chicago, USA
iRTG PhD Student
Tübingen University
Berens lab – Department of Data Science
Palmer lab – Department of Organismal Biology & Anatomy
For decades, researchers agreed that an important goal of the early visual system is image compression for efficient information transmission. However, this view overlooks how behavioural demands and energy constraints shape neural coding. Current models focus only on spiking activity in retinal ganglion cells, ignoring that photoreceptors consume the most energy. We need a comprehensive understanding of how the entire retinal circuit balances energy consumption against behavioural needs to optimize visual coding. This knowledge could reveal how diseases and aging disrupt optimal coding principles.
In this project, we will develop a computational framework combining deep reinforcement learning with biophysically detailed retinal simulations. Using our differentiable simulator Jaxley, we will model entire mouse retinal circuits with realistic energy consumption estimates for each cell type. Artificial agents with these retinal models will be trained to survive in virtual environments, allowing us to study trade-offs between energy efficiency, image compression, and behavioural demands. We will integrate statistical physics principles to create new measures of coding optimality and investigate how disease and aging processes affect these optimal coding strategies.
FR-8 The function of amacrine cells and their impact on retinal computations.
FR-8 – The function of amacrine cells and their impact on retinal computations under realistic color stimulation
IDV, Sorbonne University
IDV, Sorbonne University
iRTG PhD Student
Sorbonne University
Matías Goldin, Olivier Marre – Visual information processing: neural coding and vision restoration
This project aims at isolating the role of specific types of amacrine cells, by measuring and modelling how they impact retinal ganglion cell responses to visual stimulation. You will perform ex vivo experiments recording retinae where specific types of amacrine cells have been made light sensitive, and you will use state of the art two-photon holographic stimulation to stimulate these amacrine cells while recording the impact of this stimulation on ganglion cells. The purpose is to decompose the retinal circuit and identify how a key player, amacrine cells, influence retinal computations.
Retinal neurons process natural images after they have been transformed by the imperfect optics of the eye. While spatial and temporal aspects of retinal coding have been extensively studied, the role of chromatic aberrations remains poorly understood. These optical distortions shift the focus of different wavelengths and create colored edges in natural scenes, yet their impact on neural computations has rarely been explored. We propose that such chromatic signatures provide powerful cues for feature extraction. Understanding how the retina exploits these aberrations will uncover new mechanisms of early visual processing and expand our models of sensory coding.
FR-9 How does correlated variability impact retinal information transmission?
FR-9 – How does correlated variability impact retinal information transmission?
IDV, Sorbonne University
iRTG PhD Student
Sorbonne University
Ulisse Ferrari – Visual information processing: neural coding and vision restoration
Understanding how the retina efficiently transmits visual information despite intrinsic cellular noise is a fundamental challenge in neuroscience. Neurons in the retina communicate with sparse, stochastic electrical spikes, vastly increasing energy efficiency compared to artificial networks, but introducing noise that threatens reliable information transmission. Remarkably, biological visual systems continue to outperform artificial ones in many complex tasks, suggesting evolved mechanisms that mitigate the detrimental effects of noise – possibly through correlated activity among retinal ganglion cells.
This project seeks to reveal whether, and how, correlations in neuronal variability improve information encoding while conserving energy. The investigation combines cutting-edge machine learning models and information-theoretic analysis of retinal activity. We leverage large datasets of retinal responses to visual stimuli acquired through electrophysiological recordings and retinal calcium imaging. Deep neural network and/or Gaussian processes models – augmented with recurrent layers – are used to capture and predict correlated spiking patterns among retinal cells. Recent information estimation techniques enable us to quantify both the encoded information and the metabolic savings achieved by noise correlations in neurons’ responses. Collaborations provide access to diverse datasets and methodological expertise, supporting rigorous and multidisciplinary research.
FR-10 Perturbative approach for studying information coding in the retina.
FR-10 – Perturbative approach for studying information coding in the retina.
IDV, Sorbonne University
Matthew Chalk – Visual information processing: neural coding and vision restoration
Each moment, the retina is flooded with visual input – yet only a tiny fraction of this information is transmitted to the brain. This selective transmission allows vision to remain highly efficient, with fewer cells and less energy than would otherwise be required. But how is this efficiency achieved? A leading hypothesis is that the visual system has evolved to transmit information about natural images efficiently, by prioritising surprising visual signals that cannot be predicted from their context. In this project, we will use modern machine learning methods to test this hypothesis in the retina. Specifically, we will investigate how, and whether, the visual system is adapted to the local statistics of natural images in order to efficiently encode visual signals.
In this project, we will construct mathematical models that predict how ganglion cells – the output neurons of the retina – respond to different patterns of visual stimulation. These models will be fitted and validated on electrophysiological data. In parallel, we will build generative models of natural image statistics (diffusion models). These will be used to generate predictions about how retinal responses should depend on the local statistics of visual images, under the constraint of maximising information transmission with limited energy. We will then compare these theoretical predictions with phenomenological models trained directly on retinal data.
FR-11 Neural circuit maturation in the larval zebrafish optic tectum: From functional imaging to connectomics.
FR-11 – Neural circuit maturation in the larval zebrafish optic tectum: From functional imaging to connectomics.
IDV, Sorbonne University
Starting later.