SKiMpy: Symbolic Kinetic Models in Python

SKiMpy is a python package bridging implementing an efficient kinetic model-ing toolbox to build and analyze large-scale kinetic models for various biological domains such as signaling, gene expression, and metabolism. Furthermore, we demonstrate how this toolbox is used to parameterize kinetic models around a steady-state reference efficiently. Finally, we show how SKiMpy can implement multispecies bioreactor simulations to assess biotechnological processes.

• Non-linear ordinary equations for large scale kinetic models
• Steady state consistent parameter sampling(see ORACLE)
• Metabolic control analysis
• Modal analysis
• Uncertainty propagation in metabolic control
• Multispecies bioreactor modeling

Container-based install

You might want to use this program inside of a container. The docker/ subfolder has all the necessary information and source files to build it locally.

Also you can directly pull the docker image: docker.io/danielweilandt/skimpy Note that if you want to use a comercial solver such as CPLEX or GUROBI you will need rebuild the docker/ repository.

Setup

This step is not required if you're using the container, which bundles all this.

If you prefer a local installation of skimpy we provide a conda package bundeling binary depdendecies for linux and OSX operatings systems. Windows users can install the conda packages using a linux subsystem via WLS.

A comprehensive explanation to install anaconda/miniconda within the WLS linux subsystem can be found here.

We arre currently in the process of deploying SKiMpy and its depdency pyTFA to conda forge for a fullly atomated installation. In the meantime you can download current version of the respective packages here.

Then install pytfa and skimpy using the local source:

# Add conda-forge / bioconda channels (required)
conda config --add channels bioconda
conda config --add channels conda-forge
conda config --set channel_priority strict


# Create a new environment (highly recommended since currently only python 3.8 is supported for conda)
conda create --name skimpy-env python=3.8
conda activate skimpy-env

conda install weilandtd::skimpy

WIP: the conda package package will currenly print an error message upon plotting with bokeh although it produces the desired file see #11.

Alternatively you can install this module from source using pip: For Python 3, you might have to use pip3 instead of pip

git clone https://github.com/EPFL-LCSB/skimpy.git /path/to/skimpy
pip3 install -e /path/to/skimpy

Installation of these packages on the native windows system is challenging we thus recommend windows users to install the package only with in linux subsystem using WLS.

To use the ODE integration features scikit-odes is required to be installed beforehand following the instructions found here. To use the 'cvode' solver from the scikit-odes packages we strongly recommend to install the sundials solvers as ODE integration of large ODE system can be slow with python implemented solvers see benchmark here.

Installation from source has been tested on Ubuntu 21.10 (@eladnoor) the additional packages can be installed using:

sudo apt install gfortran libsundials-dev libflint-dev libgmp-dev

Windows users using WSL can install these dependencies in a similar fashion after starting the subsystem console.

Requirements

You will need to have Git-LFS in order to properly download some binary files:

git clone https://github.com/EPFL-LCSB/skimpy.git /path/to/skimpy
cd /path/to/skimpy
git lfs install
git lfs pull

This module was developed in Python 3.9, and it is recommended to run Python 3.9. The module also was tested in Python 3.8.

Further the following pip-python packages are required

• sympy >=1.1
• pytest
• scipy
• numpy
• bokeh
• pandas
• Cython
• markupsafe <=2.0.1
• bokeh >=0.12.0
• scikits.odes ==2.6.3
• deap
• dill
• h5py
• escher
• pytfa
• cobra <=0.24.0

The installation requires additionaly the following libraries:

• gcc
• gfortran
• libsundials-dev
• libflint-dev
• libgmp-dev

Further more using the escher plot and aninmation functions (skimpy.viz.escher) requires a Chrome installation. An installation script for linux systems can be found in docker/utils/install_chrome.sh

To use the ORACLE method for large-scale metabolic networks it is recommended to install a commercial solver package such as CPLEX or GUROBI. Note that SKiMpy requires a python version >= 3.7 the default docker version is currently 3.9. It is important that your solver supports your installed python version. We recommend installing CPLEX Studio221 which supports python versions: 3.7, 3.8, 3.9 and 3.10. A point by point installation guide for OSx and Linux is found here.

Quick start

To get right into building kinetic models please find below a simple example to get started:

import numpy as np
from skimpy.core import *
from skimpy.mechanisms import *

name = 'pfk'
metabolites = ReversibleMichaelisMenten.Reactants(substrate = 'A',
                                                   product = 'B')

parameters = ReversibleMichaelisMenten.Parameters(
    vmax_forward = 1.0,
    k_equilibrium=2.0,
    km_substrate = 10.0,
    km_product = 10.0,
    total_enzyme_concentration = 1.0,
)


pfk = Reaction(name=name,
               mechanism = ReversibleMichaelisMenten,
               reactants=metabolites,
               )

this_model = KineticModel()
this_model.add_reaction(pfk)
this_model.parametrize_by_reaction({pfk.name:parameters})
this_model.compile_ode(sim_type = QSSA)

this_model.initial_conditions['A'] = 1.0
this_model.initial_conditions['B'] = 1.0

this_sol_qssa = this_model.solve_ode(np.linspace(0.0, 100.0, 1000), solver_type='cvode')

this_sol_qssa.plot('output/uni_uni_base_out_qssa.html')

More information can be found here.

License

The software in this repository is put under an APACHE-2.0 licensing scheme - please see the LICENSE file for more details.