cobrar

The R package cobrar provides structures and functions for constraint-based metabolic network analysis, e.g. the prediction of metabolic fluxes using fluxes using Flux Balance Analysis (FBA). cobrar is inspired by the former CRAN R package sybil(1).

Installation

Please note that cobrar requires the two system libraries libSBML and glpk. The following installation instructions for different operating systems install the dependencies first, then cobrar. If you already have libSBML and glpk installed, you can skip to the last part of the instructions.

cobrar is under development and the installation instructions described will install the latest development version.

Ubuntu/Debian/Mint

Install libSBML and glpk:

sudo apt install libsbml-dev libglpk-dev

Install cobrar (in R):

# install.packages("remotes")
remotes::install_github("Waschina/cobrar")

Centos/Fedora/RHEL

Install libSBML and glpk:

sudo yum install libsbml-devel glpk-devel

Install cobrar (in R):

# install.packages("remotes")
remotes::install_github("Waschina/cobrar")

MacOS

libSBML and glpk can be installed using homebrew:

brew install glpk brewsci/bio/libsbml

Install cobrar (in R):

# install.packages("remotes")
remotes::install_github("Waschina/cobrar")

Windows

TODO

Conda

TODO

Usage

The full documentation including illustrative examples is available here.

The following (very brief) example illustrates how to use cobrar to perform flux balance analysis for a core metabolism model of Escherichia coli. In a first simulation, fluxes and growth are predicted for the default constraints of the model, which represents a minimal medium with glucose as the sole carbon source and the presence of oxygen. The second simulation performs the same flux balance analysis but without oxygen to simulate an anoxic growth environment.

# First, load the cobrar package and the metabolic model:
library(cobrar)
#> Loading required package: Matrix
#> cobrar uses...
#>  - libSBML (v. 5.18.0)
#>  - glpk (v. 4.65)
fpath <- system.file("extdata", "e_coli_core.xml", package="cobrar")
mod <- readSBMLmod(fpath)

# Next, perform simulation 1: Aerobic growth of E. coli
res_aero <- fba(mod)

# Finally, simulation 2: Anaerobic growth of E. coli
mod <- changeBounds(mod, react = "EX_o2_e", lb = 0) # before: -1000
res_anaero <- fba(mod)

# Report simulation results (i.e., growth and acetate production)
cat("[Aerobic growth]\n",
    " Growth rate:        ", res_aero@obj,"\n",
    " Acetate production: ", res_aero@fluxes[mod@react_id == "EX_ac_e"],"\n\n",
    "[Anaerobic growth]\n",
    " Growth rate:        ", res_anaero@obj,"\n",
    " Acetate prodcution: ", res_anaero@fluxes[mod@react_id == "EX_ac_e"],"\n",
    sep = "")
#> [Aerobic growth]
#>  Growth rate:        0.8739215
#>  Acetate production: 0
#> 
#> [Anaerobic growth]
#>  Growth rate:        0.2116629
#>  Acetate prodcution: 8.503585

Key differences to sybil

• cobrar is fully functional from reading SBML files until optimisation of linear programs, without the need of additional packages such as sybilSBML or glpkAPI.
• cobrar links to libsbml via libsbml’s C++ API, not the C API as the sybilSBML package.
• The GNU glpk library is a system requirement. There are no links/references to IBM’s CPLEX in the cobrar package, which were one reason why sybil was discontinued on CRAN.
• roxygen2 is used for documenting functions and classes.
• Feature trim: A range of functions in sybil are not part of cobrar.
• In cobrar, R’s garbage collector handles memory management, including memory associated to C++-objects and the pointers to these.
• Simplifications in class and function architecture
  • No more Class “SysBiolAlg” nor sub-classes. Different algorithms have their own function and detailed documentation of their return values
• In sybil, columns named “annotation” were actually concatenated CVTerms (https://synonym.caltech.edu/software/libsbml/5.20.0/cpp-api/class_c_v_term.html). To avoid confusion with other levels of annotation, the columns (e.g. in react_attr or met_attr) are named ‘CVTerms’ in cobrar.
• cobrar allows to assign SBOTerms to reactions, metabolites, genes.
• Performance (i.e., computation time) improvements in certain procedures:
  • identification of dead-end metabolites
  • reading/exporting SBML files,
  • pFBA algorithm (a.k.a MTF ‘Minimization of Total Flux’).
  • FVA; also now allows relaxed constraints on optimal growth (e.g. flux variability with 90-100% optimal growth)

Notes

• cobrar exports SBML files level 3 version 2 with fbc version 3 and groups version 1.
• Group assignments are only supported for reactions.
• currently only glpk is supported as solver. A plugin for IBM’s ILOG CPLEX is planned.
• Multiple objectives. The SBML standard with its fbc extension allows to specify more than one objective (class ListOfObjectives). However, cobrar can only handle one current objective function per model, which is defined as an objective coefficient vector in slot obj_coef of an object of class modelorg. Note that when reading SBML models, cobrar will only use the first objective defined in the SBML document.

References

1. Gelius-Dietrich, G., Desouki, A.A., Fritzemeier, C.J., Lercher, M.J. sybil – Efficient constraint-based modelling in R. BMC Syst Biol 7, 125 (2013). https://doi.org/10.1186/1752-0509-7-125