Sampling

calcSampleDifference(sample1, sample2, nPts)[source]

Selects randomly nPts flux vectors from sample1 and sample2 and calcutes the difference between the flux vectors

USAGE:

[sampleDiff, sampleRatio] = calcSampleDifference (sample1, sample2, nPts)

INPUTS:

  • sample1 – First flux sample

  • sample2 – Second flux sample

OPTIONAL INPUTS:

nPts – Number of flux difference profiles desired (default: 10% of the samples)

OUTPUTS:

  • sampleDiff – Difference between the flux vectors

  • sampleRatio – Ratio of the flux vectors

Example

example 1: [sampleDiff, sampleRatio] = calcSampleDifference(sample1, sample2) example 2: [sampleDiff, sampleRatio] = calcSampleDifference(sample1, sample2, 10)

calcSampleStats(samples)[source]

Calculate sample modes, means, standard devs, and medians of the sample

USAGE:

sampleStats = calcSampleStats (samples)

INPUT:

samples – Samples to analyze

OUTPUT:

sampleStats – Structure with the following fields:

  • mean

  • std

  • mode

  • median

  • skew

  • kurt

Example

example 1: sampleStats = calcSampleStats(sample) example 2: sampleStats = calcSampleStats({sample1, sample2})

compareSampleTraj(rxnNames, samples, models, nBins)[source]

Compares flux histograms for two or more samples for one or more reactions

USAGE:

compareSampleTraj (rxnNames, samples, models, nBins)

INPUTS:

  • rxnNames – List of reaction names to compare

  • samples – Samples to compare

  • models – Cell array containing COBRA model structures

OPTIONAL INPUTS:

nBins – Number of bins (Default = nSamples / 25)

compareTwoSamplesStat(sample1, sample2, tests)[source]

Compares statistically the difference between two samples. Does the Kolmogorov-Smirnov, rank-sum, chi-square, and T-tests.

USAGE:

[stats, pVals] = compareTwoSamplesStat (sample1, sample2, tests)

INPUTS:

  • sample1 – First sample (nVar x nSample matrix) to compare

  • sample2 – Second sample (nVar x nSample matrix) to compare

  • tests – {test1, test2,…} (Default = all tests)

    • ‘ks’ - Kolmogorov-Smirnov test

    • ‘rankSum’ - rank-sum test

    • ‘chiSquare’ - chi-squre test

    • ‘tTest’ - T-test

OUTPUTS:

  • stats – Struct array with statistics of the selected tests.

  • pVals – Struct array with p values of the selected tests.

Example

example 1: [stats, pVals] = compareTwoSamplesStat(sample1, sample2) example 2: [stats, pVals] = compareTwoSamplesStat(sample1, sample2, {‘ks’, ‘rankSum’})

Output will be in order that tests are inputed. i.e. {‘ks’,’rankSum’}

identifyCorrelSets(model, sample, corrThr, R)[source]

Identifies correlated reaction sets from sampling data

USAGE:

[setsSorted, setNoSorted, setSize] = identifyCorrelSets (model, sample, corrThr, R)

INPUTS:

  • model – COBRA model structure, with fields:

    • .rxns - reaction identifiers

  • sample – Sample to be used to identify correlated sets

OPTIONAL INPUTS:

  • corrThr – Minimum correlation (\(R^2\)) threshold (Default = 1-1e-8)

  • R – Correlation coefficient

OUTPUTS:

  • setsSorted – Sorted cell array of sets (largest first)

  • setNoSorted – List of set numbers for each reaction in model (0 indicates that there is no set)

  • setSize – List of set sizes

loadSamples(filename, numFiles, pointsPerFile, numSkipped, randPts)[source]

Loads a set of sampled data points

USAGE:

samples = loadSamples (filename, numFiles, pointsPerFile, numSkipped, randPts)

INPUTS:

  • filename – The name of the files containing the sample points.

  • numFiles – The number of files containing the sample points.

  • pointsPerFile – The number of points to be taken from each file.

OPTIONAL INPUTS:

  • numSkipped – Number of files skipped (default = 0)

  • randPts – Select random points from each file (true/false, default = false)

OUTPUT:

samples – Sample flux distributions

plotHistConv(model, sample, rxnNames, nSubSamples)[source]

Plots convergence of sample histograms

USAGE:

plotHistConv (model, sample, rxnNames, nSubSamples)

INPUTS:

  • model – COBRA model structure, with fields:

    • .rxns - reaction identifiers

  • sample – Sampled fluxes

  • rxnNames – List of reactions to plot

  • nSubSamples – Number of sub samples

Example

example 1: rxnNames = {‘R1’, ‘R2’} plotHistConv(model, sample, rxnNames, nSubSamples)

plotSampleHist(rxnNames, samples, models, nBins, perScreen, modelNames, add2Plot)[source]

Compares flux histograms for one or more samples for one or more reactions

USAGE:

plotSampleHist (rxnNames, samples, models, nBins, perScreen, modelNames, add2Plot)

INPUTS:

  • rxnNames – Cell array of reaction abbreviations

  • samples – Cell array containing samples

  • models – Cell array containing model structures or common model structure

OPTIONAL INPUTS:

  • nBins – Number of bins to be used

  • perScreen – Number of reactions to show per screen. Either a number or [nY, nX] vector. (press ‘enter’ to advance screens)

  • modelNames – Cell array containing the name of the models (used for the plot’s legend).

  • add2Plot – Struct array with additional data to show more detailed information (real measurements, FVA results, statistics results, etc.), with fields:

    • .line - x-axis position at which to draw a marker line

    • .label - text label drawn next to the marker line

Example

sampleStructOut1 = gpSampler(model1, 2150); sampleStructOut2 = gpSampler(model2, 2150); %Plot for model 1 plotSampleHist(model1.rxns,{samplePoints1},{model1})

%Plot reactions reactions in model 1 that also exist in model 2 using 10 %bins and plotting 12 reactions per screen. plotSampleHist(model1.rxns,{samplePoints1,samplePoints2},{model1,model2},10,12)

CONTROLS: To advance to next screen hit ‘enter/return’ or type ‘f’ and hit ‘enter/return’ To rewind to previous screen type ‘r’ or ‘b’ and hit ‘enter/return’ To quit script type ‘q’ and hit ‘enter/return’

printSampleStats(sampledModel, commonModel, sampleNames, fileName)[source]

Prints out sample statistics for multiple samples

USAGE:

printSampleStats (samples, commonModel, sampleNames, fileName)

INPUTS:

  • sampledModel – Samples to summarise

  • commonModel – COBRA model structure, with fields:

    • .rxns - reaction identifiers

    • .subSystems - reaction subsystem assignments (optional; printed as an extra column when present)

  • sampleNames – Names of the models

OPTIONAL INPUT:

fileName – Name of tab delimited CSV file to generate (Default = print to command window)

sampleCbModel(model, sampleFile, samplerName, options, modelSampling)[source]

Samples the solution-space of a constraint-based model

USAGE:

[modelSampling, samples] = sampleCbModel (model, sampleFile, samplerName, options, modelSampling)

INPUTS:

model – COBRA model structure with fields:

  • .S - Stoichiometric matrix

  • .b - Right hand side vector

  • .c - ‘n x 1’ linear objective coefficients (used as the bias vector for CHRR_EXP sampling)

  • .lb - ‘n x 1’ vector: Lower bounds

  • .ub - ‘n x 1’ vector: Upper bounds

  • .rxns - ‘n x 1’ reaction identifiers (renamed while preparing the ACHR model)

  • .C - ‘k x n’ matrix of additional inequality constraints

  • .d - ‘k x 1’ rhs of the above constraints

  • .dsense - ‘k x 1’ the sense of the above constraints (‘L’ or ‘G’)

  • .vMean - ‘n x 1’ vector: the mean for Gaussian sampling (RHMC only)

  • .vCov - ‘n x 1’ vector: the diagonal for the covariance for Gaussian sampling (RHMC only)

OPTIONAL INPUTS:

  • sampleFile – File name stem for the sampling output .mat files (only implemented for ACHR)

  • samplerName – {(‘CHRR’), ‘ACHR’, ‘RHMC’} Name of the sampler to be used to sample the solution.

  • options – Options for sampling and pre/postprocessing (default values in parenthesis).

    • .nStepsPerPoint - Number of sampler steps per point saved (200)

    • .nPointsReturned - Number of points loaded for analysis (2000)

    • .nWarmupPoints - Number of warmup points (5000). ACHR only.

    • .nFiles - Number of output files (10). ACHR only.

    • .nPointsPerFile - Number of points per file (1000). ACHR only.

    • .nFilesSkipped - Number of output files skipped when loading points to avoid potentially biased initial samples (2) loops (true). ACHR only.

    • .maxTime - Maximum time limit (Default = 36000 s). ACHR only.

    • .toRound - Option to round the model before sampling (true). CHRR only.

    • .lambda - the bias vector for exponential sampling. CHRR_EXP only.

    • .nWorkers - Number of parallel workers. RHMC only.

    • .useFastFVA - Use fastFVA during preprocessing (false). CHRR only.

    • .optPercentage - Optimality percentage for the FVA bounds (100). CHRR only.

    • .eps - Target relative error for the volume estimate (0.15). MFE only.

  • modelSampling – From a previous round of sampling the same model. Input to avoid repeated preprocessing, with fields (RHMC only):

    • .problem - polytope problem struct reused across runs

    • .samples - sample buffer on the returned sampler object (cleared to empty on output)

OUTPUTS:

  • modelSampling – Cleaned up model used in sampling

  • samples – n x numSamples matrix of flux vectors

  • volume – Estimated polytope volume when samplerName is ‘MFE’; otherwise a message string describing how to estimate it

Examples

%1) Sample a model called ‘superModel’ using default settings and save the % results in files with the common beginning ‘superModelSamples’

[modelSampling,samples] = sampleCbModel(superModel,’superModelSamples’);

%2) Sample a model called ‘hyperModel’ using default settings except with a total of 50 sample files % saved and with 5000 sample points returned.

options.nFiles = 50; options.nPointsReturned = 5000; [modelSampling,samples] = sampleCbModel(hyperModel,’options’,options);

sampleScatterMatrix(rxnNames, model, sample, nPoints, fontSize, dispRFlag, rxnNames2)[source]

Draws a scatterplot matrix with pairwise scatterplots for multiple reactions

USAGE:

sampleScatterMatrix (rxnNames, model, sample, nPoints, fontSize, dispRFlag, rxnNames2)

INPUTS:

  • rxnNames – Cell array of reaction names to be plotted

  • model – Model structure, with fields:

    • .rxns - reaction identifiers

  • sample – Samples to be analyzed (nRxns x nSamples)

OPTIONAL INPUTS:

  • nPoints – How many sample points to plot (Default nSamples)

  • fontSize – Font size for labels (Default calculated based on number of reactions)

  • dispRFlag – Display correlation coefficients (Default false)

  • rxnNames2 – Optional second set of reaction names

Example

%Plots the scatterplots only between the three reactions listed - %histograms for each reaction will be on the diagonal sampleScatterMatrix({‘PFK’, ‘PYK’, ‘PGL’}, model, sample); %Plots the scatterplots between each of the first set of reactions and %each of the second set of reactions. No histograms will be shown. sampleScatterMatrix({‘PFK’, ‘PYK’, ‘PGL’}, model, sample, 100, 10, true, {‘ENO’,’TPI’});