Our team follows an interdisciplinary approach, combining modeling and experiments, to tackle several issues.
How does the retina process natural scenes ?
Decades of work on the retina have taught us a lot about how the retina processes visual stimuli. However, while we know a lot about it processes relatively simple stimuli, in many cases it is still unclear how ganglion cells process complex, natural stimuli. This is because natural stimuli tap into specific non-linear processes in the retinal circuit. Our purpose is to develop models to predict how ganglion cells, the retinal output, respond to natural scenes, and understand what features they extract. As an example, recently we have developed a perturbative approach to tackle this issue, where we add to a natural image a small noise pattern and study systematically how this noise pattern changes the response of ganglion cells.
How do ganglion cells code information together ?
The activity of the different neurons of the retina is noisy and correlated. There is at the moment no consensus on the extent and purpose of the correlations observed in the population response, some studies stating that they can be beneficial whilst others showing otherwise. We aim at bridging the gap between these studies by using techniques from deep learning, experimental neuroscience and information theory to investigate extensively how this correlated noise affects the way the retina encodes visual information.
What is the retina good for ?
A long-standing challenge is to work out which functions are performed by different circuits, and how. For example, an influential theory – ‘efficient coding’ – posits that low-level sensory neural circuits have evolved so as to transmit maximal information about relevant sensory signals. But what signals are ‘relevant’ to the organism and what are the constraints faced by a given neural circuit? The answer to these questions are not given by the theory; they must be inferred from data. We work on developing ways to do this.
What are the circuits underlying retinal computations ?
To understand the role of specific cell types in the intermediate layers of the retina, we developed a combination of tools to stimulate them and record the impact of this stimulation in ganglion cells. For this we make specific cell types light sensitive with optogenetics, and use a technique called computer generated holography to stimulate them with cellular resolution. We now are using this tool to understand the specific role of specific types in shaping the computations performed by ganglion cells.
Retina across evolution: the case of the mouse lemur
If the retina is adapted to optimally process its natural input, then we should find meaningful differences between the retina of different animals living in different environments. We have started to address this issue at the level of the mouse lemur. This nocturnal primate clearly relies on vision for a variety of behaviours. Comparing the function of various ganglion cells in this primate versus other species should be a way to elucidate how retinal function matches the natural environment across evolution.
Towards clinical applications…
While we are primarily a group focused on fundamental research, we have several projects that could potentially lead to interesting clinical applications in the long term.
Vision restoration using optogenetics
Blindness affects 45 million people worldwide. In many cases of inherited retinal degeneration, photoreceptors are lost but retinal ganglion cells as well as many interneurons are spared. This opens the possibility to stimulate the remaining cells directly to restore visual function. Retinal prostheses are a promising solution and have been found to restore some useful perception in blind patients. However, the acuity of the existing devices remains very low, below the level of legal blindness, and therefore are often not sufficient to identify objects or to navigate in complex environments. Optogenetic therapies provide a possible alternative to restore vision with a higher resolution and specificity that can better mimic the natural output of the retina. In this strategy, a light sensitive protein is expressed in targeted neural populations of a blind retina. Expressing light sensitive proteins in ganglion cells can be a way to restore vision through the stimulation of these newly light-sensitive cells with patterned light to produce visual perception, although the first results show that the acuity is still low with this strategy. We are currently exploring an alternative strategy to restore richer functional selectivity, and possibly improve perceptual performance, by targeting cell types in the intermediate layers of the retina that are not affected by retinal degeneration.
Engineering novel tools for gene therapy with Machine Learning
For successful therapy in inherited retinal degenerations, we need to be able to access relevant cell populations of the retina with non-invasive yet precise viral vectors to investigate, prevent, slow-down, or even reverse the course of blinding diseases. An effective tool for gene delivery to the retina is adeno-associated virus (AAV), but it has not been optimized for gene delivery to the human retina. In collaboration with the team of Deniz Dalkara, we combine directed evolution, deep-sequencing and machine learning to engineer new viral vectors for more efficient gene delivery to the human retina for clinical gene therapy.
Myopia
The global burden of myopia is growing, affecting nearly 30% of the world population in 2020 and expected to rise to 50% by 2050. For a significant fraction of myopic patients, there is an associated risk of sight threatening myopia-related pathologies, such as glaucoma, myopic macular degeneration and retinal detachment. Myopia is due to excessive eye growth. Interestingly, it has been established across the literature that eye growth is strongly affected by the visual input that enters the eye and reaches the retina. Recently, several studies have shown that it is possible to use this feature to slow down eye growth, and therefore myopia progression. However, it is still unclear how different optical transformations of the visual input can differently modulate eye growth. To advance our understanding, our first goal is to characterize how the retina responds to optical transformations of natural stimuli using both experiments and modeling.
