janusbio 0.2.0

JANUS : Joint ANalysis for augmentation of clUSter specificity

JANUS / janusbio

JANUS is a matrix-based analysis toolkit for joint comparison of reference and target datasets. It provides three main analysis categories:

• contributions: reference and target contribution matrices
• differential: differential correlation and direction matrices
• covariation: sample-size-normalized covariation matrices

Installation

Install from PyPI

pip install janusbio

Install from Git

pip install "git+https://github.com/<YOUR_ORG_OR_USER>/<YOUR_REPO>.git"

Development install (editable)

pip install -e .[dev]

CLI Usage

The CLI accepts explicit large/reference and small/target datasets:

janusbio --large large.csv --small small.csv

Run mode

• --run all (default): compute all three categories
• --run contributions: compute only contributions
• --run differential: compute only differential correlation
• --run covariation: compute only covariation

Examples:

janusbio --large large.csv --small small.csv --run all
janusbio --large large.csv --small small.csv --run contributions

Save outputs

Use --save with one or more categories to save computed outputs as parquet files:

janusbio --large large.csv --small small.csv --run all --save contributions differential
janusbio --large large.csv --small small.csv --run covariation --save covariation --output output

If --save is omitted, JANUS computes results and prints a summary without writing files.

Validation rule: categories passed to --save must be part of what --run computes.

Output files

• contributions:
  • ref_contributions.parquet
  • target_contributions.parquet
• differential:
  • diff_corr.parquet
  • direction.parquet
• covariation:
  • ref_covariation.parquet
  • target_covariation.parquet

Python API

from janusbio import analysis, preprocessing, utils

utils.initialize({"preprocessing": {"drop_na": False}})

inputs = {
		"Large (REFERENCE)": {"path": "large.csv", "sort": "high"},
		"Small (TARGET)": {"path": "small.csv", "sort": "high"},
}

preprocessing.load_datasets(inputs)
merged, n_ref = preprocessing.merge_datasets()

ref_contr, target_contr = analysis.get_contributions(merged, n_ref)
diff_corr, direction = analysis.get_differential_correlation(merged, n_ref)
ref_cov, target_cov = analysis.get_covariation_contributions(merged, n_ref)

Tests

pytest

Build and Publish

Build distributions

uv build

Publish to TestPyPI

set TWINE_USERNAME=__token__
set TWINE_PASSWORD=<TEST_PYPI_TOKEN>
uv publish --index testpypi

Publish to PyPI

set TWINE_USERNAME=__token__
set TWINE_PASSWORD=<PYPI_TOKEN>
uv publish

After publish, verify install:

pip install janusbio
janusbio --help