Overview & Roadmap

This page is a map. It explains what braincell does, how the pieces fit together, and — most importantly — where to go next depending on what you want to accomplish.

What problem does braincell solve?

Computational neuroscience often needs to simulate the electrical behavior of neurons in biophysical detail: the flow of ions through voltage-gated channels, the spread of voltage along dendrites, and the resulting spikes. braincell provides this in a way that is:

• Detailed — from a single isopotential compartment to a full morphological reconstruction with hundreds of compartments.

• Fast and portable — every model compiles with JAX and runs on CPU, GPU, or TPU without code changes.

• Differentiable — gradients flow through whole simulations, so you can fit parameters with gradient descent.

The two kinds of cell

Almost everything in braincell is organized around two cell classes:

Class	Use it when…	Built from
braincell.SingleCompartment	the neuron can be treated as a single isopotential point (a “ball”), or you are prototyping channel dynamics.	ion species (braincell.ion) + ion channels (braincell.channel) added imperatively in __init__.
braincell.Cell	dendritic geometry matters — voltage attenuation, distributed channels, synaptic integration.	a Morphology decorated declaratively with braincell.mech mechanisms via paint / place.

Both inherit from braincell.HHTypedNeuron and share the same ion, channel, and integrator machinery — so concepts you learn for one transfer to the other.

The pipeline

A multi-compartment model moves through four conceptual stages. Understanding this pipeline makes the rest of the docs click into place:

  Morphology         Cell            Discretization        Runtime + Solver       RunResult
  (geometry)  ──▶  (declaration) ──▶ (control volumes) ──▶  (time integration) ──▶ (traces)
                   paint / place

1. Morphology — the geometry (branches, radii, connectivity), loaded from a file or built programmatically. See Morphology.

2. Declaration — you create a Cell and paint density mechanisms (channels, ions, cable properties) onto regions and place point mechanisms (clamps, synapses, probes) at locations. See Mechanisms and Regions & Locsets.

3. Discretization — the continuous cable is divided into control volumes (CVs) by a CVPolicy. See Discretization.

4. Integration — a solver from braincell.quad advances the system in time, producing a RunResult of probe traces. See Integration.

Single-compartment models skip the morphology and discretization stages: the cell is one compartment.

Where to go next

If you want to…	Go to
Run something immediately	First Steps
Understand the vocabulary and design	Architecture
Build and simulate point neurons	Modeling Tutorials
Build morphological cells	Modeling Tutorials
Pick or write a numerical solver	Numerical Integration
Load a morphology from a file	File Formats & IO
Look up an exact class or function	braincell module
Contribute or extend the library	Developer Guide
Diagnose an error	Troubleshooting & FAQ