BNMPy: Boolean network modeling and optimization

A Python library for Boolean Network (BN) and Probabilistic Boolean Network (PBN) modeling, simulation, optimization, and analysis with applications in systems biology.

See https://ilyalab.github.io/BNMPy/ for detailed documentation.

Installation

To install, run pip install -e . in this directory.

Core functionalities

1. Boolean network

Basic operations

• Network loading: Load networks from text files, SBML files, or string representations
• Network construction: Build networks from connectivity matrices and Boolean functions
• Network manipulation: Knockout/knockdown specific nodes, rename nodes, merge networks
• Network visualization: Interactive network visualizations with igraph and matplotlib

Simulation

• Deterministic update: Synchronous and asynchronous update schemes
• Stochastic update: Add noise to represent biological uncertainty
• Attractor analysis: Find steady states and limit cycles

2. Probabilistic Boolean network

Basic operations are similar to BNs.

• Stochastic Simulation: Run Monte Carlo simulations over steps
• Attractor analysis: Find steady states (state distributions) via Monte Carlo or Two-state Markov Chain (TSMC) methods

3. Knowledge Graph (KG) augmentation

BNMPy can build and extend models using the SIGNOR knowledge graph.

• Build KG-derived BN: Build a Boolean Network (BN) directly from a Knowledge Graph (KG)
• Merge models: Merge an original BN with a KG-derived model into either a BN or a PBN
• Targeted extension to PBN: Extend an existing BN by adding nodes and rules informed by KG
• Phenotype scoring from KG paths: Compute phenotype scores using gene-phenotype relations from KG

4. PBN Optimization

Parameter optimization

• Objective function: MSE-based optimization for experimental data fitting
• Optimization algorithms: Differential evolution, particle swarm optimization
• Discrete optimization: Support for discrete parameter spaces
• Result Evaluation: Compare predictions with experimental observations
• Model compression: Simplify models while preserving behavior

Examples

Tutorials:

• BN_simulation.ipynb: Basic Boolean network simulation
• PBN_simulation.ipynb: Probabilistic Boolean network simulation
• knowledge_graph.ipynb: Build from SIGNOR, extend, and merge into BN/PBN
• phenotype_score.ipynb: Calculate phenotype scores using simulation results
• Optimization.ipynb: Parameter optimization with experimental data
• workflow_example.ipynb: Complete workflow from data to optimized model

Applications:

• Optimization_Eduati2020_reducedData.ipynb: Application on a pancreatic cancer model (Eduati 2020)
• AML example

Other utilities:

• BN_PBN_steady_state.ipynb: Steady state analysis
• BN_compression.ipynb: Model compression and simplification

Quick Start

sys.path.append('./src')
import BNMPy

# Load a network (auto-detects BN vs PBN, file vs string)
network_string = """
A = A
B = C
C = !E
D = A | B
E = C & D
F = !A & B
"""
# initial state (array or dictionary)
x0 = [0, 0, 0, 0, 0, 0]  # array: order matches gene order in string
# or
x0 = {'A': 0, 'B': 1, 'C': 0, 'D': 0, 'E': 0, 'F': 0}  # dict: keys are gene names
network = BNMPy.load_network(network_string, initial_state=x0)

# Visualize the network
BNMPy.vis_network(network, output_html="SimpleBN.html", interactive=True)

# Simulate with noise = 0.05 for 10 steps
network.update_noise(p=0.05, iterations=10)

# Calculate steady states
calc = BNMPy.SteadyStateCalculator(network)
steady_state = calc.compute_steady_state(n_runs=20,n_steps=10000)

Note that when assigning initial states or obtaining steady states, gene order matches their order in the network string/file, and can be obtained by calling network.nodeDict. Currently it support both array input and dictionary input (genes not specified default to 0.)

Documentation

• Functions: https://ilyalab.github.io/BNMPy/ (in development)
• Tutorials: Check the Examples directory for tutorials
• Optimization Guide: See Optimizer README for advanced optimization features