Deep Representation Learning on Whole-Brain Population Dynamics Uncovers Geometrically Separable Neural Codes
- 1Institute of Zoology, University of Cologne, Germany
Abstract
Extracting interpretable representations from high-dimensional whole-brain neural dynamics remains a major computational challenge in systems neuroscience. Here, we develop a wiring-agnostic deep-learning framework that combines convolutional encoding with temporal transformers to learn compact representations directly from volumetric calcium imaging of the entire \textit{Drosophila melanogaster} brain, without neuronal identification or anatomical annotation. Applied to brain-wide activity recorded across 16 factorially combined sensory and internal-state conditions, the learned representations revealed a factorized organization of whole-brain dynamics: metabolic state, sensory modality, and stimulus valence emerged along three near-orthogonal axes from a classification objective based only on flat class labels and without explicit disentanglement constraints. Spatial attribution and brain region-level ablation analyses linked modality representations to anatomically distinct circuits, whereas state- and valence-related information showed a broadly distributed organization. Our approach provides a scalable framework for discovering interpretable representations of whole-brain neural dynamics.
Related
- Cross-species representation learning aligns mouse and human neural dynamics and tracks clinical drug efficacy
- Sparse component analysis: A method that uncovers separable computations within neural population activity
- Temporal resolution shapes inferred higher-order interactions and information representation in neural population activity: A data-driven approach using minimally complex models
- Toward a Transferable Representation of Protein Dynamics for Functional Prediction and Mechanistic Discovery
- High-dimensional hippocampus representation facilitates learning of low-dimensional abstract representation and supports generalization
BibTeX
@misc{abdelbaki2026deep,
title = {{Deep Representation Learning on Whole-Brain Population Dynamics Uncovers Geometrically Separable Neural Codes}},
author = {A. Abdelbaki and P. Bandow and K. Y. Cheng and I. C. Grunwald Kadow and M. P. Nawrot and V. Rostami},
howpublished = {bioRxiv},
year = {2026},
month = may,
doi = {10.64898/2026.05.12.724368},
url = {https://www.biorxiv.org/content/10.64898/2026.05.12.724368v4},
note = {CC BY},
}
RIS
TY - UNPB
TI - Deep Representation Learning on Whole-Brain Population Dynamics Uncovers Geometrically Separable Neural Codes
AU - A. Abdelbaki
AU - P. Bandow
AU - K. Y. Cheng
AU - I. C. Grunwald Kadow
AU - M. P. Nawrot
AU - V. Rostami
PY - 2026
DA - 2026/05/12
PB - bioRxiv
DO - 10.64898/2026.05.12.724368
UR - https://www.biorxiv.org/content/10.64898/2026.05.12.724368v4
AB - Extracting interpretable representations from high-dimensional whole-brain neural dynamics remains a major computational challenge in systems neuroscience. Here, we develop a wiring-agnostic deep-learning framework that combines convolutional encoding with temporal transformers to learn compact representations directly from volumetric calcium imaging of the entire \textit{Drosophila melanogaster} brain, without neuronal identification or anatomical annotation. Applied to brain-wide activity recorded across 16 factorially combined sensory and internal-state conditions, the learned representations revealed a factorized organization of whole-brain dynamics: metabolic state, sensory modality, and stimulus valence emerged along three near-orthogonal axes from a classification objective based only on flat class labels and without explicit disentanglement constraints. Spatial attribution and brain region-level ablation analyses linked modality representations to anatomically distinct circuits, whereas state- and valence-related information showed a broadly distributed organization. Our approach provides a scalable framework for discovering interpretable representations of whole-brain neural dynamics.
ER -