Community Metabolic Interdependence and Black Queen Predictions

Community metabolic interdependence is the hypothesis that organisms can rely on metabolites produced by neighboring organisms, allowing loss or reduction of otherwise costly biosynthetic functions. ₁ The Black Queen prediction tested here is that amino-acid biosynthetic capacity and environmental metabolite pools should show coordinated patterns consistent with community-level provisioning rather than uniformly independent synthesis. ₁ Morris et al. (2012) proposed the Black Queen Hypothesis as the named framework for this logic: costly genes can be lost when their functions are provided as public goods by community members. ₂ The capability-dependency project cites the same named work as Morris JJ, Lenski RE, Zinser ER (2012), “The Black Queen Hypothesis: evolution of dependencies through adaptive gene loss,” mBio 3(2):e00036-12, DOI 10.1128/mBio.00036-12. ₃ This page synthesizes evidence from discoveries, Community Metabolic Ecology via NMDC × Pangenome Integration, and Metabolic Capability vs Metabolic Dependency, now extended by Metabolic Capability vs Dependency, Metabolic Consistency of Pseudomonas FW300-N2E3 Across Four KBase Data Lakehouse Databases, and Metagenome-Prioritized Phage Cocktails for Crohn’s Disease and IBD, and now by Plant Microbiome Ecotypes — Compartment-Specific Functional Guilds and Their Genetic Architecture, and connects it to metabolic-model-gapfilling, multi-omics-integration, and costly-dispensable-gene-loss.

Evidence for Black Queen-like structure

Community-scale NMDC metabolomics supported the Black Queen prediction for amino-acid pathways: 11 of 13 pathways trended in the predicted direction, with a binomial sign-test p=0.011. ₁ The NMDC analysis supports this result across 80 GapMind pathways, with 11 of 13 testable amino-acid pathways showing negative Spearman correlations between community pathway completeness and ambient amino-acid intensity; the sign test was p = 0.011. ₄ The 220-sample × 80-pathway completeness matrix in that report is the broader pathway matrix behind the ordination described below, not the metabolomics cohort behind the amino-acid correlations. ₄ The digest describes that metabolomics test as running across 131 NMDC samples. ₁ The two associations reported as reaching significance were for leucine, which had r=-0.390, q=0.022, n=62, and arginine, which had r=-0.297, q=0.049, n=80; they were not the largest effects. ₁ The new analysis reports the same values and identifies these as the two associations surviving Benjamini-Hochberg false-discovery-rate correction. ₄ methionine shows the largest reported negative effect, r=-0.496, but is underpowered at n=18 with q=0.117, so its direction was consistent with the prediction while failing the reported significance threshold. ₁

The full H1 table states eleven of 13 (85%) testable amino-acid pathways as showing negative Spearman correlations between community pathway completeness and ambient amino-acid metabolite intensity, with a binomial sign test p = 0.011, and gives per-pathway sample sizes, correlations, unadjusted p values and Benjamini-Hochberg false-discovery-rate-adjusted q values (FDR, the multiple-testing correction used): met n = 18, r = −0.496, p = 0.036, q = 0.117; leu n = 62, r = −0.390, p = 0.002, q = 0.022 (FDR-significant); arg n = 80, r = −0.297, p = 0.007, q = 0.049 (FDR-significant); trp n = 44, r = −0.294, p = 0.052, q = 0.136; thr n = 23, r = −0.271, p = 0.211, q = 0.393; phe n = 88, r = −0.159, p = 0.140, q = 0.303; val n = 54, r = −0.151, p = 0.275, q = 0.446; ser n = 59, r = −0.061, p = 0.645, q = 0.762; asn n = 73, r = −0.061, p = 0.610, q = 0.762; ile n = 18, r = −0.057, p = 0.823, q = 0.823; chorismate n = 65, r = −0.038, p = 0.765, q = 0.823; gly n = 114, r = +0.081, p = 0.389, q = 0.562; tyr n = 26, r = +0.419, p = 0.033, q = 0.117; gln (n = 4) and pro (n = 9) were skipped. ₄

The figure figures/h1_bqh_barplot.png plots the H1 result as Spearman r per amino-acid pathway, with red marking FDR-significant negative associations as Black Queen support. ₄

In the digest’s framing the same test is a 131-sample NMDC H1 cohort in which 11/13 tested amino-acid pathways trend in the Black Queen-predicted direction, with leucine and arginine passing FDR correction. ₁ Both FDR-significant pathways are among the most energetically expensive to synthesize — leucine at 37 ATP equivalents and arginine at 26 ATP equivalents — which the digest reads as consistent with loss being most favored when the benefit of gene retention is lowest; this is an interpretive link rather than a measured energetic test. ₁ The leucine signal is reported robust to soil-only stratification (q = 0.022 unchanged) and to study-level blocking, although the dataset is 95% single-study. ₁ The cohort is also described inconsistently across the corpus: the digest calls all 131 H1 samples soil. ₁ The project report instead re-ran H1 on the 125 Soil-ecosystem samples only, excluding the 6 Unknown-ecosystem samples in the H1 dataset, so the cohort should not be described as uniformly soil. ₄ In that soil-only re-run leucine remained FDR-significant (r = −0.390, q = 0.022, n = 62 — identical to the full-cohort result, since the leucine samples are all Soil), arginine lost FDR significance (r = −0.264, q = 0.117, n = 78) while remaining directionally consistent, and the binomial sign test (11/13 negative, p = 0.011) was unchanged. ₄ The energetic reading of the two surviving pathways is an interpretation rather than a measured test of gene loss: the report states leucine at 37 ATP equivalents and arginine at 26 ATP equivalents as consistent with the prediction that costly functions are most likely to be lost when environmental supply is reliable, and its expectation that a larger metabolomics dataset would likely bring methionine (r = −0.50, n = 18, q = 0.117) to significance is itself untested. ₄ Part of the cohort’s soil weighting is structural rather than chosen: all 33 Freshwater samples lacked paired metabolomics in NMDC, so H1 is effectively a soil-only test and whether comparable Black Queen dynamics operate in freshwater communities at the same scale is untested. ₄

The result was not universal across metabolites: tyrosine was an anti-Black-Queen outlier with r=+0.42. ₁ The digest records this association as nonsignificant and offers phenylalanine hydroxylation as a proposed, untested explanation — an alternative tyrosine source that would decouple biosynthesis completeness from pool size. ₁ The NMDC analysis refines this value as r = +0.419 and likewise finds no significant isoleucine association (r = −0.057, q = 0.823, n = 18). ₄ Its proposed mechanism is specific and untested: tyrosine can be produced from phenylalanine via non-biosynthetic hydroxylation, so communities with high phenylalanine biosynthesis completeness — a quantity not captured in the tyr completeness score — may still supply tyrosine, decoupling the biosynthesis-versus-pool relationship. ₄ Thus, the dataset supports a graded, pathway-specific interpretation rather than the claim that all amino-acid biosynthesis follows one community-dependency rule. ₁ The negative associations therefore remain evidence compatible with community provisioning, not direct proof that organisms exchange the measured metabolites. ₄

A complementary organism-level analysis supports a community-context interpretation, but not direct exchange: among 1,695 complete pathway-organism pairs from 48 organisms, 267 (15.8%) were classified as fitness-neutral latent capabilities, while 881 (51.9%) were active dependencies. ₃ Latent capabilities were especially common for carbon-source utilization pathways (217 of 892, 24.3%) compared with amino-acid biosynthesis (48 of 735, 6.5%), and latent-capability rate correlated positively with pangenome openness at the clade level (Spearman ρ = 0.69, p = 0.0004, n = 22 clades). ₃ That correlation was computed after aggregating 41 organisms to 22 unique species clades, so that multiple Fitness Browser strains from the same clade share a single pangenome openness value, and the report marks this hypothesis supported, with clades carrying more fitness-neutral complete pathways tending to have more dynamic, less-conserved pangenomes. ₃ The latent class there was defined as mean |t-score| < 1.0 with fewer than 5% essential genes, and a further 32.3% of the 1,695 pairs fell in an intermediate zone that is neither clearly latent nor clearly dependent. ₃ The expectation that such losses pay is imported from cited literature rather than measured in this corpus: the report attributes to D’Souza et al. (2014) a finding that biosynthetic gene loss provides a ~13% fitness benefit when metabolites are available from neighbors. ₃ This refines the Black Queen prediction: genome dynamics and environmental context are consistent with selective sharing or reduced autonomous synthesis, but pathway presence plus fitness neutrality does not identify the producer, recipient, or exchanged metabolite.

The new pathway-capability analysis supports the condition-dependent interpretation: in a separate set of 161 organism-pathway combinations across 7 model bacteria, all 66 pairs classified as Latent Capability became fitness-important under at least one condition type, especially nitrogen limitation, stress, or carbon limitation. ₅ However, this result is partly sensitive to applying a median-based importance threshold within condition subsets, so it does not by itself establish that every latent capability is biologically shared or conditionally required. ₅

Carbon metabolism dominates community differentiation

Carbon utilization, rather than amino-acid biosynthesis, dominated metabolic differentiation in the analyzed communities. ₁ Principal-component analysis (PCA), an ordination method that summarizes major variance axes, found that PC1 explained 49.4% of variance and separated Soil from Freshwater at p<0.0001. ₁ The separation was assessed with the mann-whitney-u-test, a rank-based two-group test, on a 220-sample × 80-pathway GapMind completeness matrix, with median PC1 = +3.86 for Soil and −6.28 for Freshwater, and PC1 loadings dominated by carbon-utilization pathways rather than amino-acid pathways. ₁ Amino-acid pathways loaded more strongly on PC2, which explained 16.6%, separating sub-clusters within Soil. ₁ The NMDC analysis supports this interpretation: PC1 explained 49.4% and PC2 16.6%, while carbon pathways including glucuronate, fumarate, succinate, cellobiose, and galactose had near-uniform positive PC1 loadings. ₄ The matrix behind that ordination was built by extracting GapMind species-level pathway completeness and computing community-weighted scores for 220 samples × 80 pathways. ₄

This pattern refines the Black Queen interpretation: amino-acid associations are compatible with metabolite interdependence, but they explain less of the overall community metabolic structure than carbon-use differences in this analysis. ₁ The result therefore favors testing dependency within environmental and metabolic contexts rather than treating amino-acid exchange as the primary determinant of all community differentiation. ₁ The NMDC analysis further found significant ecosystem-type differences for 17 of 18 amino-acid pathways, but it could not determine whether these differences reflected habitat physicochemistry, taxonomic composition, or both. ₄ The digest reports the same count as 17/18 amino-acid pathways significantly differentiated across ecosystem types (q < 0.05 by kruskal-wallis-test, a rank-based test for differences among groups) while naming carbon-substrate availability, not biosynthetic capacity, as the primary driver. ₁ Because carbon-substrate range is the dominant ecological axis when community metabolic potential is used as a feature space, it will confound amino-acid pathway analyses unless controlled for; the digest recommends PC1-residualized or ecosystem-stratified analyses for Black Queen-type tests. ₁

Tensions

The metabolomics dataset was dominated by one study, with 125 of 131 samples (95%) coming from nmdc:sty-11-r2h77870, and the remaining 6 samples coming from a second study. ₁ This concentration limits how broadly the observed correlations can be generalized across environments and studies. ₁ The NMDC analysis supports this limitation: all 62 leucine samples came from that first study, although the arginine result was not driven by the minor study. ₄

Only approximately 2% of compounds had KEGG identifiers, and substring matching risked collisions such as leucine versus isoleucine. ₁ Three amino-acid pathways—cysteine, histidine, and lysine—were untestable because the relevant compounds were absent. ₁ The NMDC analysis confirms that the reported run corrected an isoleucine/leucine mapping collision, while cysteine, histidine, and lysine remained untestable. ₄ That correction mattered quantitatively: three isoleucine compounds had been misassigned to leucine and had near-zero completeness-versus-metabolite correlation, diluting the true leucine signal, so the fix moved leucine from r −0.326 to −0.390 and q 0.045 to 0.022. ₄ These limitations make the 11/13 directional result supportive but coverage-limited, because the untested pathways cannot distinguish universal from pathway-specific Black Queen behavior. ₁ A related measurement gap affects aromatic-pathway readouts: chorismate itself is rarely detected, so shikimic acid and 3-dehydroshikimic acid serve as upstream proxies that reflect precursor availability, not chorismate pool size. ₁ The NMDC analysis supports that caution for its own number, stating that these are upstream intermediates rather than chorismate itself and that the chorismate Black Queen correlation (r = −0.038) should be interpreted cautiously. ₄

The positive tyrosine association is a direct tension with the simple prediction that community provisioning should produce a common negative association pattern across amino acids. ₁ The corpus does not establish whether this outlier reflects ecological specialization, measurement or mapping limitations, or a genuinely different exchange regime. ₁ The NMDC analysis refines that uncertainty by noting that GapMind measures genomic potential rather than expression and that abiotic measurements were unavailable: pH, temperature, and total organic carbon were all NaN in the 174-sample analysis matrix. ₄ Their absence prevented partial-correlation tests controlling for environmental gradients, and the report states that these variables could confound both the amino-acid test (H1) and the ecosystem-differentiation test (H2). ₄

One numeric conflict inside the NMDC source should be recorded rather than reconciled. Its Literature Context compares the result to Gowda et al. (2022), who showed that metabolite dynamics in model communities are predictable from gene content, and quotes the environmental-scale values as leucine r = −0.33 and arginine r = −0.30; the leucine figure there conflicts with the corrected r = −0.390 reported elsewhere in the same source, and the source does not reconcile the two. ₄

A second tension comes from plant-associated communities, and its headline number has since been revised. The digest reports that co-occurring plant-associated genera show functional redundancy rather than complementarity (permutation p = 1.0, Cohen’s d = −7.54), which contradicts the expectation that co-occurrence signals complementary metabolic provisioning. ₁ The quoted p-value does not itself support a significant permutation result and that excerpt supplies no corrected estimate. ₁ The plant-microbiome project corrects the effect size: the Phase 1 NB06 value of cohens d = −7.54 is described there as an implausibly large effect and largely a Cohen’s d formula error, because NB06 divided the mean difference by the standard deviation of permutation means (which shrinks toward zero by the central limit theorem for any reasonable null) while the correct denominator is the raw pair-level standard deviation, which is much larger. ₆ The magnitude −7.54 should therefore not be carried forward as an effect size — a quality-assurance pattern treated in adversarial-research-quality-assurance — and only the direction survives. ₆

The canonical Phase 2b finding is that co-occurring plant-associated genera in NMDC soil communities are slightly less complementary than random genus pairs (Cohen’s d ≈ −0.4, permutation p < 0.001), supporting the redundancy interpretation but at a small effect size. ₆ Recomputing NB06’s max-aggregated complementarity with the consistent formula on the same data gives d ≈ −0.43, and switching the aggregation from species-max to species-prevalence (the I1 fix in NB14) further reduces |d| to 0.39; of the apparent 20× reduction in magnitude reported in the early Phase 2b draft, roughly 90% was the formula correction and about 10% the aggregation correction, and the direction is robust under both methods. ₆ Stated on its own, the prevalence-weighted aggregation produced approximately a 10% reduction in effect size with the direction unchanged. ₆

The permutation measurement itself is reported for the earlier max-aggregated NB06 complementarity analysis: of 69 genera in 348 soil/rhizosphere samples, all had gapmind data across 80 pathways, and the 1,048 co-occurring pairs had mean complementarity of 10.72 while the permutation null (1,000 iterations preserving richness) yielded 12.74±0.27. ₆ Separately, in its corrected interpretation the report states the direction as unchanged — co-occurring genus pairs are slightly less complementary than random pairs (permutation p < 0.001) — a result pointing to redundancy rather than the hypothesized complementarity and directly relevant to gene-cooccurrence-ecological-guilds; that p-value is given in the corrected discussion and is not attributed in the source to the NB06 means and null quoted above. ₆

Two resolution caveats stop this from closing the question. The report states that the GapMind pathway level (80 pathways) may be too coarse, and that the complementarity signal may exist at finer-grained metabolic resolution than the 80 GapMind pathway categories can detect, so redundancy is established at this resolution only — the evidence-resolution problem discussed in pathway-versus-reaction-evidence-resolution. ₆ It also records that core-level completeness scoring yielded 0% across all compartments, suggesting that threshold is too stringent for broad taxonomic comparisons, and leaves the core-level threshold as a separate open question even though the prevalence-weighted re-test produced a credible Cohen’s d = -0.39 in the same direction. ₆ Summarizing both corrections, the report says the prevalence-weighted re-test plus the Cohen’s d formula correction reduced the original implausible |d|=7.54 to |d|≈0.4 and that the redundancy direction remains robust at this resolution. ₆

The framing against prior literature carries the pre-correction magnitude and should be read with that in mind: the report reads the rejection of its complementarity hypothesis (H3) as consistent with Louca et al. (2018), who argued that functional redundancy is an emergent property of microbial systems, and in that passage states that its negative Cohen’s d (-7.54) indicates strong redundancy — the value its own later re-test revised to -0.39. ₆

A companion exclusion test could not be run at all, and the report treats that failure as a result. C-score analysis for plant-growth-promoting (PGP) versus pathogen exclusion was not feasible because only 0 PGP-dominant and 2 pathogen-dominant genera were represented in the NMDC co-occurrence data, the remaining 67 being dual-nature. ₆ A formal re-test on the Phase 2 refined cohorts returned 0 PGP-dominant / 3 pathogen-dominant / 66 dual-or-mixed of 69 NMDC co-occurring genera — the same pattern as Phase 1 — so the DESIGN-plan target of ≥5 PGP-only genera (verification #12) is documented as NOT MET with the failure-as-result interpretation made explicit, and it does not change H3. ₆ Because the dual-nature class swamps the others regardless of whether the Phase 1 25-marker or the Phase 2 17-marker panel is used, the C-score test is structurally underpowered in this dataset rather than a marker-panel artifact — the coverage limit discussed in callability-limited-comparative-inference, on organisms whose ambiguity is treated in dual-nature-plant-associated-bacteria. ₆

Single-organism auxotrophy predictions are likewise explained, not demonstrated, by community provisioning: the predicted serine loss in desulfovibrio-vulgaris-hildenborough is attributed to environments rich in amino acids from protein degradation, to streamlining metabolism being advantageous when nutrients are abundant, and to genomic economy in losing genes that are energetically costly when the product is freely available — all proposed explanations rather than direct findings, as also discussed in biosynthetic-prototrophy-and-auxotrophy. ₇

There is also a tension between the small condition-specific study and the broader capability-dependency analysis. Across 7 Fitness Browser organisms and 23 GapMind pathways, 41.0% of pathway-organism pairs were Latent Capabilities, and all Latent Capabilities became fitness-important under condition-specific analyses. ₁ By contrast, the broader analysis classified 267 of 1,695 complete pairs (15.8%) as latent under its stated fitness thresholds, with the latent fraction remaining 4.7%–21.1% across 16 threshold combinations. ₃ The new 161-pair analysis supports the first study’s condition-dependent result, but its 66 latent pairs and threshold caveat do not resolve the disagreement because it uses another organism set, pathway set, and median-based classification. ₅ The tension is whether latent capability is generally a condition-dependent dependency or whether a substantial fraction remains fitness-neutral across tested conditions.

Pathway-level conservation also did not distinguish latent capabilities from active dependencies: mean conservation was 0.869 for 248 latent pathways versus 0.829 for 755 active dependencies, with Mann–Whitney U p = 0.94 for active > latent and rank-biserial r = 0.052. ₃ This refines a simple gene-loss version of the Black Queen model: the reported signal appeared instead in the positive association between latent-capability rate and pangenome openness at the clade level. ₃

The report labels that pathway-level hypothesis (H2a) not supported and reads the two results together as mixed rather than uniform evidence for the Black Queen framework: the clade-level openness correlation (ρ = 0.69, p = 0.0004, n = 22 clades, H2b) provides clade-level support consistent with ongoing streamlining, while per-pathway conservation rates do not distinguish latent from active pathways, a divergence it interprets as the Black Queen signal operating at the level of genome dynamics and community context rather than being detectable in per-pathway cross-genome conservation ratios — a point that also bears on pangenome-conservation-fitness-decoupling. ₃

Four stated caveats limit how far that null can be pushed. The conservation rate captures only whether the pathway is complete in each genome in the species clade, a coarse measure that does not detect partial gene erosion — individual gene deletions within an otherwise complete pathway — and so cannot detect progressive gene loss within a pathway. ₃ The fitness experiments cover only 48 organisms, so many organisms in the same species clade may never have been fitness-tested, inflating apparent conservation of pathways that are “complete” but untested. ₃ Evolutionary lag is offered as a third explanation: the Black Queen Hypothesis predicts ongoing gene loss, but the time frame over which this is detectable may exceed what current genome sampling captures, so latent capabilities may be mid-loss in some lineages yet still appear conserved at species level. ₃ Fourth, the report attributes to Schakel et al. (2025) the finding that cross-feeding-based gene loss requires spatial structuring and is not straightforwardly predicted by the fitness cost of gene retention alone — consistent, in its reading, with why simple conservation-rate ratios do not discriminate latent from active pathways. ₃

Relationship to pathway-capability evidence

The community metabolomics result should be interpreted alongside pathway-capability and fitness evidence rather than as a standalone demonstration of metabolite exchange. ₁ Across 7 Fitness Browser organisms and 23 GapMind pathways, 35.4% of pathway-organism pairs were Active Dependencies, 41.0% were Latent Capabilities, 14.9% were Incomplete but Important, and 8.7% were Missing. ₁ All Latent Capabilities became fitness-important under condition-specific analyses. ₁

These results support the distinction between genomic capability and realized dependency: a pathway may be present yet become important only under particular environmental conditions. ₁ In FW300-N2E3, all 13 Web of Microbes metabolites that could be mapped to GapMind had complete pathways, and all 13 showed growth in Fitness Browser experiments. ₁ The overall 94% four-database concordance was structurally driven, however: Fitness Browser was 21/21, GapMind was 13/13, and BacDive was only 3/7. ₁ This means apparent agreement across resources cannot by itself establish ecological metabolite exchange. ₁ The NMDC results reinforce this caution because pathway completeness was compared with metabolite intensity, not pathway expression or experimentally demonstrated exchange. ₄

The broader capability-dependency study supports this distinction across more organisms: carbon pathways were more often complete but fitness-neutral than amino-acid biosynthesis, while only 48 organisms had fitness measurements among 293,000 genomes with pathway predictions. ₃ Its ecological clustering result is likewise observational: all 10 target species formed metabolic clusters, but environment-cluster associations were significant only for Salmonella enterica and Phenylobacterium sp., and explicit phylogenetic correction was not performed. ₃

The new analysis extends this evidence from latent fitness states to pathway variation: among 2,810 GTDB species, variable pathway count was positively associated with pangenome openness after controlling for genome count (partial Spearman ρ=0.530, p=2.83e-203). ₅ It also found that ecotype count was associated with openness (partial ρ=0.322, p=8.0e-07), supporting a link between genome fluidity and within-species metabolic diversity rather than proving metabolite sharing. ₅ Accessory-dependent completeness was strongest for several amino-acid biosynthesis routes, including exact core-versus-all gaps of 0.146 for leucine and valine and 0.141–0.140 for arginine, lysine, and threonine; this supports a Black Queen-compatible hypothesis of community-level provisioning but does not demonstrate exchange. ₅ That report names the branched-chain amino acids (leucine, valine) and the basic amino acids (arginine, lysine) as the most common candidates for community-level metabolic sharing — a candidate list read off completeness gaps rather than an established set of shared metabolites. ₅ Its mechanistic reading is that within a species, different strains have different biosynthetic capabilities, which it calls a precondition for Black Queen dynamics where metabolic functions become community-level public goods; that precondition is an interpretation of accessory-gene variation, not direct evidence of public-good exchange. ₅

Organism-level cross-feeding candidates

pseudomonas-fw300-n2e3 supplies an organism-level cross-feeding candidate that remains inferential. The isolate came from groundwater at the oak-ridge-field-research-center (ENIGMA SFA), where it exists in complex microbial communities, and tryptophan secretion by a prototroph that cannot re-assimilate it is described as a hallmark of cross-feeding potential — providing essential amino acids to auxotrophic community members — while the 231 fitness-important genes for tryptophan growth are read as complex regulatory and transport machinery rather than simple catabolism. ₈ Both the cross-feeding role and the regulatory interpretation are inferences from a within-organism discordance, not measurements of exchange. ₈

The surrounding support is literature context rather than measurement in this corpus. Ramoneda et al. (2023) mapped amino-acid auxotrophies across thousands of bacterial genomes and found tryptophan auxotrophy particularly common, from which an ecological niche for tryptophan-producing prototrophs like FW300-N2E3 is extrapolated rather than observed. ₈ Yousif et al. (2025) reported that obligate cross-feeding of metabolites is common in soil microbial communities with amino acids as a key exchanged currency, Fritts et al. (2021) reviewed extracellular metabolism as a driver of microbial cross-feeding, and Sibanyoni et al. (2025) documented amino-acid cross-feeding between Priestia megaterium and pseudomonas-fluorescens in the rhizosphere; none of these is a direct FW300-N2E3 measurement. ₈

Turning the candidate into a test is explicitly future work: the report proposes using the tryptophan overflow finding to parameterize a community metabolic model predicting cross-feeding interactions between pseudomonas-fw300-n2e3 and known tryptophan auxotrophs in the Oak Ridge groundwater community, so those predicted interactions are a plan rather than a result. ₈

Cofactor absence is not yet evidence of provisioning

A cautionary case comes from lanthanide methylotrophy, where community provisioning is only one of three live explanations for a missing pathway. Roughly 897 xoxf-carrying genomes have no detected pqq-biosynthesis, and the source proposes sequence-level resolution — per-genome ORF integrity plus CheckM2 completeness scoring — as the work needed to discriminate assembly fragmentation, pseudogenization, and genuine community-PQQ acquisition. ₉

This refines the caution running through this page from the opposite direction: where the metabolite correlations show that a present pathway does not prove exchange, these genomes show that an absent pathway does not prove dependence, because annotation and assembly artifacts leave the same genomic signature as scavenging a cofactor from neighbors — a negative-evidence problem treated in homology-search-negative-evidence and a distribution question treated in environmental-distribution-of-lanthanide-methylotrophy. ₉

Cross-feeding tests in a host-associated community

A paired-sample analysis in a human gut cohort tested metabolite exchange within a community and returned a largely negative result. In the hmp2 subject-level contrast of Crohn’s disease versus non-IBD subjects, none of 14 short-chain fatty acids (SCFAs — fermentation acids including acetate, propionate, butyrate, valerate, and hexanoate) reached differential abundance (DA, a test for differing concentrations between groups) at FDR < 0.10. ₁₀

The report treats this as an important null rather than a refutation: the absence could reflect HMP2 untargeted liquid-chromatography–mass-spectrometry (LC-MS) methods undersampling volatile, polar SCFAs (which typically need gas-chromatography–mass spectrometry), cohort-level subject-averaged contrasts masking within-subject variation, or gross SCFA pool differences between Crohn’s disease and non-IBD being smaller than within-subject variability — so the Anaerostipes caccae × pathobiont cross-feeding hypothesis cannot be confirmed by cohort-level butyrate differential abundance, and a paired within-sample test remains the right test. ₁₀ The cohort-level null therefore does not settle the within-sample cross-feeding causal claim; it sharpens the metabolic-coupling-cost annotation instead, since depleting A. caccae may not produce a measurable cohort-level butyrate change. ₁₀

The coupling that motivated the test was species-level co-occurrence inside co-occurrence modules, the intervention units discussed in species-cooccurrence-modules-as-intervention-units: 4–5 of 6 actionable Tier-A pathobionts co-cluster in a single Crohn’s-disease-specific module per ecotype, and the follow-up question was whether the non-pathobiont anchors of those modules show metabolic coupling with the pathobionts beside them. ₁₀ A. caccae, the only genuine butyrate producer among those anchors, does show strong positive species-level coupling with the pathobiont set, which the report reads as consistent with cross-feeding — an interpretation, not a demonstration. ₁₀

The test computed, for each anchor–pathobiont pair, within-IBD-substudy Spearman correlations across CMD_IBD samples, combined them by Fisher z-meta-analysis across the three robust substudies ZellerG_2014, NielsenHB_2014 and IjazUZ_2017, and scored sign concordance; an iron-context layer added a triple correlation (anchor × pathobiont × iron pathway) over 15 iron/heme pathways to test whether the iron theme was community-wide or pathobiont-specific. ₁₀

In the E1_CD stratum A. caccae correlated positively with all five module pathobionts, with sign concordance 1.0 across all three substudies: A. caccae × pathobiont Spearman ρ = +0.39 (enterocloster bolteae, the strongest pair), +0.33 (hungatella hathewayi), +0.31 (mediterraneibacter gnavus), +0.29 (flavonifractor plautii), and +0.08 (eggerthella lenta, the weakest), while B. nordii correlated negatively with M. gnavus (−0.21) and F. plautii (−0.20). ₁₀ The positive pattern was consistent with either butyrogenic cross-feeding or shared-environment co-response to the same Crohn’s-disease niche. ₁₀

The report reads the negative B. nordii couplings as niche competition for similar polysaccharide substrates — B. nordii being a generalist Bacteroides that competes with the same mucin- and glycan-degrading pathobionts, with Crohn’s disease selecting one over the other — but competition is an interpretation of correlation sign, not a measured interaction. ₁₀

The iron layer argues against an iron-mediated mechanism: A. caccae’s mean correlation over the 15 iron pathways was +0.13, much weaker than +0.45 for escherichia-coli, consistent with A. caccae being a fermentative butyrate producer rather than an iron-respiring organism. ₁₀ The substrate-, sugar- or lactate-mediated coupling the report proposed at that stage as the alternative mechanism was a hypothesis, and the later paired metabolite test did not support it, favouring shared-environment co-occurrence instead. ₁₀ The separation still matters for iron acquisition and pathobiont expansion: the A. caccae × pathobiont coupling appears in E1_CD while E. coli dominates iron-pathway co-variation, so the two signals are not the same signal. ₁₀

A. caccae is described as a well-characterized butyrate producer that grows on lactate and acetate co-substrates and on host-mucin-derived sugars, citing Schwiertz 2002, which describes Anaerostipes caccae as a saccharolytic, acetate-utilising, butyrate-producing bacterium from human faeces, and Duncan 2004, which reports human-faecal lactate-utilizing bacteria that produce butyrate as a major fermentation product. ₁₀ Its positive coupling with M. gnavus (a mucin-glucorhamnan producer), F. plautii (bile-acid 7α-dehydroxylating), H. hathewayi (a lactate and TMA producer) and E. bolteae (a sugar fermenter) is said to map onto known gut cross-feeding circuits, with pathobiont-released substrates feeding butyrogenic fermentation; the mapping is an interpretation, and the cross-feeding versus shared-environment disambiguation was explicitly deferred to the paired metabolite test. ₁₀

The coupling structure did not replicate in the other ecotype. In E3_CD only Lactonifactor × E. lenta was strong (ρ_meta = +0.27), while the two oral Actinomyces anchors coupled weakly (~ρ = 0.17–0.19) with module pathobionts; the report reads oral Actinomyces in the gut as co-trafficked ectopic colonizers under Crohn’s inflammation rather than metabolic partners, so their module membership reflects shared inflammation-driven colonization. ₁₀ This makes the cocktail-design implication E1-specific and possibly non-transferable to E3 patients, who are 31 % of the UC Davis cohort and may need a different module-anchor framework. ₁₀

The hypothesis verdict was recorded as partial: A. caccae × pathobiont coupling was clean in E1_CD with 4 strong-positive, all sign-concordant pairs, while E3_CD anchors lacked metabolic-coupling structure; the coupling by itself does not establish the proposed metabolic cross-feeding mechanism. ₁₀

A module observation cuts against a simple reading of butyrate producers as the healthy counterparty. Butyricicoccus pullicaecorum — an extensively studied butyrate-producing Clostridial-cluster-IV IBD-probiotic candidate — and A. caccae were both top-degree hubs in the E3 pathobiont module rather than in a separate healthy-commensal module, and the report notes the scope explicitly: these butyrate producers are Crohn’s-disease-depleted in most pooled analyses yet co-vary with the pathobionts under within-ecotype co-occurrence. ₁₀ The report calls that pattern consistent with a metabolic-partner or cross-feeding interpretation and a caution against treating “preserve butyrate-producers” as a naive phage-targeting goal, since in the Crohn’s ecological context these species may track with the pathobionts rather than against them. ₁₀

A pathway-level reading of H. hathewayi adds a second interpretive rather than demonstrated interaction: under Crohn’s disease the organism shifts toward pentose phosphate, anabolic amino-acid biosynthesis and purine salvage and away from lactose, galactose, propanediol and inositol utilization, and the propanediol-degradation decrease is read as niche partitioning in which E. coli may dominate that niche under Crohn’s disease. ₁₀

The design consequence drawn from the species-level coupling is a hypothesis, not a measured treatment effect: a phage cocktail depleting M. gnavus, F. plautii, H. hathewayi and E. bolteae might incidentally reduce A. caccae through loss of substrate, and because A. caccae is the only genuine butyrate producer among the module anchors and butyrate is anti-inflammatory, that incidental depletion could partially offset therapeutic benefit — motivating a metabolic-coupling-cost annotation on each actionable target asking which beneficial commensals depend on it and what the predicted net inflammatory balance is. ₁₀ The report files this hypothetical ecological cost under the ecological-durability tier of its criteria rubric, the same cost weighed in ecological-cost-of-microbiome-target-depletion and competitive-exclusion-consortium-design. ₁₀

Paired-sample disambiguation rejected cross-feeding and reframed the coupling as shared-environment co-occurrence in healthy/normobiotic samples. ₁₀ Only 7 strict cross-feeding triangles emerged, and lactate × A. caccae (+0.18) had the opposite sign of lactate × F. plautii (-0.23) and lactate × E. bolteae (-0.20), inconsistent with a lactate-mediated cross-feeding loop; shared-environment co-occurrence is the report’s preferred interpretation, with the cocktail-design implication that phage targeting of pathobionts is unlikely to substantially deplete A. caccae via metabolic coupling — an inference from correlation structure rather than a tested depletion. ₁₀ The metabolic-coupling-cost annotation had rested on species-level co-occurrence, and the paired sample-level test identified no metabolic intermediate supporting cross-feeding, so bile-acid coupling cost replaced it as the primary annotation. ₁₀

The cross-feeding-triangle test applied strict criteria (anchor × pathobiont correlations of the same sign, both |ρ| > 0.20, both FDR < 0.10) and returned only 7 triangles across the 468 paired CSM* HMP2 samples × 8 species × 583 named metabolites examined, with top candidates caffeine (B. nordii / F. plautii), linoleoylethanolamide (B. nordii / eggerthella lenta; A. caccae / E. lenta), urobilin (A. caccae / E. lenta; B. nordii / E. lenta), and cholate (A. caccae / F. plautii; B. nordii / F. plautii). ₁₀ Across those 468 paired CSM* samples, spanning 8 species and 583 metabolites, cross-feeding was not supported: only 7 strict triangles, and lactate associations of opposite sign between A. caccae and the pathobionts. ₁₀

None of the 7 candidates is a butyrogenic cross-feeding intermediate; the pattern instead matches shared-niche, health-direction co-occurrence, with A. caccae, B. nordii, F. plautii, and E. lenta correlating negatively with the primary bile acid cholate and positively with urobilin, i.e. co-occurring in healthy/normobiotic rather than Crohn’s-dysbiotic samples. ₁₀

The biochemistry behind those non-butyrogenic candidates is cited literature rather than measurement here: Hall et al. (2024) identified a gut-bacterial bilirubin reductase that produces urobilinogen and connects the intestinal microbiota to host bilirubin metabolism, Vital et al. (2018) reported a diversity of bilirubin-reducing bacteria in the human intestinal microbiota that the report calls urobilinoid-producing commensals, and Selma et al. (2014) described Gordonibacter urolithinfaciens as a urolithin-producing bacterium isolated from the human gut. ₁₀

The cohort frame is Franzosa et al. (2019), whose HMP2 metabolomics and microbiome integration the report links to a canonical bile-acid 7α-dehydroxylation deficit together with polyamine and lipid signatures in inflammatory bowel disease — the kind of paired-layer inference discussed in multi-omics-integration. ₁₀

A curated metabolite panel gives the species-level correlations behind that reading, reported in the column order A. caccae, B. nordii, H. hathewayi, F. plautii, E. bolteae, E. lenta, M. gnavus, escherichia-coli, with an asterisk marking the values the report flags as significant and unmarked values nonsignificant: butyrate +0.10*, +0.13*, -0.08, -0.02, -0.04, -0.05, -0.02, +0.04; lactate +0.18*, +0.10*, +0.03, -0.23*, -0.20*, -0.07, +0.10, +0.24*; choline +0.04, +0.01, -0.07, -0.16*, -0.15*, -0.13*, +0.11*, +0.25*; carnitine -0.04, -0.02, -0.01, -0.18*, -0.06, -0.04, +0.04, +0.19*; cadaverine, listed only for E. coli as its specialty, +0.45*; putrescine -0.04, -0.09, -0.07, -0.02, -0.15*, -0.11*, +0.13*, +0.19*; and tryptophan -0.09, -0.10*, -0.10, -0.13*, -0.10, -0.22*, +0.14*, +0.25*. ₁₀

The butyrate association with A. caccae (+0.10) is significant but very weak — consistent with A. caccae as a population-level butyrate producer, yet limited both methodologically, because HMP2 LC-MS undersamples SCFAs with only butyrate and propionate detected, and biologically, because the gut butyrate pool is dominated by other producers such as Faecalibacterium and Roseburia. ₁₀

The lactate correlations run in opposite directions — +0.18 for A. caccae against -0.23 for F. plautii and -0.20 for E. bolteae — which the report calls the strongest evidence against the cross-feeding hypothesis, since a production–consumption relationship would predict same-sign coupling; these species instead appear to occupy different metabolic niches at cohort level. ₁₀

E. coli dominates the cohort-level correlation signal in this panel (choline +0.25, carnitine +0.19, tryptophan +0.25, putrescine +0.19, and cadaverine +0.45, the strongest correlation in the entire panel), so cohort-level metabolite associations here are weighted toward one taxon rather than distributed across the proposed exchange partners. ₁₀

A parallel bile-acid panel reports paired-sample species × metabolite correlations across five bile-acid classes in the same column order, again with an asterisk for the values flagged significant: tauro-α/β-muricholate (primary tauro-conjugated) -0.03, -0.15*, 0.00, -0.04, -0.04, -0.11*, +0.20*, +0.06; cholate (primary unconjugated) -0.13*, -0.18*, -0.10, -0.26*, -0.10, -0.13*, +0.11*, -0.02; deoxycholate (secondary from cholate) -0.07, +0.05, -0.12, +0.06, +0.17*, +0.03, -0.10*, -0.05; lithocholate (secondary from CDCA) -0.05, +0.06, -0.12, +0.15*, +0.18*, +0.07, -0.20*, -0.13*; and ketodeoxycholate (secondary oxidized) -0.12*, -0.15*, -0.01, -0.19*, -0.05, -0.14*, +0.14*, +0.24*. ₁₀ The report singles out the paired sample-level substrate–product subset ρ(F. plautii × cholate) = −0.26 and × lithocholate = +0.15, ρ(E. bolteae × deoxycholate) = +0.17 and × lithocholate = +0.18, and ρ(E. lenta × ketodeoxycholate) = −0.14. ₁₀

F. plautii, E. lenta, and E. bolteae — the canonical bile-acid-7alpha-dehydroxylation bacteria — show the predicted substrate–product signature, correlating negatively with primary tauro-conjugated bile acids and positively with secondary unconjugated products (deoxycholate, lithocholate), which the report calls direct sample-level confirmation that the 7α-dehydroxylation network operates in HMP2 samples; the evidence is correlational rather than interventional, the distinction drawn in occurrence-versus-catabolic-activity. ₁₀

The network was identified in a 468 paired CSM* HMP2 samples × 8 species × 583 metabolites Spearman ρ matrix: F. plautii, E. lenta and E. bolteae carry the predicted primary-tauro-bile-acid-down and secondary-bile-acid-up substrate–product signature, M. gnavus and E. coli carry the opposite pattern, and E. coli dominates the cohort-level correlations (cadaverine +0.45, choline +0.25, tryptophan +0.25). ₁₀

A separate cohort-level contrast reads the same chemistry at the disease level: in the HMP2 Crohn’s-disease-versus-non-IBD subject-level metabolomics comparison, the report states that tauro-muricholate confirms an F. plautii 7α-dehydroxylation deficit — a disease-associated shortfall in the pathway’s output, not a measurement of F. plautii’s own enzymatic capacity. ₁₀

The enzymatic basis invoked is literature rather than measurement in this dataset: Devlin and Fischbach (2015) described the bai-operon mechanism of bile-acid 7α-dehydroxylation in clostridium-scindens-like bacteria, which the report treats as the anchor for its bile-acid network finding, and core bai-operon genes are the named genes proposed to encode the activity across F. plautii strains. ₁₀

By contrast, M. gnavus and E. coli show the opposite pattern — positive with primary tauro-conjugated bile acids and negative with secondary bile acids — and are not assigned to the 7α-dehydroxylation network in this dataset. ₁₀

Taken together, the paired sample-level evidence does not support the cross-feeding hypothesis, and shared environment is the more parsimonious explanation for the A. caccae × pathobiont coupling; the downstream statement that phage targeting of pathobionts is unlikely to substantially deplete A. caccae through metabolic coupling is an inference from correlation structure, not a tested depletion. ₁₀ This contradicts a Black Queen-style reading of positive co-occurrence as evidence of provisioning and refines the caution running through this page: co-occurrence plus plausible biochemistry is not exchange. ₁₀

Where a substrate–product signature does hold, it is converted into a coupling cost for intervention design, relevant to phage-therapy-evidence-translation: F. plautii is profiled as an active 7α-dehydroxylator with the highest bile-acid coupling cost (ρ × cholate = −0.26, × lithocholate = +0.15), whose depletion is predicted to shift the bile-acid pool toward primary tauro-conjugated inflammatory forms, prompting the suggestion of co-administering UDCA or a bile-acid-binding agent. ₁₀

That cost is the dominant design constraint for the E1 ecotype: F. plautii is present in 78 % of patients (18/23) and in all 9 E1 patients while carrying the highest bile-acid coupling cost as an active 7α-dehydroxylator, so the report gives it the lowest targeting priority and recommends deprioritization in favour of co-administering UDCA or a bile-acid-binding agent. ₁₀ Any earlier Tier-2 phage-target label for F. plautii is superseded by the report’s own phage-feasibility table, which classifies it as a phage GAP and prefers alternatives. ₁₀

A stated design principle goes further, treating F. plautii as a “leave alone” target despite a score of 3.3, on a triple penalty of 78 % prevalence, highest bile-acid coupling cost, and membership in the phage gap; the proposed substitute is bile-acid-binding co-therapy (UDCA or cholestyramine) to maintain the secondary bile-acid pool that depleting F. plautii would erode. ₁₀

The phage-gap half of that penalty is a null result about available phages, not about the bacterium or the literature on it: no phage against F. plautii appears in the literature curation or in the phagefoundry collection, and no detectable in-vivo phage correlation was found, so alternatives are preferred. ₁₀ Where targeting does proceed it is tiered and monitored — E. lenta (PMBT5) and E. bolteae (PMBT24) sit in a Tier-2 “phage targeting with monitoring” class with bile-acid pool monitoring, and M. gnavus is considered possible only if lytic-locked phage engineering succeeds. ₁₀

At cohort scale the design summary reports concrete cocktails for 14 of 23 patients (61 %), a three-strategy hybrid requirement for E1 because a pure phage cocktail is not feasible there, the F. plautii bile-acid constraint as the dominant E1 design constraint, and E. coli present in only 35 % of the cohort. ₁₀ The bile-acid coupling cost is currently estimated at ecotype level rather than per patient, so clinical translation would require per-patient bile-acid measurements. ₁₀

E. bolteae is profiled as a Tier-2 phage target with moderate bile-acid coupling cost from its positive deoxycholate (+0.17) and lithocholate (+0.18) associations; E. lenta receives a moderate annotation from a partial 7α-dehydroxylation pattern (cholate −0.13, ketodeoxycholate −0.14) consistent with Eggerthellaceae bile-acid metabolism; and H. hathewayi, M. gnavus, and E. coli are assigned low bile-acid coupling cost because they are not in the 7α-dehydroxylation network in this dataset. ₁₀

The two evidence streams differ in robustness. The paired sample-level substrate–product signature (ρ(F. plautii × cholate) = −0.26; × lithocholate = +0.15) is a species × metabolite correlation computed within paired metabolomics plus metagenomics samples, not a disease-versus-control aggregate; cohort-aggregate cross-cohort differential abundance on bile acids replicates at 57% for primary and 40% for secondary bile acids — moderately above chance for primary and sub-chance for secondary — making the aggregate evidence weaker than the paired evidence. ₁₀ For cross-cohort-microbiome-portability this matters directly: a metabolite claim that travels across cohorts is not the same claim as one measured within paired samples. ₁₀

Co-occurrence module structure adds organism-specific texture without establishing exchange: F. plautii and E. coli show ecotype-specific co-occurrence-module membership, and Butyricicoccus pullicaecorum, Anaerostipes caccae, and Lactococcus lactis in ecotype E3 are tentatively proposed as module-anchor commensals that may provide functional-driver context — a proposal, not an established driver role. ₁₀

Two figures carry this evidence: NB06_cooccurrence_networks.png shows per-subnet spring-layout co-occurrence networks (E1_all, E1_CD, E3_all, E3_CD) with actionable Tier-A species highlighted, and NB09c_cross_feeding_disambiguation.png is a two-panel figure pairing a curated 7-theme cross-feeding panel × 8 species ρ heatmap that shows the bile-acid 7α-dehydroxylation network with a scatter of the top 25 cross-feeding-triangle candidates, of which only 7 are strict triangles. ₁₀

Open Directions

• Reanalyze the NMDC metabolomics samples with study-stratified correlations and leave-one-study-out validation to test whether the 11/13 directional pattern persists after reducing the influence of nmdc:sty-11-r2h77870, which contributed 125/131 samples (95%). ₁
• Replace substring metabolite matching with identifier-validated KEGG or equivalent mappings, then retest leucine, isoleucine, and the previously untestable cysteine, histidine, and lysine pathways to determine whether the Black Queen pattern is mapping-sensitive. ₁
• Combine sample-level metabolite abundances with genome-resolved pathway presence and condition-specific fitness measurements to test whether negative metabolite associations identify producers, consumers, or both. ₁
• Test the tyrosine outlier using independent cohorts and targeted metabolomics to distinguish a real anti-Black-Queen relationship from cohort, annotation, or measurement effects. ₁
• Compare pathway presence, pathway completeness, and measured metabolite pools across the 7 Fitness Browser organisms and 23 GapMind pathways to ask when Latent Capabilities become realized community dependencies. ₁
• Add pH, temperature, and total organic carbon to partial-correlation or mixed-effects models, and pair metabolomics with metatranscriptomics, to separate abiotic confounding and genomic potential from expressed biosynthetic activity. ₄
• Obtain metabolomics for the 33 Freshwater samples and resolve the 61 Unknown-ecosystem samples through NMDC environmental annotations to test whether the Black Queen pattern generalizes beyond the predominantly soil H1 cohort; the freshwater and soil pathway-completeness profiles are described as strikingly different, which is what makes the aquatic-versus-terrestrial comparison the direct test. ₄
• Replicate the H1 amino-acid analysis across additional NMDC studies with paired taxonomy and metabolomics, since the current H1 dataset is effectively single-study, and include pH, temperature and total organic carbon in a partial-correlation or mixed-effects model once NMDC abiotic data is populated for the metabolomics-overlap samples, to test whether the leucine and arginine signals generalise beyond the current cohort or are confounded. ₄
• Reconcile the 7-organism condition-specific result with the 48-organism threshold analysis by applying matched pathway definitions and condition-stratified fitness models, testing whether latent capabilities remain neutral or become important under community-relevant media. ₃
• Link clade-level pangenome openness to genome-resolved metabolite production and consumption, testing whether the ρ = 0.69 association reflects actual exchange networks rather than shared phylogeny or sampling structure. ₃
• Replace pathway-level conservation with per-gene dN/dS or gene presence/absence across a phylogeny, directly testing the Black Queen prediction of progressive gene loss that the pathway-completeness measure cannot address. ₃
• Identify lineages with recent gene-loss events (singletons or auxiliary genes in the pangenome) in pathways classified as latent and verify that these represent actual deletion events in the phylogeny, closing the gap between a latent classification and a demonstrated loss. ₃
• Test whether organisms with higher latent fractions are more likely to encode public-goods pathways or to co-occur with cross-feeding partners in community datasets, turning the clade-level openness association into a testable community-composition prediction. ₃
• Validate the pathway-openness and ecotype-openness associations with phylogenetically independent models and genome-resolved metabolite production or consumption, testing whether accessory-dependent amino-acid pathways correspond to realized exchange. ₅
• Re-run the plant-associated complementarity permutation at a finer metabolic resolution than the 80 GapMind pathway categories, and revisit the core-level completeness threshold that scored 0% across all compartments, to test whether the small redundancy effect survives once the stated resolution caveat is removed. ₆
• Extend annotation-gap reaction detection from single genomes to microbial communities — proposed in the source as a possible application rather than reported as a finding — to test whether annotation gap reactions mediate identifiable cross-feeding interactions. ₁₁
• Measure short-chain fatty acids by gas-chromatography–mass spectrometry in paired gut metagenomic and metabolomic samples to test only the measurement question the cohort-level null leaves open — whether LC-MS undersampling of volatile, polar compounds or small pool differences explains the absence of SCFA differential abundance — since the source states that this cohort-level null cannot confirm or exclude the cross-feeding causal claim, for which the within-sample paired test remains the right design. ₁₀
• Parameterize a community metabolic model from the FW300-N2E3 tryptophan overflow result and test its predicted exchanges against measured tryptophan pools and auxotroph abundances in the Oak Ridge groundwater community, closing the gap between an inferred secretion phenotype and demonstrated exchange. ₈
• Score per-genome ORF integrity and completeness (CheckM2) for the ~897 xoxF genomes with no detected PQQ, separating assembly fragmentation and pseudogenization from genuine community PQQ acquisition before any such genome is counted as evidence of cofactor cross-feeding. ₉
• Measure bile acids per patient rather than per ecotype in the gut cohort, converting the ecotype-level bile-acid coupling cost into the per-patient estimate that cocktail design would need. ₁₀
• Re-test the E1-specific anchor-coupling result in E3 patients with an E3-appropriate module-anchor framework, to establish whether the absent metabolic-coupling structure there is real or a power limitation. ₁₀