Loading SBML models

SBML loader and simulation

SBML models can be loaded and simulated as follows.

from jaxkineticmodel.load_sbml.sbml_model import SBMLModel
import jax.numpy as jnp
import pandas as pd


filepath = ("models/sbml_models/working_models/simple_sbml.xml")

# load model from file_path
model = SBMLModel(filepath)

#replaces assignment rules, boundary conditions, etc..

S=model._get_stoichiometric_matrix()

JaxKmodel = model.get_kinetic_model()

ts = jnp.linspace(0,100,2000)


#simulate given the initial conditions defined in the sbml
ys = JaxKmodel(ts=ts,
            y0=model.y0,
            params=model.parameters)
ys=pd.DataFrame(ys,columns=S.index)

SBML test-suite passed tests

While jaxkineticmodel is compatible with SBML, not all .xml files are compatible/possible to simulate. This depends on the SBML model level and version that it was written. We here report a comparison in number of passed models with a comparison to libroadrunner using the sbml-test-suite. Libroadrunner is a high-performance simulation engine for systems biology¹. We compare the simulations to the gold-standard dataset that is provided in the sbml-test-suite. We selected 500 models from the test-suite and filtered out steady state models.

Discrepancies are models where we compared the output from JaxKineticModel to the ground-truth data and observed differences between the time-series. Overall, libroadrunner simulates more than jaxkineticmodel due to some event rules not yet being implemented. In the future we hope to address all these edge-cases.

References