Enzax

Enzax is a library of automatically differentiable equations and solvers for modelling networks of enzyme-catalysed reactions, written in JAX.

Enzax provides straightforward, fast and interoperable access to the gradients of realistic metabolic network models, allowing you to incorporate these models in your MCMC and machine learning algorithms when you want to, for example, predict the effect of down-regulating an enzyme on the yield of a fermentation experiment.

Installation

pip install enzax

Usage

Find a kinetic model's steady state

from enzax.examples import methionine
from enzax.steady_state import get_steady_state
from jax import numpy as jnp

guess = jnp.full((len(model.balanced_species),), 0.01)

steady_state = get_steady_state(methionine.model, guess, methionine.parameters)

Find a steady state's Jacobian with respect to all parameters

import jax
from enzax.examples.methionine import model, parameters
from enzax.parameters import get_parameter_position
from enzax.steady_state import get_steady_state
from jax import numpy as jnp

guess = jnp.full((len(model.balanced_species),), 0.01)

jacobian = jax.jacrev(get_steady_state, argnums=2)(model, guess, parameters)
# Each parameter is one flat array, so ask the model's labels which column
# belongs to GNMT1's turnover number.
pos = get_parameter_position(model.parameter_labelling, "log_kcat", "GNMT1")
jacobian["log_kcat"][:, pos]

Array([-3.83561719e-07, -9.66801513e-06,  3.38183091e-10,  3.15564888e-09,
        5.28588199e-08], dtype=float64)

Load a kinetic model from an sbml file

from enzax.sbml import load_libsbml_model_from_url, sbml_to_enzax

url = "https://raw.githubusercontent.com/dtu-qmcm/enzax/refs/heads/main/tests/data/exampleode.xml"

libsbml_model = load_libsbml_model_from_url(url)
# or to load an sbml file from your computer:
# libsbml_model = load_libsbml_model_from_file(path_to_file)

model, parameters = sbml_to_enzax(libsbml_model)