Anatomy of a Spike
1 language| Type | Interactive demo |
|---|---|
| Origin | Voytek Lab at UC San Diego |
| Developer | David |
| Written in | Python, spikeparam, DANDI, TypeScript |
| Tests | TypeScript fits verified against the Python spikeparam pipeline |
| Built with | Voytek Lab spike parameterization, run on DANDI:001776 |
| Hosted on | David-Internet/demos/spikes |
| Website | https://david-internet.vercel.app/demos/spikes |
Anatomy of a Spike is an interactive demonstration of spike-parameterization analysis using public marmoset patch-clamp recordings from DANDI:001776. It is hosted on David's Internet and runs fitted waveforms in the browser.[1]
The page lets visitors inspect a recorded sweep, fit individual action potentials, alter a generative waveform, and brush a population scatter of fitted spikes. The analysis uses the Voytek Lab's spikeparam model, for which David is a user rather than an author.
Overview
A cursor moves through a 50 kHz patch-clamp sweep as detected spikes are centered and windowed. The population view contains more than 2,600 fitted spikes from ten marmosets. Selecting points overlays their recorded waveforms with a mean and standard deviation display, alongside feature comparisons and group boxplots.
Spike parameterization
The main fit proceeds through a LOWESS-smoothed derivative, an inflection point, peak calipers, and a bounded exponential decay. For the alternate model, two skewed Gaussians reach r-squared of 0.999. Sliders for the fitted parameters regenerate a waveform over the recorded spike.
Development
The demo reworks a 2024 spike-parameterization project with public NWB data. Its TypeScript ports of the spikeparam patch and skewed-Gaussian fits are checked against the Python pipeline.TypeScript fits verified against the Python spikeparam pipeline.
See also
References
- ^
content/spikes/README.md, David's Internet.
- Interactive demos
- Neuroscience
- 2026 establishments