Swiftcore

blocked(S, rev, solver)

blocked finds all the irreversible blocked reactions and is utilized in swiftcc

USAGE:

result = blocked (S, rev, solver)

INPUTS:

• S – the associated sparse stoichiometric matrix

• rev – the 0-1 vector with 1’s corresponding to the reversible reactions

• solver – the LP solver to be used; the currently available options are ‘gurobi’, ‘linprog’, and ‘cplex’ with the default value of ‘linprog’. It fallbacks to the COBRA LP solver interface if another supported solver is called.

OUTPUT:

result – the result returned by the LP solver; among the last n entries, all the -1 entries are blocked, and the other entries are zero. The first m entries are its fictitious metabolite certificate.

core(model, blocked, weights, solver)

core finds a feasible flux distribution to unblock a given list of blocked reactions and is utilized in swiftcore

USAGE:

flux = core (S, rev, blocked, weights, solver)

INPUTS:

• model – the metabolic network with fields: * .S - the associated sparse stoichiometric matrix * .rev - the 0-1 indicator vector of the reversible reactions * .rxns - the cell array of reaction abbreviations * .mets - the cell array of metabolite abbreviations

• blocked – the 0-1 vector with 1’s corresponding to the blocked reactions

• weights – weight vector for the penalties associated with each reaction

• solver – the LP solver to be used; the currently available options are ‘gurobi’, ‘linprog’, and ‘cplex’ with the default value of ‘linprog’. It fallbacks to the COBRA LP solver interface if

  another supported solver is called.

OUTPUT:

flux – a feasible flux distribution

partition(model, solver, algorithm)

swiftcc++ and fastcc++ augment swiftcc and fastcc by this preprocess

USAGE:

component = partition (model, solver, algorithm)

INPUTS:

• model – the metabolic network reconstruction * .S - the associated sparse stoichiometric matrix * .lb - feasible flux distribution lower bound * .ub - feasible flux distribution uppper bound * .rxns - cell array of reaction abbreviations * .rev - the 0-1 indicator vector of the reversible reactions

• solver – the LP solver to be used; the currently available options are ‘gurobi’, ‘linprog’, and ‘cplex’ with the default value of ‘linprog’. It fallbacks to the COBRA LP solver interface if another supported solver is called.

• algorithm – the backend algorithm to be utilized between ‘swift’ and ‘fast’

OUTPUT:

component – the index set of the reactions constituting the maximum flux consistent metabolic subnetwork

Note

requires bioinformatics toolbox

swiftcc(S, rev, varargin)

swiftcc is an even faster version of fastcc

USAGE:

consistent = swiftcc (S, rev [, solver])

INPUTS:

• S – the associated sparse stoichiometric matrix

• rev – the 0-1 vector with 1’s corresponding to the reversible reactions

OPTIONAL INPUT:

solver – the LP solver to be used; the currently available options are ‘gurobi’, ‘linprog’, and ‘cplex’ with the default value of ‘linprog’. It fallbacks to the COBRA LP solver interface if another supported solver is called.

OUTPUT:

consistent – the 0-1 indicator vector of the reactions constituting the maximum flux consistent metabolic subnetwork

swiftcore(model, coreInd, weights, tol, reduction, varargin)

swifcore is an even faster version of fastcore

USAGE:

[reconstruction, reconInd, LP] = swiftcore (model, coreInd, weights, tol, reduction, solver)

INPUTS:

• model – the metabolic network with fields: * .S - the associated sparse stoichiometric matrix * .lb - lower bounds on reaction rates * .ub - upper bounds on reaction rates * .rxns - the cell array of reaction abbreviations * .mets - the cell array of metabolite abbreviations

• coreInd – the set of indices corresponding to the core reactions

• weights – the weight vector for the penalties associated with each reaction

• tol – zero-tolerance, i.e., the smallest flux value considered nonzero

• reduction – boolean enabling the metabolic network reduction preprocess

OPTIONAL INPUT:

solver – the LP solver to be used; the currently available options are ‘gurobi’, ‘linprog’, and ‘cplex’ with the default value of ‘linprog’. It fallbacks to the COBRA LP solver interface if another supported solver is called.

OUTPUTS:

• reconstruction – the flux consistent metabolic network reconstructed from the core reactions

• reconInd – the 0-1 indicator vector of the reactions constituting the reconstruction

• LP – the number of solved LPs

Note

For the choice of the weight vector, use c*ones(n, 1) where c is an arbitrary constant c > 1 if you have no preference over reactions. Also, note that the input model is assumed to be flux consistent.