RAVEN release notes#

Unreleased#

Breaking changes#

  • getModelFromHomology with complexPolicy drop removes each complex that has a subunit without an ortholog, and removes the reaction only when no complex remains (was: the reaction was removed as soon as one subunit had no ortholog). For the rule (A and B) or (C and D) without an ortholog of B, the draft has the orthologs of C and D.

Fixes#

  • getModelFromHomology reads the rule as a tree, in which and binds tighter than or, when it removes the placeholder genes of isozymes without an ortholog (was: it removed a placeholder that stood beside an or token, also when the placeholder belongs to a complex). A placeholder in an or group is removed when the group has a member that is not a placeholder, in nested groups as well. For the rule (A or B and C) and D without an ortholog of B, where nA, nC and nD are the orthologs of A, C and D, the draft rule is (nA or (OLD_T_B and nC)) and nD under flag, (nA or nC) and nD under keep and nA and nD under drop; it was nA and nC and nD under all three. The result differs from before only for rules with unbracketed and and or or with nested or groups.

Data#

  • The KEGG assets are the kegg120 release of toolbox-assets (was: kegg118), and fetchToolboxAsset reads the manifest manifest-v5.

3.0.0b2 (2026-11-01)#

Second beta toward 3.0. Changes below are relative to 3.0.0b1.

Breaking changes#

  • The folder layout changed (see Repository layout and downloads); code that adds RAVEN folders to the path or addresses binaries by path needs the new paths.

  • fetchRavenDataAsset is replaced by fetchToolboxAsset.

  • gapFillMILP returns [addedRxns, newModel, exitFlag] and has no weights option.

  • sortModel keeps the directions of the reactions unless sortReversible is requested (was: it sorted them by default).

  • parseTaskList refuses a binary .xls file.

  • gapReport, scaleBiomassFraction, getModelFromHomology and getIndexes raise errors in cases they accepted before (see below).

Repository layout and downloads#

  • The MATLAB code is in src/ (was: folders in the repository root), the tests in tests/ (was: testing/), the generated help pages in docs/html/ (was: doc/), the readme files in docs/readme/, the logo in img/ (was: RAVEN.png), and the tutorials in tutorials/ with data under tutorials/data (was: tutorial/). The external software is in src/external (was: software/). checkRaven, downloadRavenBinaries, getBlast, getDiamond, getKEGGModelForOrganism, optimizeProb, setRavenSolver and updateDocumentation use the new paths.

  • parseHPA, parseHPArna and scoreModel are in INIT/ (was: omics/, which is removed). tOmics.m is merged into tINIT.m.

  • fetchToolboxAsset replaces fetchRavenDataAsset, and assets are downloaded from CompMetEng/toolbox-assets (was: raven-data). downloadRavenBinaries, buildGlobalKEGGModel, getKEGGModelForOrganism, getModelFromKEGG, getPhylDist, keggDataVersion and checkRaven use it.

  • fetchToolboxAsset downloads to a .part file and checks the SHA-256 sum listed in the manifest-v3 of toolbox-assets. A mismatch deletes the file and raises an error; an asset that the manifest does not list, or a manifest that cannot be read, gives a warning.

  • The repository, documentation and version-check links point to CompMetEng.

I/O#

  • writeYAMLmodel writes an ec reaction without enzymes (a row of zeros in model.ec.rxnEnzMat) with no enzymes entry (was: - enzymes: !!omap with nothing under it, which is not valid YAML). readYAMLmodel reads a missing enzymes entry as a reaction without enzymes, and reads - enzymes: !!omap with no entries and - enzymes: !!omap [] from files written by earlier versions.

  • readYAMLmodel reads notes maps (as written by cobrapy and raven-toolbox) into metNotes and rxnNotes (was: the error "Unknown entry"), reads folded scalars whose continuation starts with -, and keeps one model.subSystems entry per reaction (was: empty entries were dropped, so later subsystems sat on the wrong reaction).

  • exportForGit raises RAVEN:notMainBranch when a RAVEN checkout is on a branch other than main or master.

  • writeYAMLmodel writes the enzymes of each ec-rxns entry indented under - enzymes (was: at the indentation of the key, which ruamel.yaml could not read), quotes the YAML 1.1 booleans (y, n, yes, no, on, off in any case), and writes non-integers with the shortest digits that read back exactly. printOrange prints \n as a line break.

  • exportForGit: the txt table is headed Rxn id, writes bounds and objective with the digits needed to read them back and pads no field. The output folder is created with its parents also when subDirs is false, and the logical flags are validated.

Manipulation#

  • mergeCompartments: a merged metabolite is the copy already in compartment s, if any, otherwise the copy whose id sorts first, with every field of that copy (was: the first copy in model order), so ACCOA_c and ACCOA_m become ACCOA_c whatever their order.

  • simplifyModel with constrainReversible: on a model without a reversible reaction this step changes nothing (was: an error "logical indices ... outside of the array bounds", as getAllowedBounds took the empty reaction list for all reactions). prepINITModel therefore runs on a model whose reactions are all irreversible after the first simplification, with the result of raven-toolbox's prep_init_model.

  • removeGenes: the help states that, when at least one of the requested genes is in the model, genes no remaining reaction uses are removed as well; when none is, no gene, reaction or bound changes.

Gap-filling#

  • gapFillTopological: without targets and without an objective reaction it raises RAVEN:badInput (was: an error without identifier). Help states the default seeds and targets.

  • checkProduction with checkNeededForProduction: the returned model holds excretion reactions only (was: the excretion reaction of the last metabolite in notProduced was returned as an uptake reaction). This matters for code that uses the fifth output.

Curation#

  • renameModelGenes with a file path as geneTable: the table is read and the genes are renamed (was: an error MATLAB:invalidConversion when the message that names the file was printed).

Queries#

  • getIndexes with type 'metnames' and returnLogical true: raises RAVEN:badInput (was: the error "Unable to use a value of type cell as an index"). A name can match several metabolites, so no single logical vector is defined.

Analysis#

  • findGeneDeletions with 'dgd': grRatioMuts has one element per gene pair, as the help states (was: twice as many, the second half zero). This matters for code that indexes grRatioMuts by the number of its elements.

Context-specific models (ftINIT / tINIT)#

  • ftINIT with resolveTies: each main-extraction step takes its fluxes from the least-total-flux solution for the tie-broken on/off state, so the direction and magnitude at which the next step forces a reaction do not depend on the solver's vertex. That phase stops at a relative or absolute gap of 1e-6.

  • ftINIT with resolveTies: among the least-total-flux solutions, each main-extraction step takes its fluxes from the one with the least id-rank-weighted backward flux, so the direction of a flux that carries the same total either way no longer depends on the solver. The id-rank phases rank a merged reaction by the smallest original reaction id in its group, so they do not depend on which member names it.

  • ftINITInternalAlg with resolveTies: when the primary solve stops at the time limit, the tie-break phases are skipped and the primary solution is kept, as in raven-toolbox (was: the phases ran on the unproven incumbent). The count and id-rank phases get half of params.TimeLimit each, rounded down and at least 1 s (was: the full limit).

  • ftINIT with resolveTies: the last step skips the minimal-flux and direction phases, as no step reads its fluxes (new ftINITInternalAlg argument pinFluxes). The reaction set does not change; the fluxes and metProduction of the last step are those of the id-rank phase.

  • ftINITInternalAlg: new optional positional arguments rankKeys (14th, the key each reaction is sorted by in the id-rank phases, default model.rxns) and pinFluxes (15th, default true). The "tie resolution did not converge" warning of resolveTies also names phase2c-minflux and phase2d-direction when the minimal-flux or direction phase does not prove optimal.

  • mergeLinear: groups are numbered 1, 2, ... in the order of the first reaction of each group in origRxnIds (was: the order in which the reactions were merged).

  • mergeLinear: a reaction flipped after the first merge pass is the one marked in reversedRxns (was: the reaction at the same index of the original model).

Utilities#

  • grRuleToDNF: help states the grammar it reads, that of parseGrRule (and/or in any case, &, &&, |, ||, operators as whole words, so G1&G2 is one gene id), and that a malformed rule raises RAVEN:badGrRule.

  • checkRaven: when the binaries cannot be downloaded, the message names the RAVEN-<tag>-full.zip bundle attached to each release for an offline install (was: "the *-binaries RAVEN release").

Homology and curation#

  • getModelFromHomology has the options complexPolicy (flag, keep or drop: how a reaction is treated when an and-linked subunit has no ortholog), keepGeneFree (reactions of the template without a grRule are transferred) and preserveNotes (rxnNotes and rxnConfidenceScores of the template are kept; rxnReferences are always kept). A gene that several template genes reach appears once per grRule. The draft has no version, date or description, and the descriptive annotation of the template is not copied (was: those of the templates). A preferred order that removes no gene is accepted (was: an error).

  • getModelFromHomology raises an error that names the gene ids of the new organism that split in a grRule (whitespace, parentheses, | or &, for example the UniProt id sp|P33284|HXK_KLULA) (was: the ids were accepted).

  • New: hitsFromSuppOrth reads a suppOrth.tsv into a blast structure for getModelFromHomology, in both directions, with filters on support, predictors, identity and E-value and the option proteinToGene for a protein to gene mapping. runSuppOrth runs the suppOrth command and reads its output.

  • curateModelFromTables reads the tables by header name in any order, with optional columns, and writes every supplied field to existing entities (was: metNotes and rxnReferences were not written). The MIRIAM columns follow the core columns, and grRules go to the right reaction when existing and new reactions are mixed. An empty grRules, lb, ub or rev cell leaves the existing value; the new option overwriteWithEmpty (default true) decides whether other empty cells clear a field. Cells are read without evaluating text: numeric cells accept plain decimals, Inf and NaN, and lb, ub and rev accept true and false; commas, hexadecimal numbers and expressions raise RAVEN:badInput. A file, column, gene-rule or repeated-id error raises RAVEN:badInput.

  • scaleBiomassFraction raises an error when the balancing component would get a fraction of 0 or less (was: the balancing pseudoreaction was inverted), when it has no current fraction (was: balancing was skipped) or when it is not a component of the biomass. The new argument targetTotal sets the balanced total of the biomass.

  • setGAM and scaleBiomassPseudoreaction raise an error before the model changes: setGAM when the NGAM reaction is missing and scaleBiomassPseudoreaction when the proton metabolite is missing. setGAM refuses ngamValue without ngamRxn, and an empty ngamRxn counts as not given.

  • sortModel keeps the directions of the reactions unless sortReversible is requested (was: the default sorted them). generateNewIds takes the numerically largest id and counts only finite integer suffixes.

  • replaceMets: the final contractModel call is restricted to the surviving metabolite (was: unrelated duplicate reactions were merged as well).

  • addIdentifierPrefix prefixes whole gene ids in grRules (was: an id with a dash, dot or colon, for example 5-HT1A, got a prefix at every word start, so exportModel failed).

  • getIndexes with an ambiguous name[comp] raises an error that names the matching metabolites.

  • gapFillMILP adds database reactions only and does not reverse draft reactions, which keep their bounds. The second output reversedRxns and the option weights are removed. bigM replaces infinite bounds of universal reactions; a finite bound is used as it is (was: bounds capped at bigM).

  • prepINITModel raises RAVEN:badInput when the task structure has no id field or a non-text id.

  • getKEGGModelForOrganism works with the kegg118 reference model, which has no rxnNotes (was: an error when adding the provenance note).

  • gapReport reports the metabolites that are produced or consumed with the boundary closed for every model (was: only for models with an unconstrained field). It closes the reactions of getExchangeRxns when the model has no unconstrained field, and stops with RAVEN:infeasibleModel when the model is infeasible under its own bounds or with its boundary closed, for example with a forced ATP maintenance flux.

Solvers#

  • optimizeProb returns the logical res.hitTimeLimit for every solver. runINIT and ftINITInternalAlg raise "Time limit reached without finding a solution" (was: "The problem is infeasible" with every solver except Gurobi). getMinNrFluxes returns exit code -2 for a solution that is not proven optimal at the time limit (was: 1 with Gurobi), and -1 with a warning when no solution came back.

  • optimizeProb solves MILPs, including those of gapFillMILP, fillGaps, getMinimalMedium, getMinNrFluxes, ftINIT and runINIT, with the bundled GLPK and on the GLPK of the COBRA Toolbox, with a RAVEN:glpkMILP warning (was: refused).

Sampling and reporting#

  • reporterMetabolites uses the population standard deviation for the background, which gives the scores of raven-toolbox (was: the sample standard deviation). The score is 0 when the spread of the background is 0 or at rounding level, P-values outside [0, 1] are refused, and metabolites without a score are dropped from all result fields.

  • sampleChebyshevCenter passes a sparse constraint matrix, so randomSampling with the method 'chrr' works with Gurobi (was: Gurobi rejected the dense matrix).

  • sampleACHR corrects the flux vectors towards S*v = 0 every few steps and before each sample, and keeps them within the bounds. It raises the warning sampleACHR:steadyStateNotRestored when a sample keeps a residual above 1e-9 relative (was: the residual of S*v grew exponentially, up to 9 after 100 steps).

  • haveFlux returns all reactions as blocked when the simplification removes every reaction (was: a GLPK error on an empty matrix).

Tasks#

  • parseTaskList reads false and 0 as false, trims ; lists, ignores the case of headers, keeps empty CHANGED bounds as NaN, and reads no as false. fitTasks leaves a bound unchanged when the CHANGED LB or UB is NaN. An absent IN LB, IN UB, OUT LB or OUT UB column gives the default bounds 0 and 1000. Comment rows and orphan rows are skipped with a warning, a non-numeric bound raises RAVEN:badInput, and a binary .xls file is refused with RAVEN:badInput (save it as .xlsx or .txt).

  • checkTasks warns when a task metabolite matches several model metabolites. removeLowScoreGenes keeps the first gene of the rule on a tie.

  • ftINIT removes the genes of removed reactions from the output.

Documentation#

  • Help blocks of 176 files are rewritten in impersonal, present-tense wording with British spelling; the code is unchanged. Statements that differed from the code are corrected, for example simplifyModel deleteDuplicates (first occurrence kept), removeBadRxns refModel, replaceMets outputs, getMinNrFluxes params, checkRaven developMode and the downloadRavenBinaries offline bundle. reverseRxns has the RAVEN 3 help structure.

3.0.0b1 (2026-09-07)#

First beta toward 3.0. Changes below are relative to RAVEN 2.9.3. The last 2.x release is 2.11.4: a reorganized codebase, fewer toolbox dependencies, hardened I/O and MILP algorithms, plus capabilities backported from the raven-toolbox (Python) port.

Repository & tooling#

  • Repo reorganized from a flat layout into functional folders (analysis/, annotation/, biomass/, comparison/, conversion/, gapfilling/, io/, manipulation/, queries/, solver/, utils/, etc.).

  • Function help blocks reformatted to NumPy-style docstrings; design-rationale, version-history and cross-language comments trimmed to current-state prose.

  • Allow for positional-or-named function arguments via parseRAVENargs. Essential arguments are positional, non-compulsory arguments are named.

  • Dropped the Statistics, Bioinformatics, Text Analytics and Optimization Toolbox dependencies.

  • All 317 dispEM call sites across 67 files replaced with native warning()/error(); new ravenList for formatted item lists (fixed: was silently dropping its own indent).

  • checkRaven replaces checkInstallation, moved to the repo root, faster checks.

  • Universal progress reporting (progressReport).

  • Poor performing visualization functions removed.

Reconstruction (KEGG & homology)#

  • KEGG HMM libraries and external binaries moved out of git to on-demand, SHA256-verified fetch from the raven-data release repo.

  • KEGG artifact set updated 116 → 118.

  • Legacy KEGG artifact-builder functions removed, moved to raven-toolbox (Python).

  • Recalibration of default HMM cutoffs.

  • MetaCyc reconstruction removed, as it persistently yielded high false positive annotations.

  • getModelFromHomology: recalibration of minLen default, token-based GPR rewrite (replaces a buggy cleanup regex).

  • New assignCompartments: functionality-constrained compartment-assignment MILP (WIP).

Curation#

  • Genome data utilities for homology-based reconstruction: getGeneData (gained a UniProt output column from the GFF Dbxref field; fixed crash when a CDS's Parent resolves to no gene), downloadGenomeData, processProteinFastaFile.

  • curateModelFromTables: fixed missing metabolite-support check.

I/O#

  • Legacy pre-L3V1/FBCv2 SBML import removed, only L3V1/FBCv2 remains supported.

  • importExcelModel removed, to discourage the practice of curating models in Excel.

  • New model.ec export to Excel (ENZYMES/ENZRXNS sheets).

  • Apache POI Excel I/O replaced with a dependency-free OOXML writer.

  • Changed YAML file format, to better align with cobrapy and have full compatibility with raven-toolbox (Python): fixed quoting; multi-line entries; position-dependent defaults; objective_coefficient import; empty/missing gene sections; compartment annotations; EC-code annotation round-trip.

  • Support for rxnDeltaG/metDeltaG in SBML.

Manipulation#

  • addRxns: new string keyword for metsBy; fixed gene ids shredded by and/or.

  • addTransport: fixed row-oriented metNames; crash on a compartment with no counterpart metabolite; rxnConfidenceScores/rxnDeltaG defaults.

  • mergeCompartments: fixed inverted unconstrained warning; stale compMiriams.

  • mergeModels: fixed metFrom handling; same-batch id-rename collisions.

  • convertToIrrev: fixed wrong reverse-copy index.

  • setExchangeBounds: fixed missing multi-exchange metabolites.

  • closeModel: fixed b growth and metNotes padding.

  • replaceMets: fixed three bugs.

  • deleteUnusedGenes is now a thin wrapper over removeReactions.

  • removeLowScoreGenes: fixed gene field alignment.

  • permuteModel: fixed compartment remap.

  • simplifyModel: fixed dropping task constraints during gap-filling.

Queries#

  • checkModelStruct: struct-array issues output; fixed two bugs; grRules start/end and/or handling.

  • findPotentialErrors made public, with parse-tree detection; grRules now parsed into a tree, also fixing removeGenes and expandModel.

  • getExchangeRxns: fixed index return in both branches, and exchange/transport misclassification for boundary-metabolite models.

  • getIndexes: hash-map lookup, islogical fix.

  • getElementalBalance: empty-reaction fix, new charge-balance check.

  • constructS/parseFormulas: fixed output corruption.

  • findDuplicateRxns rewritten with unique(…,'rows'): O(n log n) instead of O(n²).

Gap-filling#

  • New LP/MILP/topological gap-filling algorithms: gapFillFastCore, gapFillSwiftCore, gapFillMILP and gapFillTopological, in addition to the original fillGaps function.

  • makeSomething/consumeSomething merged into findLeakMetabolite(model, direction, …) (originals kept as thin wrappers).

Context-specific models (ftINIT / tINIT)#

  • getINITModel/runINIT deprecated, in favor of ftINIT (a separate, faster reimplementation that does not call runINIT). Code is still available for now, as published Human-GEM-derived models were built with tINIT.

  • ftINIT gained per-reaction big-M, essential-reaction clamping, tree-based gene removal; runINIT gained per-reaction big-M.

  • ftINIT gained resolveTies/proveAbsGap for deterministic MILP tie-breaking.

  • ftINIT: fixed metabolomics spdiags crash; MILP handling under the free solver; gap-fill failures reported as success; inverted allowExcretion sense in ftINITInternalAlg.

  • ftINIT task gap-filling now edits the reference model in place instead of copying it.

  • getMinNrFluxes: fixed crash on all-non-negative scores.

  • sortReactionOrder: fixed scoring the wrong columns.

  • checkTasks: task loop parallelized via parfor.

Analysis & flux sampling#

  • FSEOF regression-based target selection with knockdown/knockout classification.

  • reporterMetabolites exact closed-form background correction.

  • getFluxZ: fixed inverted sign in the zero-variance branch.

  • New analysis utilities: modelSummary, walkFluxes, traceFluxPath, compareFluxes.

  • New getMinimalMedium: MILP-based minimal medium finder.

  • randomSampling: FVA-based goodRxns selection, reproducible seed parameter; fixed undefined variable; per-reaction LPs parallelized via parfor; CHRR and ACHR flux samplers added.

  • compareRxnsGenesMetsComps folded into compareMultipleModels.

  • New diffModels: structured model comparison.

Biomass#

  • fitParameters: fixed every xRxn's upper bound set from the first.

  • guessComposition: fixed unknown metabolite's own coefficient.

Localization#

  • WoLF PSORT localization loader removed, replaced by DeepLoc2, MULocDeep, COMPARTMENTS, UniProt (WIP).

Solver#

  • optimizeProb: fixed dead no-op block; SCIP translation; gurobi results; MILP guard.