cell-growth-sim

A simulation of single-cell growth and division under different size-control models, with a Rust simulation engine and a Python (marimo) analysis layer.

The biological question: how do cells decide when to divide? Three canonical strategies are implemented, plus a generalisation:

Model	Division rule	Intuition
Timer	divide after a fixed time τ	"wait a set interval"
Sizer	divide on reaching an absolute volume	"grow to a target size"
Adder	divide after adding a fixed volume increment	"add a constant amount, then split"
AdderAlpha	Δv ≥ α·v_birth + v_c — interpolates the above	α = 0 → adder, α = −1 → sizer

Cells grow exponentially (V(t) = V_b · e^(r·t)), divide with stochastic partitioning (split fraction ~ Normal(0.5, split_noise)), and inherit their mother's control model. Run-to-run reproducibility is guaranteed by a fixed RNG seed.

Repository layout

This is a monorepo — one git repository, two cooperating projects that communicate by file handoff:

cell-growth-sim/
├── sim/            Rust simulation engine (Cargo project)
│   ├── Cargo.toml
│   └── src/        cell.rs · config.rs · simulation.rs · main.rs
├── analysis/       Python analysis & visualisation (uv project, marimo)
│   ├── pyproject.toml
│   └── notebooks/
├── data/           simulation output consumed by the analysis layer
│   └── events.json
├── LICENSE
└── README.md

Data flow:

  sim/  ──writes──▶  data/events.json  ──reads──▶  analysis/
 (Rust)                                            (marimo / Polars)

The two halves are intentionally decoupled: the simulation knows nothing about the notebooks, and the notebooks just read the emitted JSON. Each division event records birth_volume, division_volume, added_volume, generation_time, generation, daughter_volumes, and time.

Running the simulation (Rust)

Requires a recent Rust toolchain (edition 2024).

cd sim
cargo run            # runs the sim, writes ../data/events.json
cargo test           # unit tests for the cell model and config

Parameters (growth rate, noise, seed, …) live in sim/src/config.rs. Per-cell division thresholds are derived from the fundamental parameters rather than set directly — configure r and V₀, and the timer period / sizer target / adder increment follow.

Analysis (Python)

Requires uv and Python 3.14+.

cd analysis
uv sync              # create the environment
uv run marimo edit   # open the notebook

The analysis layer uses Polars for data wrangling and Altair for plots, reading ../data/events.json.

Status

Active learning project (Rust). The AdderAlpha model is implemented in the cell logic but not yet wired into Config/main — see the todo! in sim/src/cell.rs.