Abstract
Large-scale in silico experimentation depends on the generation of connectomes beyond available anatomical structure. We suggest that linking research across the fields of experimental connectomics, theoretical neuroscience, and high-performance computing can enable a new generation of models bridging the gap between biophysical detail and global function. This Focus Feature on "Linking Experimental and Computational Connectomics" aims to bring together examples from these domains as a step toward the development of more comprehensive generative models of multiscale connectomes.
The Completeness Gap in Neuroscience
The central challenge in modern neuroscience is the "completeness gap." Currently, researchers rely on two primary, yet disconnected, data streams:
- Experimental Connectomics: This provides anatomical data regarding neuronal connectivity. However, these datasets are often incomplete, contain high levels of uncertainty, and cannot yet map every synapse in a living system.
- Functional Data: This captures brain activity, but the relationship between structure and function is not "uniquely invertible." Many different wiring diagrams could theoretically produce the same activity pattern, making it difficult to infer structure from function alone.
To run biologically realistic simulations—in silico experimentation—computational neuroscientists require high-resolution, complete connectomes. Bridging this gap requires moving beyond messy, incomplete structural data toward functional models that behave like biological brains.
Generative Connectomics: A Methodological Framework
Rather than presenting a single algorithm, this editorial outlines "Generative Connectomics"—a framework that uses computational rules based on development and plasticity to fill experimental blanks. Key approaches include:
- Simulated Development and Evolution: Instead of manual mapping, researchers use biological rules (e.g., neuronal growth and target finding) to "grow" synthetic connectomes that match known experimental constraints.
- Multi-scale Integration: Models must bridge the gap between local biochemical events (milliseconds) and global plasticity or behavior (hours to years).
- High-Performance Computing (HPC): Combining HPC with theoretical neuroscience is essential to handle the massive data requirements of large-scale neuronal network simulations.
- Interactive Tooling: Advanced visualization and simulators allow researchers to interact with and refine these generative models in real-time.
Key Insights and Clinical Utility
A core insight of this framework is the Structure-Function Loop: structure does not just dictate function; functional activity (plasticity) constantly rewires the structure. By discovering the rules the brain uses to wire itself, researchers can create models that are more robust than static maps.
The ultimate goal of this in silico approach is predictive clinical utility. By simulating brain function, researchers hope to predict the consequences of lesions or the impact of therapeutic interventions, such as brain stimulation, before they are performed on a patient.
Highlights from the Focus Feature
Several studies demonstrate the success of this interdisciplinary approach:
- Primate Organization: Hilgetag et al. (2019) identified an "architectonic type principle" capable of predicting primate brain organization across different scales.
- Therapeutic Recovery: Lu et al. (2019) used computational models to demonstrate how transcranial direct current stimulation (tDCS) triggers "cell assembly" formation, providing a mechanism for long-term recovery after injury.
- Cross-Species Application: Rojas et al. (2019) applied a wavelet-transform framework to analyze signals in the cockroach brain, proving these computational tools are applicable across diverse biological systems.
Future Directions
Despite progress, hurdles remain. Anatomical datasets still contain significant uncertainties, and human-scale "biologically realistic" simulations present massive technical challenges. Future work must focus on better visualization technology to allow scientists to steer the generation of neural networks. Most importantly, experimentalists and computationalists must work in a tighter feedback loop to ensure models remain grounded in biological reality.
Summary updated