Macular: a multi-scale simulation platform for the retina and the primary visual system

Abstract: We developed Macular, a simulation platform with a graphical interface, designed to produce in silico experiment scenarios for the retina and the primary visual system. A scenario consists of generating a three-dimensional structure with interconnected layers, each layer corresponding to a type of 'cell' in the retina or visual cortex. The cells can correspond to neurons or more complex structures (such as cortical columns). The inputs are arbitrary videos. The user can use the cells and synapses provided with the software, or create their own using a graphical interface where they enter the constituent equations in text format (e.g., LaTeX). They also create the three-dimensional structure via the graphical interface. Macular then automatically generates and compiles the C++ code and generates the simulation interface. This allows the user to view the input video and the three-dimensional structure in layers. It also allows the user to select cells and synapses in each layer and view the activity of their state variables. Finally, the user can adjust the phenomenological parameters of the cells or synapses via the interface. We provide several example scenarios, corresponding to published articles, including an example of a retino-cortical model. Macular was designed for neurobiologists and modelers, specialists in the primary visual system, who want to test hypotheses in silico without the need for programming. By design, this tool allows natural or altered conditions (pharmacology, pathology, development) to be simulated.

• The paper introduces Macular as a multi-scale simulation platform that models the retina and primary visual system via layered cellular graphs and parameterized synaptic dynamics.
• It utilizes a user-friendly GUI to construct and customize detailed cell models, differential equations, and connectivity schemes without requiring programming expertise.
• The platform facilitates rapid hypothesis testing in computational neuroscience by combining realistic visual stimuli, flexible model configurations, and integrated in silico experiments.

Macular: A Multi-Scale Simulation Platform for the Retina and Primary Visual System

Introduction

The paper introduces Macular, an open-source, multi-scale simulation platform that addresses the complexities of retinal and primary visual system modeling. Designed for use by both neurobiologists and computational modelers, including those with limited programming expertise, Macular features a graphical interface enabling construction, parameterization, and simulation of interconnected layers representing cellular types in the retina and cortex. The platform facilitates hypothesis testing, in silico exploration, and the creation of physiologically grounded models informed by realistic visual stimuli.

Platform Architecture and Modeling Philosophy

Macular distinguishes itself by providing a high-level abstraction for biological structures. The central architecture comprises "Cells" (generalized neuronal or multi-unit objects) organized in spatially explicit, layered graphs. Each cell type is defined by its own set of differential equations, parameters, and state variables, all of which can be customized through the GUI or extended using the Macular Template Engine (MTE). This approach enables simulation scenarios encompassing a broad array of cellular mechanisms and synaptic interactions.

Visual information is input as movies, mapped to spatial input currents through receptive field models defined as spatio-temporal kernels. The default convolution employs Deriche filters, integrating the VirtualRetina framework for efficient OPL modeling. However, current limitations include reliance on spherical symmetry, restricting simulations with asymmetric or orientation-selective kernels.

Existing cell models in Macular span linear, rectified, and nonlinear elements, including classical circuits (e.g., Hodgkin-Huxley, Morris-Lecar), gain-modulated bipolar and ganglion cells, and mean-field columns for cortex. Layer composition is fully user-definable, supporting multi-layered, multi-population retino-cortical architectures, with spatial positions and connectivity explicitly specified in macular scenario files.

Synaptic Connectivity and Computational Flexibility

Synaptic dynamics are equally extensible, with both chemical and electrical (gap junction) synapses parameterized by their kinetics and transmission equations. Macular supports user-defined synapse types via the MTE, with instantaneous or delayed signal propagation (the latter using a customizable conduction velocity).

Connectivity schemes encompass one-to-one, nearest neighbor, radius-based, Gaussian (distance-dependent weights), and fully connected graphs, enabling rich spatial interaction motifs both within and across layers. This flexibility supports realistic modeling of local and long-range circuits, lateral inhibition, gain control, and recurrent cortical loops.

The core simulation engine leverages the GNU Scientific Library for ODE integration, offering a broad suite of fixed and adaptive solvers (RK4, Bulirsch-Stoer, Adams, BDF, etc.). Stochastic integration is not natively supported; extension for noise-driven models requires further development.

Usability, GUI, and Batch Operation

Macular is engineered for accessibility without code-writing. The GUI delivers tools for constructing layered architectures, parameter manipulation, selection of visual inputs, and direct inspection of cell and synapse dynamics. Scenario creation and modification—including the integration of new cell and synapse models with equations entered in LaTeX—do not require programming. The platform provides 2D and 3D visualization of activity across layers, and the direct monitoring of state variables during simulations.

For high-throughput or automated experimentation, Macular includes a headless batch mode for executing and recording simulation sessions defined in configuration files.

Example Simulations

The manuscript details usage scenarios that highlight Macular's scope:

• Retinal Waves: Emulation of spontaneous, wave-like activity in starburst amacrine cell (SAC) networks during development, as described in prior publications. This example demonstrates Macular's facility for developmental modeling and the flexibility to simulate intrinsic activity absent visual input.
• Retino-cortical Integration: Modeling of joint retina-V1 dynamics under video stimulation. The simulated architecture includes gain-controlled bipolar, amacrine, and ganglion cell layers, coupled to excitatory and inhibitory cortical populations operating under mean-field dynamics. This scenario exemplifies cross-modality modeling, enabling realistic in silico replication of experimental paradigms focused on cortical anticipation and motion processing.
• Custom Networks: By exposing the Wilson-Cowan and Amari mean-field models via the MTE, Macular supports rapid prototyping and analysis of canonical neural circuit motifs under complex spatio-temporal stimulation.

Comparison with Existing Simulators

Relative to existing retinal simulators (COREM, VirtualRetina, ISETBIO, PULSE2PERCEPT, etc.), Macular provides unique capabilities:

• A GUI-driven workflow targeted at users lacking programming skills.
• Integrated multi-scale modeling, bridging retina and early cortex.
• In situ parametric modulation and scenario creation with real video input.
• Direct code generation and compilation for new cell and synapse types.
• Versatile connectivity and physical units management.

However, Macular currently lacks support for color processing, multi-compartment neuron models, and direct stochastic integration. Kernel generalization (asymmetric, orientation-selective) is presently limited compared to platforms such as Convis.

Practical and Theoretical Implications

Practically, Macular enables faster, broader hypothesis testing for visual system mechanistic studies, including pharmacological interventions, developmental dynamics, prosthetic stimulation, and the exploration of connectivity rules. The abstraction of "Cells" and flexible graph definitions afford modelers the ability to construct both microcircuit and population-level models. The potential to interface Macular with large-scale cortical simulators (e.g., The VirtualBrain, NEST) opens pathways to hierarchy-spanning models of sensorimotor integration and cognition.

Theoretically, Macular provides a controlled platform to examine multi-scale interactions in visual processing, test the consequences of parameter perturbations, and validate dynamic behavior against experimental data. Its use may drive development of new analysis tools for high-dimensional model outputs and foster cross-disciplinary collaboration between experimental and computational neuroscientists.

Future Directions

Key avenues for extension include:

• Support for color information in OPL modeling and downstream layers.
• Implementation of multi-compartment and extended neuron models.
• Parallelization and GPU support for large-scale simulation.
• Generalization to non-symmetric, orientation-selective receptive fields.
• Integration with complementary simulators for full brain modeling.
• Enhanced noise modeling and stochastic integration support.

Conclusion

Macular represents a significant advancement in the accessibility and versatility of retina and early visual system simulation platforms. By bridging detailed biophysical modeling and user-friendly scenario construction, it empowers researchers to rapidly construct, test, and refine multi-scale models with realistic stimuli. While several limitations remain—in particular, the support for advanced synaptic and neuronal properties—the platform's extensibility and focus on practical experiment replication position it as a valuable tool in computational neuroscience and visual system modeling.

Source: "Macular: a multi-scale simulation platform for the retina and the primary visual system" (2512.13052).