Noisy Logic Bayesian Inference

A Bayesian Networks approach for inferring active Transcription Factors using logic models of transcriptional regulation.

Examples

The package includes several examples to help you get started:

Jupyter Notebooks

• TF-inference.ipynb - Basic TF inference workflow
• TF-inference-simulation.ipynb - Simulation-based inference example

Python Scripts

• tf_inference_simulation.py - Command-line example demonstrating TF inference with simulated data, including:
  • Random network generation
  • Evidence simulation
  • OR-NOR model inference
  • Performance evaluation (ROC/PR curves)
  • Terminal-based visualization

Installation Instructions

Prerequisites

The package requires the GNU Scientific Library (GSL). Install it for your operating system:

Linux (Debian/Ubuntu)

sudo apt install -y libgsl-dev

macOS

brew install gsl

Windows

Setup GSL using vcpkg:

1. install vcpkg:

git clone https://github.com/microsoft/vcpkg.git
cd vcpkg
./bootstrap-vcpkg.bat

2. Install GSL:

.\vcpkg install gsl:x64-windows

3. Set the environment variable VCPKG_ROOT to the path of the vcpkg directory:

• PowerShell:

$env:VCPKG_ROOT = "C:\path\to\vcpkg"

• Command Prompt:

set VCPKG_ROOT="C:\path\to\vcpkg"

Package Installation

Install nlbayes directly from GitHub:

pip install git+https://github.com/umbibio/nlbayes-python.git

Usage

from nlbayes import ORNOR

# Create and fit the model
model = ORNOR(network, evidence, n_graphs=5, uniform_prior=False)
model.fit(n_samples=2000, gelman_rubin=1.1, burnin=True)

# Get inference results
results = model.get_results()
print(results)

Input Data Format

The input data for the ORNOR class should be formatted as follows:

• network: A dictionary where keys are TF IDs and values are dictionaries mapping target gene names to regulation modes (-1 for repression, 1 for activation)
• evidence: A dictionary mapping gene names to expression states (-1 for down-regulated, 1 for up-regulated)

For more detailed examples, check the example notebooks and scripts in the nlbayes/examples directory.