ACHR

ACBSampler(model, warmupPoints, fileName, nFiles, pointsPerFile, nMixPts, nWarmupNeeded, saveMatFlag, biasOpt)

Artificial centering boundary sampler

USAGE:

ACBSampler(model, warmupPoints, fileName, nFiles, pointsPerFile, nMixPts, nWarmupNeeded, saveMatFlag, biasOpt)

INPUTS:

model: Model structure warmupPoints: Warmup points fileName: Base fileName for saving results nFiles: Number of sample point files created pointsPerFile: Number of points per file saved nMixPts: Number of steps initially used for mixing (not saved)

OPTIONAL INPUTS:

nWarmupNeeded: Number of warmup points needed (Default = 20000) saveMatFlag: Save points in mat format vs txt format (Default = true) biasOpt: Options for biasing sampler (Default = no bias)

ACHRSampler(model, warmupPoints, fileName, nFiles, pointsPerFile, stepsPerPoint, initPoint, fileBaseNo, maxTime, printLevel, minFlux, maxFlux)

Artificial Centering Hit-and-Run sampler

USAGE:

ACHRSampler(model, warmupPoints, fileName, nFiles, pointsPerFile, stepsPerPoint, initPoint, fileBaseNo, maxTime)

INPUTS:

model: Model structure warmupPoints: Warmup points fileName: Base fileName for saving results nFiles: Number of files created pointsPerFile: Number of points per file saved stepsPerPoint: Number of sampler steps per point saved printLevel: 0 silent / 1 active

OPTIONAL INPUTS:

initPoint: Initial point (Default = centerPoint) fileBaseNo: Base file number for continuing previous sampler run

(Default = 0)

maxTime: Maximum time limit (Default = 36000 s) minFlux: Lower bound of the solution space given by FVA maxFlux: Upper bound of the solution space given by FVA

ACHRSamplerParallelGeneral(sampleStruct, nLoops, stepsPerPoint, maxtime, proc, fdirectory)

Artificial Centering Hit-and-Run sampler with in place (memory) point management

USAGE:

sampleStruct = ACHRSamplerParallelGeneral(sampleStruct, nLoops, stepsPerPoint, maxtime, proc, fdirectory)

INPUTS:

sampleStruct: Sampling structure nLoops: Number of iterations stepsPerPoint: Number of sampler steps per point saved maxtime: Amount of time to spend on calculation (in seconds)

OPTIONAL INPUTS:

proc: Number of processes if > 0. Otherwise, the proces #. fdirectory: Do not use this parameter when calling function directly.

OUTPUT:

sampleStruct: Sampling structure with sample points

convRevSamples(model, samples)

Converts signs for reactions that are only running in reverse direction

USAGE:

[model, samples] = convRevSamples(model, samples)

INPUT:

model: Constraint-based model

OPTIONAL INPUT:

samples: Sample set

OUTPUTS:

model: COBRA model structure with negative-direction fluxes reversed samples: Sample set with negative-direction fluxes reversed

createHRWarmup(model, nPoints, verbFlag, bias, nPointsCheck)

Creates a warmup point set for hit-and-run sampling by combining orthogonal and random points

USAGE:

warmupPts = createHRWarmup(model, nPoints, verbFlag, bias, nPointsCheck)

INPUTS:

model: Model structure

OPTIONAL INPUTS:

nPoints: Number of warmup points (Default = 5000); verbFlag: Verbose flag (Default = false) bias: Structure with fields:

  • method - Biasing distribution: ‘uniform’, ‘normal’

  • index - The reaction indexes which to bias (nBias total)

  • param - nBias x 2 matrix of parameters (for uniform it’s min max, for normal it’s mu, sigma).

OUTPUT:

warmupPts: Set of warmup points

gpSampler(sampleStruct, nPoints, bias, maxTime, maxSteps, threads, nPointsCheck)

Samples an arbitrary linearly constrained space using a fixed number of points that are moved in parallel The space is defined by

\[\begin{split}A x \leq, =, \geq b \\ lb \leq x \leq ub\end{split}\]

USAGE:

[sampleStructOut, mixedFraction] = gpSampler(sampleStruct, nPoints, bias, maxTime, maxSteps, threads, nPointsCheck)

INPUTS:
sampleStruct: Structure describing the space to be sampled and

previous point sets:

  • A - LHS matrix (optionally, if not A script checks for S)

  • b - RHS vector

  • lb - Lower bound

  • ub - Upper bound

  • csense - Constraint type for each row in A (‘G’, ‘L’, ‘E’)

  • warmupPoints - Set of warmup points (optional, generated by default)

  • points - Currently sampled points (optional)

OPTIONAL INPUTS:
nPoints: Number of points used in sampling

(default = 2 * nRxns or 5000 whichever is greater)

bias: Structure with fields:

  • method - Biasing distribution: ‘uniform’, ‘normal’

  • index - The reaction indexes which to bias (nBias total)

  • param - nBias x 2 matrix of parameters (for uniform it’s min max, for normal it’s mu, sigma).

maxTime: Maximum time alloted for the sampling in seconds

(default 600 s, pass an empty number [] to set maxSteps instead)

maxSteps: Maximum number of steps to take (default 1e10). Sampler

will run until either maxStep or maxTime is reached. Set maxStep or maxTime to 0 and no sampling will occur (only warmup points generated).

threads: number of threads the sampler will use. If you have a

dual core machine, you can set it to 2 etc. The speed up is almost linear w/ the number of cores. If using this feature and 2009a or newer, a futher speedup can be obtained by starting matlab from the command line by “typing matlab -singleCompThread” New feature: if threads < 0, use distributed toolbox.

nPointsCheck: Checks that minimum number of points (2 * nRxns) are

used. (Default = true).

OUTPUTS:

sampleStructOut: The sampling structure with some extra fields. mixedFract: The fraction mixed (relative to the warmupPts). A value of 1

means not mixed at all. A value of .5 means completely mixed.

mixFraction(sample1, sample2, fixed)

Compares two sets of sampled points and determines how mixed they are.

USAGE:

mix = mixFraction(sample1, sample2, fixed)

INPUTS:
sample1, sample2: Ordered set of points. The points must be in

the same order otherwise it does not make sense.

OPTIONAL INPUT:
fixed: The indexes of sampled reactions which are fixed

and are not expected to mix. They are ignored.

OUTPUT:
mix: The mix fraction. Goes from 0 to 1 with 1 being

completely unmixed and .5 being essentially perfectly mixed.

Example

example 1: mix = mixFraction(sample1, sample2) example 2: fixed = findRxnIDs({R1, R2, …, Rx}) mix = mixFraction(sample1, sample2, fixed)

removeCorrelRxns(model, R, correlCutoff)

Removes fully or almost fully correlated reactions

USAGE:

[selRxns, rxnSets, rxnList, Rfilt] = removeCorrelRxns(model, R, correlCutoff)

INPUTS:

model: COBRA model structure R: Correl coefficient matrix

OPTIONAL INPUT:

correlCutoff: Cutoff level for fully correlated rxns (Default 0.99999)

OUTPUTS:

selRxns: true/false vector that allow selecting non-redundant data rxnSets: Correlated reaction sets rxnList: Reaction list with correlated reactions concatenated Rfilt: Filtered R

verifyPoints(sampleStruct)

Verify that a set of points are in the solutoin space of sampleStruct. Typically, this method would be called to check a set of warmup points or points generated via gpSampler. Also verifies if points moved from warmup points.

USAGE:

[errorsA, errorsLUB, stuckPoints] = verifyPoints(sampleStruct)

INPUT:

sampleStruct: LPProblem containing points and warmup points

OUTPUTS:
errorsA: Row index of the constraint in sampleStruct that

is not consistent with the given points

errorsLUB: Upper and lower bounds of the constraint + tolerance stuckPoints: Index of points which did not move.