What this entity is
ModelSEED is a metabolic-reconstruction and biochemical-reaction resource used with rast annotations to build draft metabolic models and perform gapfilling. 1
- Canonical name: ModelSEED. 1
- Known aliases: ModelSEED metabolic reconstruction resource; ModelSEED biochemical-reaction resource. 1
- Stable external identifier: No stable external identifier was reported in the source document. 1
The BERDL Data Atlas inventories 56,012 ModelSEED reactions and 45,708 ModelSEED compounds, refining the resource’s documented role with a measure of its catalog scale. 2 The atlas also identifies ModelSEED as a reference layer for cross-tenant genome, annotation, pathway, and biochemistry analyses, supporting its use as an integration point rather than only a standalone reconstruction resource. 2
The ENIGMA Carbon Census further supports this integration role by using ModelSEED alongside PubChem, KEGG, and genome-depot annotations in a compound-to-genome knowledge census. 3 Its annotation-ceiling analysis refines the interpretation of ModelSEED coverage: simple, common, pollutant-adjacent compounds were more likely to be represented in ModelSEED and related resources, so apparent biological darkness can partly reflect resource coverage rather than absence of catabolism. 3
Role in annotation-gap discovery
The study used ModelSEED/RAST annotations and cobrapy to construct draft metabolic models for 14 organisms across 574 organism–carbon-source combinations. 1
Baseline flux-balance analysis (FBA), a constraint-based method for predicting metabolic flux and growth, achieved 42.5% overall accuracy, with recall of 86.5% corresponding to 244 of 282 growth-positive conditions correctly predicted and precision of 42.5% corresponding to 244 of 574 growth predictions correct. 1
The baseline models produced 330 false positives, meaning predicted growth on carbon sources that the organisms could not use experimentally. 1
Conditional ModelSEED gapfilling was applied to 38 false-negative cases and added 219 reactions: 201 enzymatic reactions, 14 transport reactions, and 12 exchange reactions, averaging 5.8 added reactions per case. 1
Among the 201 gapfilled enzymatic reaction–organism pairs, 96 pairs, or 47.8%, received candidate genes through integration with kescience-fitnessbrowser phenotypes, pangenome annotations, gapmind pathway evidence, and BLAST homology. 1
The resulting assignments included 44 high-confidence pairs, or 21.9%; 19 medium-confidence pairs, or 9.5%; and 33 low-confidence pairs, or 16.4%, while 105 pairs, or 52.2%, remained unresolved. 1
The study generated 574 gapfilling results, 219 gapfilled reaction details, 109 carbon-source-to-ModelSEED-compound mappings, and 201 master reaction–gene candidate records. 1
Compound-linking and integration limits
The Web of Microbes (WoM) explorer supports ModelSEED’s role as a cross-collection compound layer but refines expectations about identifier quality. Of 257 identified, non-unknown WoM compounds, 69 (26.8%) had definitive ModelSEED links through exact name matching; 107 (41.6%) had formula-only matches, yielding 176 compounds with any link (68.5%) and leaving 81 unmatched (31.5%). 4 The 107 formula-only compounds expanded to 900 ModelSEED molecules, an average of 8.4 ModelSEED molecules per WoM compound, so exact name matches are high-confidence 1:1 mappings whereas formula-only matches are candidate sets requiring manual curation. 4
This compound-link ambiguity refines the interpretation of the 109 carbon-source-to-ModelSEED-compound mappings produced in the annotation-gap study: a mapping can provide a useful modeling candidate without establishing a unique biochemical identity. 1,4 The WoM assessment also found that GapMind pathway matching was blocked by internal pathway identifiers rather than simple metabolite names, indicating that pathway-to-substrate/product lookup is needed before ModelSEED-linked analyses can be interpreted consistently. 4
Limitations and interpretation
ModelSEED gapfilling is non-unique because multiple valid reaction sets may explain a false-negative case, and the study used the default procedure, which minimizes the number of added reactions without guaranteeing biological optimality. 1
The draft models contained systematic errors from automated annotations, and their 42.5% baseline FBA accuracy was dominated by false positives, indicating overly permissive model behavior. 1
The study’s knockout validation was inconclusive because the models could not grow on carbon-source minimal media without the gapfilled reactions, making tests of genes assigned to those reactions circular in this setting. 1
The Carbon Census extends this limitation from individual draft models to a broader resource-defined gap: 74/83 enrichment compounds were organism-dark, including 33 KEGG-linked compounds with no reaction in queried genomes and 29 fully orphan compounds. 3 These categories identify ModelSEED- and related-resource pathway coverage, rather than confirmed biological non-use, as a target for metabolic-model gapfilling and enrichment experiments. 3
The atlas’s cross-tenant inventory supports the resource’s broader integration value, but its warning that schema-level join-key presence does not establish valid value-space overlap refines how ModelSEED-linked bridges should be interpreted: proposed joins require live validation rather than relying on table compatibility alone. 2
The evidence-triangulation results support metabolic-model-gapfilling by showing how ModelSEED gapfilling can supply candidate reactions for integration with phenotype, annotation, pangenome, pathway, and sequence evidence, while also exposing model-quality and solution-nonuniqueness limitations. 1
Related pages
- Annotation-Gap Discovery via Phenotype-Fitness-Pangenome-Gapfilling Integration — source report describing the integrated annotation-gap discovery study. 1
- BERDL Data Atlas — Inventory, Topic Map, and Cross-Reference Synergies — atlas inventory quantifying ModelSEED’s reaction and compound reference layers. 2
- ENIGMA Carbon Census — Tiered Knowledge Census of 83 Enrichment Compounds — compound-to-genome census that evaluates annotation coverage and organism-dark compounds. 3
- Web of Microbes Data Explorer — WoM compound-link assessment and cross-collection integration analysis. 4
- metabolic-model-gapfilling — cross-document synthesis of metabolic-model gapfilling. 1
- multi-omics-integration — synthesis of evidence integration for annotation and functional inference. 1
- cobrapy — modeling framework used with ModelSEED/RAST annotations. 1
- rast — annotation source used in draft-model construction. 1
- gapmind — pathway-completeness resource integrated with gapfilling evidence. 1