logternary

Log-ternary plots for fold-change visualization between three conditions.

A log-ternary plot maps positive triples (a, b, c) — where only ratios carry meaning — to ℝ² via a symmetric isometric log-ratio transform. The resulting plot has three axes at 120° separation, each representing fold-changes in one condition relative to the geometric mean of the other two.

Installation

pip install logternary

Quick start

import matplotlib.pyplot as plt
import logternary  # registers the 'logternary' projection

fig, ax = plt.subplots(
    subplot_kw={'projection': 'logternary', 'base': 2, 'max_level': 3,
                'labels': ('Wild type', 'Knockout', 'Rescue')}
)
ax.scatter(a, b, c, color='steelblue', s=12)
plt.show()

The three-argument forms ax.scatter(a, b, c), ax.plot(a, b, c), and ax.annotate('label', a, b, c) are automatically transformed to log-ternary coordinates. Two-argument calls fall through to standard matplotlib behaviour.

Examples

Key properties

Mirrored points, collinear trajectories, and fold-change transformations:

Transcriptome data

Simulated gene expression data across three conditions, with differentially expressed genes highlighted:

Exchange rate trajectories

Monthly exchange rate trajectories visualized as a path through log-ternary space:

Configuration

All parameters are passed via subplot_kw:

Grid visibility: ax.grid(True) / ax.grid(False)

Development

git clone https://github.com/gatoniel/logternary
cd logternary
uv sync --dev
uv run pre-commit install
uv run pytest

License

MIT