Model code for Cu Efflux in Halobacterium salinarum

Developed by

Wyming Lee Pang, PhD at the Institute for Systems Biology

Overview

The following is code that defines a quantitative model of copper efflux in the halophillic archaea Halobacterium salinarum, with specific focus on how the dynamics of gene expression of efflux machinery and intracellular copper levels change in response to varying levels of metallochaperone expression.

A publication that for which this model was created and provides more scientific detail is currently under review at PLoS Computational Biology.

Requirements:

R (2.11 or higher)
R packages
- deSolve

It is assumed that users have a basic understanding of the statistical computing environment R (freely available for download).

In order to perform simulations users must first install the deSolve package:

install.packages('deSolve')

Basic Simulation

For an example of basic simulation code, refer to the file:

odeCuEfflux.R

Before beginning, utility functions must be imported from:

CuEfflux_Func.R

which provides:

dxdt() : the differential step function
.T() : for the calculation of 'total' abunance of a species
setStoic() : sets individual reaction stoichiometries
set.a(), set.nu() : creates the propensity/rate vector a, and stoichiometry matrix nu from reaction definitions (see below)

The model is defined as reaction network that is simulated as a system of ordinary differential equations.

This definition is separated into three files:

CuEfflux_GlobalParams.R
CuEfflux_Init.R
CuEfflux_rxnDef.R

which are imported into the simulation code using source().

GlobalParams: specifies parameters used throughout the model. This includes all default reaction specific rate constants, and boolean flags that determine the type of genetic background being simulated. After import, a named vector called params will be added to the workspace.

Init: specifies initial conditions for all species. This file also initializes parameter values based on the 'modelName' parameter defined in GlobalParams. For example, if a reaction is not used in a model, its reaction rate is set to a value of 0. After import a named vector called x0 will be added to the workspace and the vector params will have been modified accordingly.

rxnDef: specifies all reactions used by all models. After import, a named list called rxn will be added to the workspace. Each element of rxn will have the following structure:

rxn$ReactionName
 .. $ a  : chr()
 .. $ nu : num()

where

$a is a character string defining the reaction rate law
$nu is a numeric vector whose names are the species involved in the reaction and values are their respective stoichiometries. The stoichiometry convention is reactants < 0, products > 0.

After reactions are defined, the rate vector a and stoichiometry matrix nu are created by calling:

a = set.a(rxn)
nu = set.nu(rxn)

The elements of the rate vector a are the defined rate laws from rxn and retain the reaction names defined in rxn.

The rows and columns of nu correspond to species and reactions, respectively, and are named accordingly.

Simulations are carried out via a call to ode() (from the deSolve package):

out = ode(x0, 
          seq(0,18000,by=100), 
          dxdt, 
          c(list(nu=nu, a=a), parms), 
          method='daspk')

Here, the simulation is performed using initial conditions x0 from timepoints 0 to 18000 seconds, advancing every 100 seconds. The differential step function dxdt() is called with additional parameters c(list(nu=nu, a=a), params). The simulation algorithm is set to daspk which is best suited for stiff ODE systems.

The result out is a matrix-like object with rows and columns as timepoints and species, respectively.

To plot the simulated time profile of the species Cu, the intracellular copper abundance, first a time vector in units of minutes is generated:

t.abs = out[,'time'] / 60

Then the raw Cu trajectory is plotted via:

plot(t.abs, out[,'Cu'], type='l')

Here it is important to note that Cu exists as both an isolated species and bound to other molecules such as metallochaperones and export pumps. Thus, it is necessary to 'totalize' all Cu species to get the correct intracellular abundance:

plot(t.abs, .T('Cu', out), type='l')

Finally, the model accounts for cellular growth. To get the intracellular Cu abundance on a per cell basis (assuming a volume of 1 is one cell):

plot(t.abs, .T('Cu', out)/out[, 'V'], type='l')

Summary

The sequence of steps above describe the core simulation workflow. All other analyses - e.g. parameter sweeps, strain comparisons, sensitivity analysis - merely extend this workflow.

Appendix 1: SBML Export

Code is available to convert the model as it exists in R to SBML a commonly used format for easy exchange and simulation of models in a variety of computational tools.

Requirements

Cu Efflux Model R Code
R 2.11 (or higher)
R Packages:
- rsbml

Coversion

To convert the R model to an SBML file simply run:

source('makeSBML.R')

This will produce a file called CuEfflux.xml that is SBML L2V1 compliant which can then be imported and further modified/tested using SBML aware software tools such as:

COPASI
CellDesigner

Caveats

If any modifications to the model are made one must take the following into consideration prior to conversion to SBML.

rsbml does not check SBML consistency. An SBML file may be produced but may generate errors during import.
Mathematical expressions for reaction KineticLaws converted from R to SBML/MathML must be in fully expanded infix form - e.g.:

a*(b + c)

will produce libSBML errors, whereas

a*b + a*c

will covert to SBML/MathML.  This limitation is not specified anywhere in
either the rsbml, SBML, or libSBML documentation.

Object names cannot have '.'. Substituting with '_' passes.

wleepang / CuEfflux