AutoReduce: An Automated Model Reduction Toolbox

Python toolbox to obtain reduced model expressions using time-scale separation, conservation laws, sensitivity analysis, and projection-based interfaces to established reduction libraries.

Overview

AutoReduce is a Python package for automated model reduction of nonlinear dynamical systems. It provides tools for:

• Automated model reduction using QSSA (Quasi-Steady State Approximation)
• Conservation-law based reductions
• Local sensitivity analysis
• SBML import/export through python-libsbml
• Integration with BioCRNPyler for synthetic biology models
• Optional python-control and PyDMD interfaces

See the bioRxiv paper and International Journal of Robust and Nonlinear Control paper for the model-reduction background.

Quick Start

import numpy as np
from sympy import symbols

from autoreduce import System, solve_timescale_separation

S, C, P, k1, k2, k3, E_total = symbols("S C P k1 k2 k3 E_total")

E = E_total - C
x = [S, C, P]
f = [
    -k1 * E * S + k2 * C,
    k1 * E * S - (k2 + k3) * C,
    k3 * C,
]

system = System(
    x,
    f,
    params_dict={k1: 1.0, k2: 0.5, k3: 0.25, E_total: 1.0},
    x_init=[10.0, 0.0, 0.0],
    C=np.array([[1, 0, 0], [0, 0, 1]]),
)

reduced_system, collapsed_system = solve_timescale_separation(
    system,
    [S, P],
    fast_states=[C],
)

params_dict can be used to set, get, and update parameters:

system.set_param(k1, 2.0)
system.get_param(k1)
system.set_param_dict({k2: 1.0, k3: 0.1})

For more examples, check out the documentation.

Installation

Supported Python versions are 3.9 - 3.13.

Install the latest version of AutoReduce:

pip install autoreduce

Install with all optional dependencies:

pip install autoreduce[all]

Install optional integrations explicitly:

pip install "autoreduce[bio]"
pip install "autoreduce[control]"
pip install "autoreduce[dmd]"

For development installation:

git clone https://github.com/ayush9pandey/autoreduce.git
cd autoreduce
pip install -e ".[dev]"

Documentation

Full documentation is available at autoreduce.readthedocs.io.

Contributing

We welcome contributions. Developer notes and release instructions are in the documentation.

Contact

For questions, feedback, or suggestions, please contact:

• Ayush Pandey (ayushpandey at ucmerced dot edu)
• GitHub Issues

License

Released under the BSD 3-Clause License (see LICENSE)