SpikeLab Manuscript — Figure Code

Code to reproduce all figures from the manuscript: "SpikeLab: Agentic tools for spike data analysis".

Figures

Figure	Description	Script
Fig. 2	Benchmark: plain models vs SpikeLab-assisted analysis	fig2/
Fig. 3	Cross-species recordings: mouse, human, and organoid	fig3/
Fig. 4	Neuronal firing and burst properties across diazepam doses	fig4/
Fig. 5	Pairwise correlations and network structure	fig5/
Fig. 6	Burst structure and reproducibility	fig6/
Fig. 7	GPLVM latent state analysis of burst dynamics	fig7/

Structure

Each figure directory contains:

• compute.py — data processing and intermediate result computation
• panels.py — individual panel plotting functions
• assemble.py — multi-panel figure assembly

Shared utilities at the root:

• load_recordings.py — loads spike-sorted recordings into an AnalysisWorkspace
• compute_shared.py — shared computations used across multiple figures
• plot_config.py — matplotlib style configuration (eLife formatting, color palette, figure sizing)

Prerequisites

• Python 3.10+
• SpikeLab (pip install -e SpikeLab/)
• matplotlib, numpy, scipy, scikit-learn, Pillow
• Spike-sorted recording data (SpikeData .pkl files)

Usage

Run from the repository root:

# Compute intermediate results for a figure
python -m fig3.compute

# Assemble the final figure
python -m fig3.assemble

Figures that depend on shared computations should run compute_shared.py and the figure-specific compute.py before assembly.

SpikeLab Skill

The spikelab-skill/ directory contains the distributable SpikeLab skill for agentic coding tools. Skill files are plain markdown documents that can be used with any agentic coding tool that supports custom instructions — simply provide the skill file as context. See spikelab-skill/INSTALL.md for library installation instructions.