New

createGroupIncidenceMatrix(model, trainingModel, param)

Build a combined group (fragment) incidence matrix for a test model and the training model, as used by the component contribution method

USAGE:

combinedModel = createGroupIncidenceMatrix (model, trainingModel, param)

INPUTS:

• model – test COBRA model structure with fields:

  • .mets - m x 1 cell array of metabolite identifiers

  • .inchi - structure whose .nonstandard field is an m x 1 cell array of nonstandard InChI

• trainingModel – training data model structure with fields:

  • .S - p x n stoichiometric matrix of the training data

  • .mets - p x 1 metabolite abbreviations

  • .rxns - n x 1 reaction abbreviations

  • .metKEGGID - p x 1 cell array of metabolite KEGG IDs

  • .inchi - structure whose .nonstandard field is a p x 1 cell array of nonstandard InChI

  • .cids_that_dont_decompose - KEGG compound IDs that do not decompose (manual fragmentation)

  • .groups - g x 1 cell array of group definitions (written)

  • .G - p x g group incidence matrix (written)

  • .groupDecomposableBool - p x 1 boolean of group-decomposable metabolites (written)

  • .trainingMetBool - p x 1 boolean flagging training metabolites (written)

  • .testMetBool - p x 1 boolean flagging test metabolites (written)

OPTIONAL INPUT:

param – structure of parameters with fields:

• .printLevel - verbosity level (default 0)

• .fragmentationMethod - ‘abinito’ (default) or ‘manual’

• .modelCache - cache filename for the test-model fragmentation

• .radius - number of bonds around each central atom (default 1)

• .dGPredictorPath - absolute path to a clone of dGPredictor

• .canonicalise - boolean, canonicalise SMILES fragments (default 0)

OUTPUT:

combinedModel – combined model structure (training padded with test-only metabolites), with fields including:

• .S - k x n stoichiometric matrix

• .G - k x g group incidence matrix

• .groups - g x 1 cell array of group definitions

• .mets - k x 1 cell array of metabolite identifiers

• .rxns - n x 1 reaction abbreviations

• .inchi - structure whose .nonstandard field is a k x 1 cell array of nonstandard InChI

• .groupDecomposableBool - k x 1 boolean of group-decomposable metabolites

• .inchiBool - k x 1 boolean of metabolites with an InChI

• .trainingMetBool - k x 1 boolean flagging training metabolites

• .testMetBool - k x 1 boolean flagging test metabolites

• .test2CombinedModelMap - m x 1 mapping of model.mets to combinedModel.mets

getGroupVectorFromInchi(inchi, printLevel)

Decompose an InChI into a group-contribution vector via the inchi2gv python script

USAGE:

group_def = getGroupVectorFromInchi (inchi, printLevel)

INPUT:

inchi – InChI string of the metabolite to decompose

OPTIONAL INPUT:

printLevel – verbosity level (default 0): 0 no output, 1 progress only, 2 progress and warnings

OUTPUT:

group_def – row vector of group counts (the group-contribution vector), empty if the InChI cannot be decomposed

getMappingScores(model, trainingModel)

Finds the best mapping between the model metabolites and the training model metabolites, the higher the confidence score, the more reliable the mapping

USAGE:

mappingScore = getMappingScores (model, trainingModel)

INPUTS:

• model – test COBRA model structure with fields:

  • .mets - m x 1 cell array of metabolite identifiers

  • .inchi - structure of InChI strings, with fields .standard, .standardWithStereo and .standardWithStereoAndCharge (each m x 1 cell arrays)

• trainingModel – training COBRA model structure with fields:

  • .metKEGGID - p x 1 cell array of metabolite KEGG IDs

  • .inchi - structure of InChI strings, with fields .standard, .standardWithStereo and .standardWithStereoAndCharge (each p x 1 cell arrays)

OUTPUT:

mappingScore – nMet x nTrainingMet sparse matrix giving the best mapping score between model and training metabolites (higher is more reliable)

regulariseGroupIncidenceMatrix(combinedModel, printLevel)

Identify similar and duplicate metabolites in the combined model, comparing group-decomposition vectors and InChI, to regularise the group incidence matrix

USAGE:

[groupM, inchiM] = regulariseGroupIncidenceMatrix (combinedModel, printLevel)

INPUT:

combinedModel – combined model structure with fields:

• .G - k x g group incidence matrix

• .inchi - structure whose .nonstandard field is a k x 1 cell array of nonstandard InChI

• .inchiBool - k x 1 boolean of metabolites with an InChI

• .trainingMetBool - k x 1 boolean flagging training metabolites

• .testMetBool - k x 1 boolean flagging test metabolites

• .mets - k x 1 cell array of metabolite identifiers

OPTIONAL INPUT:

printLevel – verbosity level (default 0)

OUTPUTS:

• groupM – nMet x nMet logical matrix, true where two metabolites share a group decomposition

• inchiM – nMet x nMet logical matrix, true where two metabolites share an InChI