Installation¶

sbmlsim needs python 3.14 or newer and runs on Linux, macOS and Windows. Install it from PyPI with pip:

pip install sbmlsim

The simulator libroadrunner comes as a binary wheel for all three platforms, so you need no compiler. The one exception is python 3.15: h5py, which writes the result files, has no wheels for it yet, so pip builds it from source. That needs a C compiler and the HDF5 library (libhdf5-dev on Debian and Ubuntu, brew install hdf5 on macOS).

With uv¶

With uv, add sbmlsim to your project, which locks it together with the rest of your environment:

uv add sbmlsim

or install it into a virtual environment:

uv venv --python 3.14
uv pip install sbmlsim

Check the installation¶

Import the package and run a short simulation of one of the models that ship with it:

import sbmlsim
from sbmlsim.resources import REPRESSILATOR_SBML
from sbmlsim.simulation import Simulation
from sbmlsim.simulator import Simulator

print(sbmlsim.__version__)
res = Simulator().run(REPRESSILATOR_SBML, Simulation(end=100))
print(res)

Output

0.18.2
ScanResult(_point: 1084; PX, PY, PZ, X, Y, Z, eff, n, ... (41 variables))

Now continue with the Tutorial.

Optional packages¶

Five features need packages which are not installed with sbmlsim:

feature	install
interactive HTML figures of experiments (`figure_formats=["html"]`, see Plots and reports)	`pip install plotly`
hybrid models with neural networks (PEtab SciML)	`pip install "sbmlsim[sciml]"`
COPASI as a second simulation engine (`Simulator(engine="copasi")`, see COPASI)	`pip install "sbmlsim[copasi]"`
AMICI as a third simulation engine (`Simulator(engine="amici")`, see AMICI)	`pip install "sbmlsim[amici]"`
the analyses as task types of simrun, which runs them on a cluster (see simrun)	`pip install "sbmlsim[simrun]"`

AMICI has no wheels and is built from source when it is installed. It needs a C++ compiler, a BLAS library with its cblas header and the headers of HDF5 from the system, e.g. on Debian or Ubuntu:

sudo apt install build-essential libopenblas-dev libhdf5-dev

CMake and SWIG come from PyPI with the build of AMICI, and the extra installs ninja, with which AMICI compiles a model.

Development version¶

Install the current state of the develop branch from GitHub:

pip install git+https://github.com/matthiaskoenig/sbmlsim.git@develop

or with uv:

uv add "sbmlsim @ git+https://github.com/matthiaskoenig/sbmlsim.git@develop"

To work on sbmlsim itself, with the tests and the documentation, see Development.

Dependencies¶

pip installs these packages together with sbmlsim:

package	used for
libroadrunner	simulating the SBML models
sbmlutils, python-libsbml	reading, validating and changing SBML models
pymetadata	COMBINE archives of PEtab problems
numpy, pandas, xarray, scipy, sympy	numerics, data and results
h5netcdf, h5py, hdf5plugin	the netCDF files of the results, compressed with zstd
sbmlmath	the math of SBML as sympy expressions
pint, pydantic	units and validated data
petab, pyyaml, SALib	PEtab problems and global sensitivity analysis
pkpdutils	non-compartmental pharmacokinetic analysis
matplotlib, seaborn, jinja2, rich	figures, reports and console output
certifi	the certificates with which sbmlutils and pymetadata reach the network

Logging¶

sbmlsim logs its messages with the logging module of the standard library, under the logger sbmlsim, and never configures logging itself. Your application decides what is shown:

import logging

logging.basicConfig(level=logging.INFO)
logging.getLogger("sbmlsim").setLevel(logging.WARNING)

In scripts and interactive sessions you can switch on the colored output of rich:

from sbmlsim import log

log.enable_rich_logging()