Neuronal classification from extracellular electrophysiological recordings is challenging due to intrinsic waveform variability, noise, and technical differences across experiments, technologies, and species. We introduce HIPPIE (High-dimensional Interpretation of Physiological Patterns In Intercellular Electrophysiology), a deep learning framework that combines self-supervised pretraining on unlabeled datasets with supervised fine-tuning to classify neurons from extracellular recordings. Using conditional convolutional joint autoencoders, HIPPIE learns technology-adjusted representations of waveforms and spiking dynamics. Here we show, across mouse, rat, and macaque recordings, that HIPPIE classifies cell types competitively with existing methods while additionally supporting generative analyses that discriminative models cannot perform, including counterfactual decoding of electrophysiological signals under changed experimental conditioning, cross-species latent interpolation, and a cross-modal analysis revealing that spike-timing modalities and waveform morphology encode largely independent dimensions of neuronal identity. HIPPIE is available as both a Python package and a coding-free web application, providing a unified framework for multimodal neuronal classification across technologies, experimental conditions, and species.
Extracellular recordings carry information about neuronal location, and pooling per-unit predictions across neighboring electrodes improves its readout, but whether this local pooling should be learned is untested. We present MYCEL (Message-passing Yields Cellular Embeddings of Location), a graph neural network whose nodes are spike-sorted units carrying waveform and spike-timing features selected on cell-type ground truth, and whose edges encode recording geometry. On the Allen Brain Observatory Visual Coding Neuropixels dataset, learned message passing outperforms fixed aggregation, with the largest gains where single units are least informative. On the International Brain Laboratory Brain Wide Map, MYCEL matches the strongest population model at one-tenth the pairwise interactions. On held-out animals, the embeddings recover laminar depth within visual cortex. Retrained with only the edge geometry changed, the same framework decodes anatomy in macaque visual cortex and human medial temporal lobe. Thus, how neighboring neurons are combined matters as much as the features that describe them.
SUMMARY: Brain organoids develop self-organizing neural networks that can be used to model neurodevelopment and disease. Here we introduce PULSE, a foundation model developed using over 900 hours of high-density microelectrode array recordings of in vitro and ex vivo neural tissue. We show that this single model can address two open questions: how disease-associated variants alter network development, and how naive circuits encode information. PULSE revealed accelerated network maturation in held-out human 16p11.2-deletion organoids. It also revealed a change in network coordination shared with SHANK3 deficiency. Because organoids develop without sensory input, they let us ask how circuits encode information before any representation is learned. Closed-loop control of a virtual inverted pendulum engaged existing spontaneous activity patterns rather than creating new ones. PULSE provides a common reference frame for neural circuits across genotypes, protocols and tasks.