Metabolic evidence is not directly interchangeable when assays measure different biological capabilities or environmental conditions. Exometabolomics measures extracellular metabolite emergence or increase, whereas growth assays measure whether organisms can use a compound to support growth; this distinction supports interpreting metabolic overflow as release or accumulation of metabolites that a microorganism does not necessarily re-assimilate. 1 The Metabolic Consistency of Pseudomonas FW300-N2E3 Across Four KBase Data Lakehouse Databases compared Web of Microbes exometabolomics, Fitness Browser mutant-fitness experiments, BacDive species-level utilization phenotypes, and GapMind pathway predictions for Pseudomonas FW300-N2E3, showing that apparent agreement depends on assay scope and growth medium. 1 The Metabolic Capability vs Dependency refines this conclusion by distinguishing genomic capability from experimentally observed dependency: among 161 organism–pathway pairs spanning 7 model bacteria and 23 GapMind pathways, complete pathways could be either active dependencies or latent capabilities under the tested conditions. 2 The Essential Metabolome: GapMind Pathway Analysis Across Essential-Gene Organisms supports this distinction while showing that pathway-completeness results can look highly conserved in a small organism panel: 17 of 18 amino-acid biosynthesis pathways were complete in all 7 mapped organisms, but this computational result did not establish pathway essentiality or activity. 3 The Web of Microbes Data Explorer extends the database-scope qualification: its 2018 Web of Microbes snapshot distinguishes metabolites that increased from those that emerged de novo, but records no organism consumption actions, so production evidence cannot be treated as evidence of substrate use. 4
Production, utilization, and ecological secretion
The FW300-N2E3 analysis supports the hypothesis that some produced metabolites serve ecological exchange, signaling, or other extracellular functions rather than self-catabolism. This interpretation is strongest when a metabolite is detected as produced, supports organismal growth in a matched Fitness Browser condition, but has little or no species-level utilization in BacDive. 1 Production plus non-utilization is not intrinsically contradictory: extracellular production can represent overflow metabolism, biosynthetic byproduct release, or active secretion for ecological purposes, whereas BacDive growth assays measure utilization as a growth capability. 1
This comparison does not show that the producing strain cannot synthesize or use the metabolite. It separates biosynthetic production, catabolic utilization, and regulatory or ecological roles that the production-versus-growth comparison alone does not resolve. 1 The proposed transfer of a released metabolite to another organism therefore remains a hypothesis, not a demonstrated cross-feeding event. 1 This supports community-metabolic-interdependence as a framework for testing whether released metabolites become resources for neighboring organisms, and refines occurrence-versus-catabolic-activity by showing that extracellular detection and growth-based utilization can describe different biological roles rather than contradictory observations. 1
Evidence from FW300-N2E3
The Metabolic Consistency of Pseudomonas FW300-N2E3 Across Four KBase Data Lakehouse Databases integrated Web of Microbes exometabolomics, Fitness Browser mutant-fitness data, BacDive utilization phenotypes, and GapMind pathway predictions for pseudomonas-fw300-n2e3. 1 FW300-N2E3 produced or increased 58 Web of Microbes metabolites: 27 emerged and 31 increased. Only 21 of the 58 metabolites could be cross-referenced against at least one other database, so ecological-secretion interpretations apply to a tested subset rather than the full exometabolome. 1 The analysis drew on three data collections. The kescience_webofmicrobes collection supplied the FW300-N2E3 exometabolomic profile through its observation, organism, compound, environment and project tables. The kescience_fitnessbrowser collection supplied gene fitness on C/N sources through its genefitness, experiment, gene and seedannotation tables. The kescience_bacdive collection supplied species-level metabolite utilization through its metabolite_utilization, strain and taxonomy tables. 1 The report’s concordance_summary.png figure has two panels for the 58 WoM-produced metabolites: a pie chart of concordance categories and a histogram of cross-database coverage. Its consistency_heatmap.png figure is a 58-row heatmap showing WoM, Fitness Browser, BacDive and GapMind signals per metabolite. 1
Across the 21 testable metabolites, 17/21 (81%) were fully concordant across all matched databases, 4/21 (19%) were partially concordant, and none were fully discordant; the mean concordance score was 0.94. 1 Across 41 individual metabolite-database comparisons, 37 were concordant (90.2%). Fitness Browser comparisons were concordant for 21/21 (100%) metabolites and GapMind comparisons for 13/13 (100%), whereas BacDive utilization was concordant for 3/7 (43%) matched metabolites. 1 The matched-set results support cross-database integration when metabolite identity, organismal context, and measured capability are aligned, but do not establish concordance for the 37 metabolites observed only in Web of Microbes and not tested in BacDive, predicted by GapMind, or used as Fitness Browser conditions for this organism. 1 Both the report and the central digest describe the 94% mean concordance as largely structural. Fitness Browser (21/21 = 100%) and GapMind (13/13 = 100%) are invariably concordant with WoM production, because Fitness Browser always shows growth for metabolites used as sole C/N sources and GapMind predicts complete pathways for common amino acids. The only variable component is BacDive (3/7 = 43% utilized), and overall 37/41 comparisons are concordant (90.2%). A binomial test found that BacDive utilization of WoM-produced metabolites (3/7 = 43%) did not differ significantly from the overall P. fluorescens baseline (22/80 = 27.5%) (p = 0.40). The report interprets this as utilization at a rate consistent with the species norm. 1,5 This null result qualifies the high concordance: it reflects assay design rather than independent confirmation. 1,5 A sensitivity analysis excluded two approximate Fitness Browser matches, Cytosine→Cytidine and Uracil→Uridine, which are base→nucleoside mappings. This reduced testable metabolites from 21 to 19 and shifted mean concordance from 0.937 to 0.930. The change is negligible because both approximate matches were already fully concordant. 1
The 3/7 BacDive utilization result was compared with an overall Pseudomonas fluorescens (pseudomonas-fluorescens) baseline of 22/80 (27.5%), with p = 0.40 for the binomial comparison. 1 The BacDive evidence is aggregated across a broad GTDB-reclassified pseudomonas e fluorescens_E clade and therefore cannot be transferred directly to FW300-N2E3. 1
The clearest production-versus-utilization discordance was tryptophan. FW300-N2E3 increased tryptophan in its exometabolome, had 231 genes with significant Fitness Browser effects when grown on tryptophan, and had a complete tryptophan-biosynthesis pathway prediction from GapMind. In contrast, 0 out of 50 Pseudomonas fluorescens strains in BacDive could utilize tryptophan as a carbon source. 1 This pattern supports the hypothesis that tryptophan production can coexist with growth-associated genetic requirements without species-level tryptophan catabolism, potentially because of overflow, signaling, or cross-feeding. 1 It is consistent with the hypothesis that tryptophan overflow could support cross-feeding or signaling rather than catabolism, but does not show that FW300-N2E3 itself cannot grow on tryptophan. 1
The report retains a separate literature-context value of n=52 alongside the primary 0+/50- result with n=50 because it does not reconcile those values. 1 Trehalose provides a weaker parallel: it was increased by FW300-N2E3, but the absorbed report states that 1 of 6 BacDive strains was positive for utilization, potentially reflecting osmoprotection rather than carbon-source metabolism. 1 The main cross-database analysis reports the corresponding result as 1+/5- with moderate confidence. 1 Lysine and glycine also showed production without BacDive utilization, reported as 0+/3- and 0+/1-, respectively; these sample sizes are insufficient for a confident ecological-secretion conclusion. 1 The central discoveries digest gives the same tryptophan comparison with a different BacDive count. It reports that FW300-N2E3 produces tryptophan (WoM: Increased), has 231 genes with significant fitness when grown on tryptophan, and has a complete tryptophan biosynthesis pathway (GapMind). By contrast, 0/52 P. fluorescens strains in BacDive can utilize tryptophan as a carbon source. The digest calls this convergence a compelling case for overflow metabolism serving a cross-feeding or signaling function. That function is an inferred explanation rather than a measured transfer. The digest’s 0/52 is retained alongside the report’s 0+/50- without reconciliation. 5,1
The FW300-N2E3 report proposes that secretion by a prototroph unable to re-assimilate tryptophan could provide an amino acid to auxotrophic members of the groundwater community associated with the oak-ridge-field-research-center. This is a proposed ecological mechanism, not a demonstrated transfer to a recipient organism. 1 It connects to metabolic-capacity-specialization because pathway completeness and measured growth capacity do not establish that a produced metabolite is consumed by the same organism in its ecological setting. 1
Why assay outputs disagree
Web of Microbes exometabolomics measured compounds that FW300-N2E3 produced or increased, whereas BacDive growth assays measured whether Pseudomonas fluorescens strains could utilize compounds as carbon sources. 1 Web of Microbes used R2A rich medium, whereas Fitness Browser experiments used minimal medium with single carbon or nitrogen sources. 1 Thus a rich-medium exometabolome and a single-substrate fitness effect are condition-specific observations rather than identical tests of metabolism. 1 Because metabolic profiles are condition-dependent, some metabolites detected in WoM may be produced only on rich media. The report does not establish which metabolites are produced constitutively and which depend on the medium. 1 The cystic-fibrosis (CF) formulation study supports the same assay-scope caution in a different system. It reported a Pseudomonas aeruginosa PA14 carbon-utilization ranking of proline > histidine > ornithine > glutamate and said that this ranking mirrors the composition of SCFM, the synthetic CF sputum medium used in the study. However, its inhibition assays measured planktonic competition, whereas PA14 grows primarily in biofilms in CF lungs, so translation to structured biofilms is untested. 6 Its 22 tested carbon sources also missed mucins, lipids, iron, polyamines and the sugar alcohols that its genomic analysis predicted as important. The report proposes testing the 5 core species on N-acetylglucosamine, putrescine/spermidine, iron-limited conditions and pH 6.5, none of which the current panel covered. 6
The pathway-capability study supports and generalizes this limitation: carbon-source pathways were latent in 217 of 892 (24.3%) complete pathway-organism pairs, compared with 48 of 735 (6.5%) amino acid biosynthesis pairs, while 32.3% of all pairs were intermediate rather than clearly latent or active. 7 In the new 161-pair classification, 57 (35.4%) complete-and-important pairs were Active Dependencies and 66 (41.0%) complete-but-latent pairs were Latent Capability; all 66 Latent Capability pairs became fitness-important under at least one condition type, with nitrogen limitation, stress, and carbon limitation the most frequent triggers. 2 Because the condition-specific median importance threshold can cause reclassification by construction, independent calibration is needed rather than treating every reclassification as biological confirmation. 2 The latent fraction ranged from 4.7% to 21.1% across tested threshold combinations, reinforcing condition- and threshold-sensitivity. 7 Metabolic-model evidence also supports condition dependence. FBA (flux balance analysis) predicts metabolic fluxes from a constraint-based model. In Acinetobacter baylyi ADP1, 20% of genes (177/866) changed FBA flux class between rich and minimal media, which the report interprets as condition-dependent metabolic rewiring. 8 In a separate ADP1 analysis, glucarate showed a positive correlation (ρ = +0.246), opposite to prediction, which the report suggests reflects condition-specific model inaccuracies. This is a caveat from a single organism and a single condition, not a general finding. 9
The essential-metabolome analysis refines this condition dependence: its essential-gene measurements came from RB-TnSeq, or random-barcode transposon sequencing, in rich media, where nutrient supplementation can make biosynthetic genes appear non-essential. 3 Its complete GapMind pathways therefore do not prove that corresponding biosynthetic functions are essential for viability. The 7-organism panel contained 18 amino-acid pathway records and found 17 of 18 pathways complete in all 7 organisms, with serine incomplete in 1 organism; these were computational pathway calls rather than matched-condition activity measurements. 3
The Web of Microbes snapshot further refines the distinction by encoding E as emergence from an initially absent medium component and I as increase of a component already present. Across 10,744 observations these actions were mutually exclusive, but no organism had a decreased or consumption action. 4 Its proposed E/(E+I) metabolic-novelty fraction is a future phenotype for comparison with pangenome content, not an established predictor of pathway use. 4 The 19 produced metabolites for pseudo3_N2E3 that also appear as Fitness Browser carbon or nitrogen sources—5 emerged and 14 increased—provide a production-to-fitness bridge, but the absence of consumption actions prevents the stronger inference that the producing organism takes up the same compounds. 4 The central digest supports this reading of the encoding. D (Detected) and N (Not detected) are actions for the control medium, and all 742 D observations belong to the control, so the 2018 snapshot records only production, not consumption. 5
Coverage, annotation, and taxonomic limits
Only 21 of 58 FW300-N2E3 metabolites were cross-referenceable; the 21/58 (36%) testable fraction and 37/58 (64%) Web-of-Microbes-only fraction show that database coverage constrains comparability. 1 Of 257 identified non-unknown Web of Microbes compounds, 69 (26.8%) had definitive exact-name ModelSEED links, whereas 107 (41.6%) had formula-only links; formula-only links expanded to 900 ModelSEED molecules and therefore represent candidate sets rather than definitive identifications. 4 GapMind integration was blocked by internal pathway identifiers that did not expose simple metabolite names, so pathway-to-substrate/product mapping is required. 4 For FW300-N2E3, 28/58 (48%) WoM metabolites matched Fitness Browser carbon/nitrogen source conditions, 8/58 (14%) matched BacDive P. fluorescens utilization data, and 13/58 (22%) matched GapMind pathway predictions. 1 These low matching rates partly reflect naming differences among WoM compound names, Fitness Browser condition names and BacDive compound names. They also reflect differences in scope: many WoM metabolites, such as N-acetylated amino acids and nucleotide derivatives, have not been tested in Fitness Browser, BacDive or GapMind. 1 Manual curation identified 28 WoM-FB and 8 WoM-BacDive matches. Two of the Fitness Browser matches were approximate base→nucleoside mappings, and other matches may have been missed because of naming differences. 1 Only 21/58 WoM metabolites (36%) could be tested against any other database, and only 3 metabolites achieved four-way coverage. The results are robust where data exist, but the untested 64% may harbor additional discordances. 1 Name-mapping errors are a general risk. The annotation-gap analysis relied on manual curation to map 109 carbon sources from Fitness Browser condition names to ModelSEED compound IDs, and errors in that mapping would produce spurious false negatives. 10 Only ~2% of compounds in nmdc_arkin.metabolomics_gold have KEGG compound IDs populated, so pathway matching for that table must rely on compound-name substring patterns. Such matching risks collisions; for example, 'leucine' matches inside 'isoleucine', so longer, more specific patterns should be matched first. 5 One identifier-based alternative to name matching is a structural re-match of Fitness Browser Tier-1 carbon-source experiments through name→PubChem→InChIKey. 11
BacDive counts were computed after per-strain consensus, with duplicate records resolved by majority vote within strain. 1 The tryptophan, trehalose, lysine, and glycine comparisons had sample sizes of 50, 6, 3, and 1, respectively, so positive and negative proportions do not have equal evidentiary strength. 1 These results support reporting sample size with every species-level utilization rate and are relevant to taxonomic-resolution-dependent-functional-inference and environmental-resistome. 1
The pathway-capability study supports the same caution: only 48 organisms were fitness-tested among 293,000 genomes with pathway predictions, and its matched Tier 1 panel comprised only 7 of those 48 Fitness Browser organisms. 7 The essential-metabolome pilot likewise mapped only 7 organisms from an underlying 45-organism essential-gene collection; 0 GapMind predictions were available for the Escherichia coli K-12 Keio genome GCF_000005845.2, leaving the remaining 38 organisms unknown. 3
The Web of Microbes snapshot covered 37 organisms, including 20 experimental organisms, and was drawn from 5 ENIGMA-funded projects and a single 2018 archive. 4 All Web-of-Microbes genera had species-clade pangenome representation, but species-level matching required strain-to-genome mapping that was not attempted; genus-level comparisons are therefore exploratory rather than isolate-specific validation. 4 These limitations refine multi-omics-integration and cross-tenant-data-bridging: successful joining requires semantic alignment, identifier quality, and matched biological context in addition to matching names. 3
Fitness evidence and four-way agreement
Across 21 Web-of-Microbes metabolites with matching Fitness Browser experiments, FW300-N2E3 had significant effects for 601 unique genes and 4,764 total significant gene-condition hits, using |fit| > 1 and |t| > 4. 1 Of 31 Fitness Browser conditions mapped in the crosswalk, 7 returned no data because no FW300-N2E3 experiments exist for them: 4-aminobutanoate, 5-oxo-proline, Betaine, Guanine, Nicotinamide, Sarcosine and trans-Aconitate. The downstream analysis therefore used only the 24 conditions with actual fitness data. 1 A total of 231 genes were significant in 3 or more metabolite conditions, while the top 18 genes were significant in all 21 conditions. Broadly pleiotropic functions included methionine, histidine, branched-chain amino acid, leucine, and aromatic amino acid biosynthesis and were interpreted as housekeeping fitness requirements rather than substrate-specific catabolism. 1
Approximately 370 genes significant in only 1–2 conditions were treated as the more substrate-specific component. 1 Emerged metabolites had a mean of 184.4 +/- 115.5 significant genes, including means of 221.0 detrimental and 21.0 beneficial effects; increased metabolites had a mean of 163.5 +/- 74.1, including means of 206.5 detrimental and 12.8 beneficial effects. Both groups had detrimental-to-beneficial ratios of approximately 10:1, indicating that most significant genes were required for growth on tested substrates rather than inhibitory. 1
These findings support condition-specific-fitness while cautioning that a large fitness landscape does not identify the mechanism of production or utilization. 1 Pathway-level conservation likewise did not distinguish latent capabilities from active dependencies: mean conservation was 0.869 for latent pathways and 0.829 for active pathways, with Mann–Whitney U p = 0.94. 7 The deferred NB04 analysis, which would map fitness-important genes to specific GapMind pathway steps, leaves biosynthetic, catabolic, and regulatory contributions unresolved. 1 SEED-proxy pathway membership can introduce false positives relative to direct GapMind per-step assignments, and the new analysis mapped Fitness Browser genes through Fitness Browser-native KEGG annotations to GapMind pathways, leaving annotation and mapping completeness as constraints. 7 2
The essential-metabolome report supports this caution because GapMind can miss non-canonical or divergent enzymes and unannotated genes; its apparent Desulfovibrio vulgaris serine gap remains a hypothesis requiring growth testing, not evidence of established auxotrophy. 3 The Web-of-Microbes bridge supports pursuing the same analysis for pseudo3_N2E3, but refines interpretation by distinguishing its 5 emerged from 14 increased products and by documenting the lack of consumption observations. 4
Malate, arginine, and valine were the three metabolites with four-way coverage across production, Fitness Browser growth, BacDive utilization, and GapMind prediction. Malate was increased, supported growth involving 92 Fitness Browser genes, was utilized by 49/49 BacDive strains (100%), and had a complete GapMind pathway. Arginine was increased, supported growth involving 270 genes, was utilized by 40/48 strains (83%), and had a complete pathway. Valine emerged, supported growth involving 160 genes, was utilized by 1/1 strain, and had a complete pathway. 1 These are gold-standard examples of four-way consistency, but only 3 of 58 Web-of-Microbes metabolites; strong agreement is demonstrated for particular metabolite-condition combinations, not for all exometabolomic signals. 1
Tensions
The FW300-N2E3 matched subset shows 100% concordance for 21 Fitness Browser and 13 GapMind comparisons, whereas the broader metabolic-capability analysis finds complete pathways that are latent or intermediate under measured fitness criteria. 1 7 This is a scope difference rather than a numerical contradiction: one is a selected matched subset, while the other classifies complete pathway-organism pairs across organisms and conditions. The new analysis refines the distinction with 57 (35.4%) Active Dependencies and 66 (41.0%) Latent Capabilities. 2
The essential-metabolome pilot adds an assay-scope and coverage tension rather than contradicting these results: it found 17 of 18 amino-acid pathways complete in all 7 mapped organisms, but used rich-media RB-TnSeq and computational GapMind calls that cannot determine whether pathways are required under defined conditions. 3 The Web of Microbes evidence adds a related scope tension: it provides a production-to-Fitness Browser bridge for 19 metabolites, but its 2018 snapshot has no organism consumption action, whereas other comparisons include utilization or single-substrate growth evidence. 4
The trehalose observations, 1 of 6 BacDive strains positive in the absorbed FW300 report and 1+/5- in the main comparison, are compatible representations of the same result rather than a conflict. 1 For the scope tensions above, resolving work is to repeat matched comparisons across more organisms, pathways, and explicitly aligned media, while adding direct consumption measurements, strain-level mapping, and identifier-based compound curation.
Open Directions
- Map the 601 unique Fitness Browser genes and 4,764 significant gene-condition hits to GapMind pathway steps using the deferred NB04 analysis, and test whether genes associated with FW300-N2E3 tryptophan production are biosynthetic, catabolic, or regulatory. 1
- Repeat the four-database comparison for additional ENIGMA isolates, including pseudomonas-stutzeri-rch2, and test whether concordance and production-versus-utilization patterns are isolate-specific or reproducible. 1
- Expand Web-of-Microbes–BacDive matching with InChIKey or CHEBI identifiers, and test whether identifier-based joining increases the 8/58 BacDive match count without conflating biologically distinct compounds. 1
- Repeat Web-of-Microbes profiling across growth media to determine whether tryptophan and trehalose production is constitutive or medium-dependent. 1
- Build a community metabolic model for the Oak Ridge groundwater community to test whether FW300-N2E3 tryptophan release can support predicted auxotrophic recipients. 1
- Apply direct GapMind gene-step assignments and matched-condition fitness assays across the 48-organism panel to test whether latent classifications reflect ecological inactivity or missing laboratory conditions. 7
- Recalibrate condition-specific importance thresholds against independently known essentials, then test whether the 66 Latent Capability pairs remain conditionally important without median-threshold circularity. 2
- Extend matched capability–dependency comparisons beyond the 7 model organisms and compute phylogenetically independent relationships between pathway variation, pangenome openness, and metabolic ecotype counts. 2
- Test the apparent D. vulgaris serine auxotrophy on serine-free minimal medium, inspect lower-confidence GapMind steps-missing calls, and expand organism-to-genome mapping beyond the 7-organism pilot. 3
- Combine GapMind with eggNOG EC-to-KEGG pathway analysis and direct essential-gene mapping to determine whether conserved pathway completeness predicts condition-specific dependency across remaining mapped organisms. 3
- Obtain a current Web of Microbes or GNPS2/Northen laboratory dataset with consumption measurements, then repeat production-versus-use comparisons under explicitly matched media and organisms. 4
- Build a pathway-to-metabolite lookup table for Web of Microbes internal pathway identifiers and manually curate the 107 formula-only ModelSEED candidate sets before testing whether compound-resolution improvements alter concordance. 4