FreeFlux is a Python package designed for 13C metabolic flux analysis of biological systems at isotopic steady state or transient state, making it suitable for both heterotrophic and autotrophic organisms. With FreeFlux, you can:
- Estimate metabolic fluxes
- Simulate labeling patterns of metabolite (fragments)
- Conduct constraint-based optimizations such as flux balance analysis and flux variability analysis
Our goal is to increase the accessibility of 13C fluxomics techniques to researchers in the metabolic phenotyping and engineering community.
To get started, check out our documentation, which provides an overview of FreeFlux's fundamental functions using a toy model. We've also included tutorials for practical models of E. coli and Synechocystis.
FreeFlux now supports 15N labeling data, which can be enabled by specifying nitrogen as the labeled atom when defining the labeling strategy.
FreeFlux was tested in Python 3.7, 3.8, 3.9 and 3.10. It can be installed using pip from PyPI:
python -m pip install --upgrade pip
pip install freefluxor from source (assuming you have git installed):
git clone https://github.com/Chaowu88/freeflux.git /path/to/freeflux
pip install /path/to/freefluxInstallation within an virtual environment is recommendated.
If FreeFlux is installed on a Linux platform with Python>=3.8, using JAX can significantly speed up the preparation step before flux estimation. JAX can be installed using the following command:
pip install "jax[cpu]>=0.4,<=0.4.18"FreeFlux requires the numerical optimization framework OpenOpt for nonlinear regression. It can be installed by the following commands:
pip install openopt
pip install FuncDesignerNote that OpenOpt is known to work well in Python 3.7, but may have compatibility issues in Python 3.8 and above. FreeFlux now automatically applies the necessary patches to ensure compatibility. If you prefer to apply the patch manually, please refer to this link for solutions.
FreeFlux uses the modeling language Pyomo to formulate linear optimization problem. By default, solvers are not installed together with Pyomo and should be installed independently. For example, to install the glpk solver, run the following command:
conda install -c conda-forge glpkA typical workflow with FreeFlux starts by building a model through reading metabolic reactions with atom transitions. Users can then call a handler, such as the fitter, simulator, or optimizer, to perform flux estimation, labeling pattern simulation, or constraint-based flux analysis, respectively. Various methods are provided for these handlers for data input and computation.
Below is an example script that performs flux estimation at steady state using the toy model:
MODEL_FILE = 'path/to/reactions.tsv'
from freeflux import Model
model = Model('demo')
model.read_from_file(MODEL_FILE)
with model.fitter('ss') as fit:
fit.set_labeling_strategy(
'AcCoA',
labeling_pattern = ['01', '11'],
percentage = [0.25, 0.25],
purity = [1, 1],
label_atom = 'C'
)
fit.set_flux_bounds('all', bounds = [-100, 100])
fit.set_measured_MDV(
'Glu_12345',
mean = [0.328,0.276,0.274,0.088,0.03,0.004],
sd = [0.01,0.01,0.01,0.01,0.01,0.01]
)
fit.set_measured_flux('v1', mean = 10, sd = 1)
fit.prepare()
res = fit.solve()For more information, please refer to the documentation.
NumPy's underlying BLAS/LAPACK implementation may automatically utilize multiple CPU cores to improve computational efficiency. It can introduce additional overhead and may cause high system CPU usage, especially when using the solve_with_confidence_intervals method for confidence interval estimation. In such cases, it is recommended to explicitly limit the number of threads. For example, if OpenBLAS is used:
export OPENBLAS_NUM_THREADS=1FreeFlux is released under the GPL version 3 license, please see here for more details.
Feel free to provide feedback at chao.wu@nrel.gov or chaowu09@gmail.com.