Decoding Cortical Microcircuits: A Generative Model for Latent Space Exploration and Controlled Synthesis

TL;DR

A generative model is introduced, to learn this underlying representation from detailed connectivity maps of mouse cortical microcircuits, and it is found that specific, interpretable directions within this space directly relate to understandable network properties.

Abstract

A central idea in understanding brains and building artificial intelligence is that structure determines function. Yet, how the brain's complex structure arises from a limited set of genetic instructions remains a key question. The ultra high-dimensional detail of neural connections vastly exceeds the information storage capacity of genes, suggesting a compact, low-dimensional blueprint must guide brain development. Our motivation is to uncover this blueprint. We introduce a generative model, to learn this underlying representation from detailed connectivity maps of mouse cortical microcircuits. Our model successfully captures the essential structural information of these circuits in a compressed latent space. We found that specific, interpretable directions within this space directly relate to understandable network properties. Building on this, we demonstrate a novel method to controllably generate new, synthetic microcircuits with desired structural features by navigating this latent space. This work offers a new way to investigate the design principles of neural circuits and explore how structure gives rise to function, potentially informing the development of more advanced artificial neural networks.

Figures (28)

• Figure 1: Overall model structure.
• Figure 2: Generated Results Demonstrate the Authenticity of Our Model
• Figure 3: SHAP analysis for clustering coefficient. The results of SHAP analysis show the entangled relationship between the latent dimensions and the graph metrics.
• Figure 4: t-SNE visualizations of the latent space reveal that variations in specific topological metrics correspond to distinct gradient directions within the embeddings. The darkness of each point's color corresponds to the magnitude of its associated bin index for the specific metric being considered. To further aid visualization, the red line overlaid on the t-SNE plot represents the direction of metric variation, obtained by fitting an auxiliary 2D linear regression model to the 2D t-SNE embeddings. This 2D regression is solely for visual convenience and does not replace the 32-dimensional analysis described in the main text.
• Figure 5: Metrics of generated graphs when setting different targets (4 of 6 target graph metrics).