signifikante 0.1.0

Scalable gene regulatory network inference using tree-based ensemble regressors with p-values

SignifiKANTE

SignifiKANTE builds upon the arboreto software library to enable regression-based gene regulatory network inference and efficient, permutation-based empirical P-value computation for predicted regulatory links.

Quick install

The tool is installable via pip and pixi

git clone [e-mail]:bionetslab/SignifiKANTE.git
cd SignifiKANTE
pip install -e .

To create a pixi environment, download pixi from pixi.sh, install and run

git clone [e-mail]:bionetslab/SignifiKANTE.git
cd SignifiKANTE
pixi install

Create jupyter kernel using pixi.toml/pyproject.toml, which will install a jupyter kernel using a custom environment (including ipython)

git clone [e-mail]:bionetslab/SignifiKANTE.git
cd SignifiKANTE
pixi run -e kernel install-kernel

FDR control

We provide an efficient FDR control implementation based on GRNBoost2, which computes empirical P-values for each edge in a given or to-be-inferred GRN. Our implementation offers both a full and a (more efficient) approximate way of P-value computation. An example call to our FDR control includes the following steps:

import pandas as pd
from signifikante.algo import grnboost2_fdr

# Load expression matrix - in this case simulate one.
exp_data = np.random.randn(100, 10)
exp_df = pd.DataFrame(data, columns=columns)

# Run approximate FDR control.
fdr_grn = grnboost2_fdr(
            expression_data=exp_df,
            cluster_representative_mode="random",
            num_target_clusters=5,
            num_tf_clusters=-1
        )

A more detailed description of all parameters of the grnboost2_fdr function can be found in the respective docstring.

License

This project is licensed under the GNU General Public LICENSE v3.0.