EUVIMEDEUVIMED entdecken
Crossref

Robust Embeddings of Genetics, Anatomy, and State Decoded from a Neuron’s Activity

Aidan M Schneider

Open Scholarship Institutional Repository (Washington University in St. Louis) · 29.09.2025

DOI: 10.7936/kka8-1a55 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Open Scholarship Institutional Repository (Washington University in St. Louis), erschienen am 29.09.2025. Verfasst von Aidan M Schneider.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Neural spike trains are shaped by both extrinsic inputs and intrinsic structural constraints. While prior work has focused on how spike timing encodes externally driven variables, such as stimuli or behavior, this dissertation explores whether principles that organize neural activity in space and time—such as genetic cell type, anatomical location, and arousal state—are also embedded in the spiking output of individual neurons. I hypothesize that these organizing principles are not quenched as irrelevant variability, but rather multiplexed within the neural code itself. To test this, I attempt to decode these principles from the spike trains of neurons recorded in behaving mice, and show that, like signals of stimulus or behavior, these principles are reliably embedded in the neural code. I demonstrate that the spike timing of individual neurons carries enough information to classify: (1) genetic cell type, using a novel deep-learning architecture (LOLCAT) on datasets from the Allen Institute; (2) anatomical location, across diverse brain regions, structures, and cortical layers in datasets spanning multiple labs and behavioral paradigms; and (3) arousal state, revealing a surprisingly local scale of this presumed global signal in freely-behaving mice. These signatures are consistently observed across neurons, indicating that any individual neuron may carry this embedded information. These findings expand beyond a moment-to-moment, stimulus-and-response-centric view of the neural code, revealing that individual spike trains carry high-dimensional, temporally embedded signatures of the principles that organize the brain. The spike train reveals not just what the brain does, but what it is. In doing so, this work reframes neural coding as the simultaneous representation of both function and form—suggesting that these embedded signatures of structure and state may be foundational, though often overlooked, dimensions of the neural code.

Schlagwörter

spike trainsspike timing codingneural codedeep learning decodingbehaving micecortical layers

Zitieren

APA

Schneider, A. M. (2025). Robust Embeddings of Genetics, Anatomy, and State Decoded from a Neuron’s Activity. Open Scholarship Institutional Repository (Washington University in St. Louis). https://doi.org/10.7936/kka8-1a55

Vancouver

Schneider AM. Robust Embeddings of Genetics, Anatomy, and State Decoded from a Neuron’s Activity. Open Scholarship Institutional Repository (Washington University in St. Louis). 2025. doi:10.7936/kka8-1a55

BibTeX
@article{Schneider2025,
  author = {Schneider, Aidan M},
  title = {Robust Embeddings of Genetics, Anatomy, and State Decoded from a Neuron’s Activity},
  journal = {Open Scholarship Institutional Repository (Washington University in St. Louis)},
  year = {2025},
  doi = {10.7936/kka8-1a55}
}
Kostenlos starten →