Taxonomic Resolution Controls Detectable Functional Associations

Taxonomic resolution controls which genotype-to-function relationships can be detected because aggregation determines both how community taxa are bridged to pangenome clades and how much abundance is retained for functional scoring. ₁ Mapped coverage—the fraction of community abundance linked to the functional-feature construction—can therefore make community-level associations reflect representation rather than biological shifts. ₁ Functional redundancy, in which different taxa provide overlapping functions, is a compatible hypothesis: environmental selection may change community membership while aggregated functional summaries remain stable, but this dataset does not directly demonstrate redundancy. ₁

The Functional Dark Matter — Experimentally Prioritized Novel Genetic Systems supports this coverage-and-resolution interpretation from a complementary genome-centered analysis. Across 48 Fitness Browser organisms, 57,011 of 228,709 genes (24.9%) were classified as dark, but dark-gene fractions ranged from more than 35% to less than 15%, a pattern interpreted as annotation-depth variation rather than established differences in functional content. ₂ Full GTDB r214 pangenome propagation increased dark-gene conservation coverage from 32,791 (57.5%) to 37,997 (66.6%) by recovering 5,206 additional genes, showing that detectable functional inference changes substantially with database and taxonomic coverage. ₂ The report therefore refines the existing claim: taxonomic resolution is not only an aggregation choice, but also a function of annotation depth, ortholog-group breadth, and which organisms are represented in the reference collection. ₂

The ENIGMA workflow linked enigma-coral geochemistry, community composition, kbase-ke-pangenome clades, and eggnog-derived functional proxies across 108 samples. ₁ The geochemistry matrix had shape (108, 49), the community data contained 41,711 taxon rows, and the analysis covered 212 distinct communities and 1,392 distinct genera. ₁ The analysis is summarized in Contamination Gradient vs Functional Potential in ENIGMA Communities, while the related plant-association analysis is summarized in Plant Microbiome Ecotypes — Compartment-Specific Functional Guilds and Their Genetic Architecture. _{1 3}

Resolution, Coverage, and Bridge Constraints

The taxonomy bridge parsed 27,690 GTDB species rows, mapped 530 genera, left 862 genera unmapped, and produced 530 strict clades, 7,380 relaxed clades, and 150 species-proxy clades restricted to genera mapping to exactly one clade. ₁ It generated 8,242 genus-to-clade rows through many-to-many expansion. ₁ Ambiguity was substantial: 380 mapped genera were multi-clade genera, and the clade-count distribution had a long right tail with a maximum of 433 clades per genus. ₁ The most ambiguous genera were pseudomonas with 433 clades, streptomyces with 378, prevotella with 358, streptococcus with 214, and mycobacterium with 186. ₁

Of the 1,392 ENIGMA genera observed, the bridge mapped 530 and left 862 unmapped; 150 mapped genera were unique-clade (species-proxy resolvable) and 380 were multi-clade ambiguous, and the 8,242 bridge rows supported 3,630 functional feature rows across the strict_single_clade, relaxed_all_clades, and species_proxy_unique_genus modes. ₁ Strict and relaxed modes retained a mean mapped abundance fraction of 0.343, whereas the species-proxy mode retained a mean of 0.031. ₁ Across functional-feature rows, strict and relaxed modes had a mapped-abundance range of 0.031 to 0.854, while the species-proxy mode had a range of 0.0002 to 0.543. ₁ The species-proxy mode retained 150 unique-clade genera compared with 530 mapped genera overall, and the analysis generated 3,630 functional feature rows across mapping modes. ₁ Thus, the more restrictive unique-clade mode reduced taxonomic ambiguity but also discarded much of the abundance available for functional inference. ₁

The functional-dark-matter analysis supports this trade-off between specificity and coverage. Its environmental analyses tested 151 accessory dark-gene clusters across 31 species, but only 10 of 137 clusters with environmental-category tests were significant at FDR < 0.05 and only 1 of 67 showed significant AlphaEarth embedding divergence. ₂ A pre-registered lab–field mapping found 29 of 47 testable clusters concordant (61.7%), while NMDC, the National Microbiome Data Collaborative dataset, independently matched 5 of 6 carrier genera and confirmed 4 of 4 testable abiotic predictions. ₂ These results show that finer gene- and carrier-level resolution can recover environmentally structured signals, but sparse metadata and incomplete mapping constrain how broadly they can be detected. ₂

The source taxonomy table, enigma_coral.ddt_brick0000454, provided labels through Genus but no species or strain labels, preventing direct species-level bridge testing in this dataset slice; the species_proxy_unique_genus mode was added to approximate higher resolution by keeping only genera mapping to exactly one GTDB species clade. ₁ Genus-level aggregation is therefore necessary but potentially masking: strain-level adaptation may be hidden when distinct functional states are combined within a genus. ₁ The bridge ambiguities can redistribute functional features across clades and weaken genus-level resolution of species- or strain-specific environmental responses. ₁ The ENIGMA Carbon Census — Tiered Knowledge Census of 83 Enrichment Compounds supports this limit independently: it describes the pangenome-to-ENIGMA taxonomic bridge as genus-level at best and avoids isolate-specific transfer of pangenome Tier-2 reconstructions in favor of direct genome_depot annotation. ₄

The pitfalls digest generalizes this bridge limit. ENIGMA taxonomy in ddt_brick0000454 currently includes Domain through Genus levels only, so attempting species-level bridge logic directly will either fail or silently reuse genus-level mappings while appearing higher resolution. ₅ The digest’s sanctioned workaround is the one this project took: use an explicit species-proxy mode (unique genus to a single GTDB clade) and report the mapped-coverage loss it incurs. ₅ This supports the page’s reading of the species-proxy coverage collapse as a structural consequence of the available taxonomy rather than an artifact of one analysis. _5,1

Resolution mismatch also strikes on the reference side of a join, where it can manufacture a null. Both the NMDC covstats_taxonomy_rollup and Planet Microbe run_to_taxonomy abundance tables key on a taxonomy.name that is species-level, not genus-level. ₅ In enigma_carbon_census_1 an initial marine-abundance contrast showed 0/68 genera present in Planet Microbe — not because they were absent, but because the genus-name filter matched only genus-rank reference rows instead of aggregating their constituent species. ₅ The remedy is an ordering rule: parse or derive the genus from the species name, GROUP BY genus and SUM(abundance), and only then filter to the genera of interest. ₅ This refines the page’s coverage argument down to query construction: an apparent absence can be produced by the rank at which a table is keyed. ₅

The plant-association analysis supports this masking concern but also refines its expected magnitude. Across 607 root, rhizosphere, and phyllosphere species, the refined marker panel detected compartment separation with PERMANOVA R² = 0.071, while location-only db-RDA R² = 0.060; 84% of the PERMANOVA signal was attributable to location and 16% to dispersion. ₃ The earlier R² = 0.527 fell to 0.072 after excluding three genome-rich species per compartment, an 86% loss, indicating that apparent broad functional separation can be dominated by taxonomic sampling. ₃ This supports the existing concern that aggregation can conceal or manufacture associations, while showing that even refined compartment effects may be small rather than universally strong. ₃

The Ecotype Functional Differentiation supports this masking concern from a complementary within-species analysis: valid gene-content ecotypes were detected in 12 of 15 sampled species (80%), and 170 of 257 COG-category tests (66.1%) were significant after BH-FDR correction across those species. ₆ This strengthens the case that finer within-species structure can carry detectable functional differences, while not showing that those differences are ecological adaptations. ₆

The plant analysis further refines the inference by finding that marker associations were often genus-scale: after genus-cluster-robust modeling, 8/14 markers passed the stated significance and effect-size thresholds, but within-genus label shuffling retained only nitrogen fixation, ACC deaminase, and T3SS. ₃ Its phylogenetic-tree coverage was also limited: only 18 of 65 species with at least 20 genomes had tree distance data, and only 5 of 17 testable species passed Bonferroni-corrected Fisher tests. ₃ Thus, finer resolution can reveal structure, but the detectable signal and its ecological interpretation remain constrained by within-genus controls and coverage. ₃

The cross-species COG Functional Category Analysis refines the resolution argument. Across 32 species spanning 9 phyla and 357,623 genes, novel or singleton genes were enriched in mobile-element COG L by +10.88%, defense COG V by +2.83%, and unknown-function COG S by +1.64%, while the report lists translation COG J (-4.65%, 97% consistency) and several metabolic categories under core genes being consistently enriched. ₇ The report attaches negative values to these core-enriched categories and labels J the strongest depletion, so the sign convention of these figures is ambiguous; the report’s stated direction is core enrichment of translation and metabolic functions. ₇ COG, or Clusters of Orthologous Groups, categories therefore retain systematic core-versus-novel functional structure, but genus-level or broad-category aggregation can conceal structured differences in gene novelty and ecological function even when the signal is consistent across taxa. ₇

Functional-dark-matter conservation analysis supports the importance of choosing an informative taxonomic scale but refines the meaning of conservation. An initial eggNOG breadth classification assigned 30,721 of 30,756 dark-gene clusters (99.9%) to universal breadth and was poorly discriminative; species counts ranged from 1 to 33, with median 1 and mean 2.2, and the species-count variant had Spearman ρ = 0.982 with the original ranking but only 62% top-50 overlap and 58% top-100 overlap. ₂ After GTDB r214 propagation, root ortholog groups were distributed across kingdom-, class-, family-, genus-, mobile-, phylum-, order-, and species-level tiers, with 55.9% classified as kingdom-level and 0.5% as species-level. ₂ A broad taxonomic label can therefore indicate conservation without identifying the resolution at which ecological function is expressed. ₂

Marker-gene surveys impose a resolution ceiling before any functional mapping begins. In the Lab Fitness Predicts Field Ecology at Oak Ridge, 16S amplicon data resolves to genus, but Fitness Browser organisms are specific strains, so genus-level community profiles cannot be matched directly to the strains whose fitness was measured. ₈ In the Lignin Enrichment and Ecological Memory in Microbial Communities, 97% OTU clustering — operational taxonomic units grouped at 97% identity, computed with vsearch — was used instead of ASV-level resolution (amplicon sequence variants, resolved with DADA2) because of environment constraints, which may merge closely related species. ₉ Both support the ENIGMA position that attainable resolution is set by the assay and the reference tables rather than chosen freely. _8,9,1

The Metal Resistance Ecology — Phylogenetic Conservation vs. Environmental Selection refines the masking concern by testing it rather than assuming it. Genus-level pruning — one representative tip per genus, chosen as the GTDB species cluster representative — discards within-genus phylogenetic structure, and its sensitivity analysis S3 tested whether within-genus SD of metal types is a significant covariate in PGLS (phylogenetic generalized least squares, regression that accounts for shared ancestry); it is not, p = 0.18, which the report reads as confirming that genus aggregation does not hide a species-level confound. ₁₀ That S3 model retained a positive BH-FDR-significant coefficient, with 606 genera, +0.0189, SE 0.0060, p = 0.0016 and q = 0.007. ₁₀ The same report nevertheless preserves the proxy caveat: genus-level aggregation of OTU niche scores is a proxy, because a genus may appear broad-niched because different species within it occupy different habitats rather than any single organism being a true ecological generalist, so the PGLS coefficient reflects covariation between inferred niche breadth and inferred metal tolerance capacity, not confirmed physiology. ₁₀

Resolution also governs what a failed prediction would mean. The MicrobeAtlas report supplies falsification criteria for its predicted OTU enrichment: if the OTU does not increase, the candidate explanations are that the resistance genes are on accessory elements absent in the soil ecotype, or that the metal concentration exceeds the allele-specific MIC (minimum inhibitory concentration). ₁₀ Both candidates sit below the genus label at which the association was estimated, so a negative field test would not discriminate between them without finer data. ₁₀

Taxonomy-inferred functional scores make the genus assumption explicit and quantifiable. The Polyhydroxybutyrate Granule Formation Pathways — Distribution Across Clades and Environmental Selection used a two-tier taxonomy mapping (Tier 1: gtdb_metadata NCBI taxid to GTDB genus bridge, 2,336 taxon columns; Tier 2: direct genus name matching via taxonomy_dim, 678 additional columns) to map 3,014/3,492 (86.3%) NMDC taxon columns to GTDB genera with known PHB (polyhydroxybutyrate) status, then computed per-sample inference scores as the abundance-weighted sum of genus-level phaC prevalence across 6,365 NMDC metagenomic samples, with a median 87.2% of taxonomic abundance matched to pangenome genera. ₁₁ The Prophage Gene Modules and Terminase-Defined Lineages Across Bacterial Phylogeny and Environmental Gradients reused the device: taxonomy-based inference of prophage burden across 6,365 NMDC metagenomic samples at 87.2% median matching coverage reveals 57 significant module-abiotic correlations at FDR < 0.05. ₁₂

Those prophage scores were not measured in the samples. Genus-level prophage burden scores derived from the pangenome were applied to the NMDC samples via two-tier taxonomy mapping (GTDB metadata bridge plus taxonomy_dim genus matching), so burden is inferred rather than observed. ₁₂ The report states that this inference assumes genus-level conservation of prophage content, which may not hold for recently acquired or lost prophages, and that the approach was validated for the PHB granule ecology project but has not been independently validated for prophage genes. ₁₂ This refines the page’s coverage argument into an assumption about rank: a high matched-abundance fraction says nothing about whether the trait being propagated is conserved at the rank used to propagate it. _12,11

Coverage-Sensitive Contamination–Defense Associations

The contamination index combined arsenic, cadmium, chromium, copper, lead, nickel, uranium, and zinc using per-metal log1p z-scoring followed by a row-wise mean. ₁ Across the 108 samples, it ranged from -0.448 to 3.836, with median -0.271 and IQR [-0.363, 0.053]. ₁

Predeclared genus-level Spearman tests of site defense score against contamination were null in both broad mapping modes. ₁ In the relaxed-all-clades mode, rho was 0.0587, the 95% bootstrap confidence interval was [-0.128, 0.250], Spearman p was 0.546, and FDR q was 0.862. ₁ In the strict-single-clade mode, rho was 0.0682, the 95% bootstrap confidence interval was [-0.111, 0.253], Spearman p was 0.483, and FDR q was 0.849. ₁ FDR, or false-discovery-rate correction, controls the expected proportion of false discoveries among results called significant. ₁

The confirmatory null was unchanged across four contamination definitions: the composite all-metals index, uranium-only, the top-3-variance-metals index, and the first principal component of metal z-scores. ₁ All 8 confirmatory variant tests remained non-significant after FDR, with q = 0.546 across the tests, including uranium-only. ₁

Coverage-aware modeling refines the confirmatory null by showing that detectable signal depends on both mapped abundance and mapping mode. In adjusted ordinary least-squares models, adjusting for contamination, depth, latitude, longitude, and mapped abundance fraction, the relaxed-all-clades defense-score beta was 0.000751 with 95% bootstrap confidence interval [0.000224, 0.001779], p = 0.000398, and FDR q = 0.0462. ₁ The corresponding strict-single-clade estimate was beta = 0.000640 with 95% bootstrap confidence interval [0.000169, 0.001538], p = 0.00354, and FDR q = 0.130. ₁ These results support a coverage-sensitive defense association in the relaxed mode, but do not establish a robust association across both mapping modes after global FDR. ₁

The functional-dark-matter report supports treating coverage as an inferential constraint rather than as a neutral preprocessing detail. Its Fitness Browser collection contained 37/48 Pseudomonadota organisms, and the top 500 candidates included 417 Gammaproteobacteria, 71 Alphaproteobacteria, 17 Deltaproteobacteria, 12 Bacteroidetes, 10 Betaproteobacteria, 5 Firmicutes, 2 Cyanobacteria, and 0 Archaea, Actinobacteria, or Epsilonproteobacteria. ₂ The report explicitly notes that uneven condition coverage and taxonomic composition can make organisms with deeper profiles appear more informative. ₂ This refines interpretation of the ENIGMA mode difference: apparent resolution effects can combine retained abundance, taxonomic representation, annotation depth, and experimental coverage.

Fraction-aware adjusted models showed the same directional pattern without resolving mode dependence. ₁ The relaxed-all-clades beta was 0.000607 with 95% bootstrap confidence interval [0.000213, 0.001221], p = 0.00144, and FDR q = 0.0838, while the strict-single-clade beta was 0.000566 with 95% bootstrap confidence interval [0.000214, 0.001113], p = 0.00548, and FDR q = 0.130. ₁

The species-proxy mode further weakened detectable association: of 530 mapped genera it retained only the 150 unique-clade genera, excluding the 380 multi-clade genera, its mean mapped-abundance fraction was 0.031 versus 0.343 in strict/relaxed modes, and its defense trend was positive but non-significant, with rho = 0.169 and Spearman p = 0.081. ₁ No high-coverage test was feasible at mapped-abundance fraction >= 0.25 because of low retained coverage, and the species-proxy high-coverage analysis was underpowered at n = 1, whereas coverage-aware fraction-level models increased effective rows. ₁ The discoveries digest records the same genus-only taxonomy and species-proxy coverage collapse and concludes that true species-level gains likely require higher-resolution ENIGMA taxonomy or metagenomic mapping. ₁₃ This is sensitivity evidence that increased clade specificity can reduce power when it sharply lowers coverage, rather than evidence against contamination-linked functional differentiation. ₁

Robustness and Functional-Redundancy Interpretation

Fraction-aware analyses used 212 sample-fraction rows, consisting of 106 rows for the 0.2_micron_filter fraction and 106 rows for the 10_micron_filter fraction. ₁ Within-fraction defense tests were non-significant in relaxed mode, with p = 0.767 for the 0.2-micron fraction and p = 0.898 for the 10_micron fraction, and in strict mode, with p = 0.780 and p = 0.793, respectively. ₁ This supports the interpretation that the aggregate monotonic defense signal is not robustly reproducible within individual fraction strata. ₁

High-coverage subsets produced nominal defense p-values of 0.0207 for relaxed mode and 0.00980 for strict mode, but their global FDR q-values were 0.301 and 0.189, respectively. ₁ Most non-defense outcomes remained non-significant, although site stress score in the strict high-coverage subset had rho = 0.2489 and p = 0.0407. ₁ These results refine the interpretation of taxonomic-resolution effects: changing the mapping mode changes both the functional score’s coverage and the evidential threshold reached by downstream models. ₁ The discoveries digest supports this directly: when strict and relaxed modes were computed independently rather than by reusing strict outputs, 1,140 of 1,590 genus-feature pairs differed between modes (~71.7%) and site stress scores differed for all 108 samples, so strict-versus-relaxed sensitivity should not be approximated by copied feature tables. ₁₃

The plant analysis supports the hypothesis that aggregation can conceal environmental differentiation: its 11.7% Jaccard overlap between 487 pangenome-derived plant genera and 438 MGnify rhizosphere genera reflects different isolation-metadata and metagenomic-detection processes, while compartment labels were available for only 7,995 of 293,059 genomes (2.7%). ₃ Its corrected complementarity analysis found co-occurring genus pairs slightly less complementary than random pairs, with Cohen’s d approximately -0.4 and permutation p < 0.001, rather than demonstrating strong functional complementarity. ₃ This refines functional-redundancy interpretations: stable aggregate functions may reflect taxonomic or measurement coverage, but the plant result does not establish redundancy as the mechanism. ₃

Together, the null confirmatory results, coverage-sensitive exploratory effects, and weak within-fraction reproducibility support the hypothesis that coarse functional summaries can conceal environmental selection or confound it with represented abundance. ₁ Environmental differentiation may instead be detectable in taxonomic composition or gene content before it is visible in aggregated categories, connecting this result to ecotype-environment-gene-content. ₁ The broad conclusion is not that contamination has no ecological or functional effect. ₁ Rather, this ENIGMA subset did not show a robust community-level shift in inferred stress-related functional potential when coarse COG-category proxies were aggregated at genus resolution. ₁ The result is compatible with contamination-driven taxonomic turnover that is functionally redundant or operates at species, strain, or pathway resolution finer than the current mapping. ₁

COG provides broad functional categories rather than curated metal-resistance pathways. ₁ The bridge and annotation stack also depends on GTDB and eggNOG conventions, and at genus-level aggregation broad COG-style signals can dilute pathway-specific metal-response effects. ₁ The COG analysis refines this limitation rather than contradicting the ENIGMA null: COG annotations covered approximately 70% of genes, and its 32-species comparison found structured enrichment of mobile, defense, and unknown functions among novel or singleton genes alongside conserved metabolic and housekeeping functions among core genes. ₇ Thus, broad COG scores can contain real, taxonomically widespread functional structure while remaining incomplete and too coarse to establish a contamination association. ₇

The ecotype analysis refines this interpretation by showing that functional differentiation can be widespread within species, but its use of gene-content clusters and COG profiles also retained important uncertainties: only approximately 38% of gene clusters had COG annotations, and without within-species phylogenetic controls it could not distinguish ecological adaptation from phylogenetic or demographic substructure. ₆

Functional-dark-matter prioritization further supports pathway- and gene-level follow-up as a way to test whether broad categories conceal signal. GapMind identified 1,256 organism–pathway pairs across 44 species in which nearly complete pathways co-occurred with strongly fitness-active dark genes, but these were organism-level co-occurrences rather than direct gene-to-enzyme assignments. ₂ Supplementary domain matching produced 42,239 gene–pathway candidates across 3,186 dark genes, including 5,398 high-confidence EC-prefix matches, 4,687 medium-confidence Pfam-family matches, and 32,154 low-confidence keyword matches. ₂ This refines the redundancy hypothesis: pathway-level and gene-level resolution may recover ecologically relevant differences, but computational assignments require experimental validation.

Composite COG assignments containing multiple functional letters were treated as biologically meaningful rather than annotation artifacts in the cross-species analysis; the LV composite showed +0.34% enrichment with 76% consistency, and composite categories were counted once per gene rather than split across component letters. ₇ This supports retaining biologically meaningful composite categories in future resolution comparisons while testing whether their interpretation changes with taxonomic aggregation. ₇

This finding supports pangenome-integration, because bridge ambiguity and unmapped genera directly determine the fraction of community abundance contributing to inferred functions. ₁ It also connects to callability-limited-comparative-inference and environmental-resistome, since missing taxonomic detail can limit comparability and detection of within-species functional differences. ₁ It supports multi-omics-integration by demonstrating that integrating geochemistry, community composition, taxonomy, pangenome clades, and functional annotations does not eliminate coverage, mapping, or resolution constraints. ₁

Rank-Dependent Signals Across Projects

The SSO Subsurface Community Ecology — Spatial Structure, Functional Gradients, and Hydrogeological Drivers supports the coverage-versus-resolution trade-off with an explicit multi-resolution design in which biogeochemical processes were inferred from taxonomy rather than directly measured: class-level annotation covered 22 classes with 78% coverage, whereas genus-level annotation covered 65 annotated genera with 21% coverage. ₁₄ The class-level redox index ranged from 0.047 (M6) to 0.227 (U3). ₁₄ Genus-level functional annotation covered only 21% of total reads (65 of 1,038 genera), so process abundance estimates are lower bounds and the unclassified 56% of reads could harbor additional functional capacity. ₁₄ Species-level classification was ~0%, genus-level classification covered only 44% of sediment reads, and functional inference relied on literature-based trait assignments rather than direct genomic evidence. ₁₄ The report states that the higher-coverage class-level traits (78%) show consistent redox patterns but that sensitivity to the genus coverage gap still needs a formal comparison, and it proposes spatially matched shotgun metagenomics to confirm the taxonomy-to-trait inferences and capture the 56% of reads without genus-level classification. ₁₄ As in the ENIGMA contamination bridge, finer taxonomic resolution came at a steep coverage cost. _14,1

The Gene Function Ecological Agora supports the claim that the rank of aggregation can change a functional verdict, using photosystem ii (PSII) KEGG orthologs (KOs, KEGG Orthology gene-family identifiers) in Cyanobacteria. Each (rank × clade × KO) receives a producer score, a z-score of within-clade paralog expansion against a null, and a consumer score, a z-score against a parent-rank null in which a higher value means the KO is less phylogenetically clumped; verdicts compare these with Cohen’s d (a standardized effect size) against the atlas non-housekeeping median. ₁₅ At genus rank (137 Cyano genera pooled, n=2,350), producer d=+0.08 and consumer d=−0.53 with both Mann–Whitney p values 1.00 (STABLE); at family rank (42 Cyano families, n=705), producer d=+0.20 and consumer d=−1.28 with p=0.97 and 1.00 (STABLE); and at order rank (19 Cyano orders, n=301), producer d=+0.19 and consumer d=−1.42 with p=0.99 and 1.00 (STABLE). ₁₅ At cyanobacteriia class rank, 21 PSII KOs gave producer d=+1.50 and consumer d=+0.70 with p=2×10⁻⁵ and 2×10⁻⁴, supporting H1 as INNOVATOR-EXCHANGE; at Cyanobacteriota phylum rank, 21 KOs gave producer d=+1.60 with p=7×10⁻⁶, but consumer d was NaN and p n/a because no consumer null was available at phylum, giving INNOVATOR-ISOLATED. ₁₅ The class-rank analysis pools all 21 PSII KOs within the single Cyanobacteriia clade. ₁₅

The report interprets this rank gradient as consistent with PSII paralog expansion being a whole-Cyanobacteriia-class property: the genus, family, and order analyses pool 137, 42, and 19 individual Cyano clades respectively, most individual genera do not carry the full PSII paralog repertoire, and pooling therefore dilutes the signal, so the Cyanobacteria PSII signal is class-rank-only and washes out at family, order, and genus ranks. ₁₅ It treats the genus null as expected because class rank is biologically appropriate per Cardona 2018 (PSII evolved before Cyanobacteria diversification), and attributes the phylum reversal to a substrate artifact of insufficient reference consumer data. ₁₅ Both pre-registered class-rank criteria passed at α/4 = 0.0125, with producer d=1.50 and consumer d=0.70 (the consumer result meaning PSII KOs are less phylogenetically clumped than typical non-housekeeping KOs at class rank), but the support is donor-undistinguished. ₁₅ The report itself flags the n=21 sample size and rank-dependent contradictions as a real concern, defending the class result through producer α=2×10⁻⁵, consumer α=2×10⁻⁴, the Bonferroni α/4=0.0125 threshold and the very large effect d=1.50, while stating that the rank-gradient (STABLE at genus/family/order, INNOVATOR-EXCHANGE at class) requires a clearer explanation. ₁₅ This refines the ENIGMA genus-aggregation concern and suggests the hypothesis that the informative rank depends on where a trait is fixed: finer ranks may dilute a clade-wide signal just as coarse ranks may mask strain-level ones. _15,1

The report clarifies that class-rank n=21 counts the 21 PSII KOs themselves (K02703–K02727 minus a few absent in atlas). Each KO is treated as one observation against an atlas-wide reference of 554,280 (KO × clade) tuples at class rank, so n=21 is not a count of independent species observations; the class-rank answer was d=1.50, p=2×10⁻⁵. ₁₅ The report accepts as substantively correct the review critique that the PSII result depends on rank. ₁₅ Its supporting rank-dependence figure shows producer d at genus (n=2,350, d=0.08, STABLE), at class (n=21, d=1.50, INNOVATOR-EXCHANGE) and at phylum (n=21, d=1.60, INNOVATOR-ISOLATED); the figure calls the phylum change a reversal and frames the genus null as expected, not anomalous, for a class-defining innovation per Cardona 2018. ₁₅ The figure description reports the rank-dependent change but does not itself establish its cause. ₁₅

The report gives a rationale for the finer-rank verdicts. Within Cyanobacteria, PSII is a near-universal feature with deep ancestry, so genera should not be innovating it, and the producer signals of d ≈ 0.08–0.20 at finer ranks reflect expected near-uniformity. At phylum rank the verdict shifts from EXCHANGE to ISOLATED solely because reference consumer data are insufficient (NaN); the phylum producer d=1.60 is consistent with the class d=1.50. ₁₅ The report preserves a caveat: PSII Innovator-Exchange is specifically a class-rank phenomenon and not a generic Cyanobacteria-wide signature of HGT — horizontal gene transfer, the movement of genes between lineages rather than from parent to offspring — at every taxonomic resolution. The supported claim is that Cyanobacteriia as a class exchange PSII KOs more than reference clades do. That claim is anchored to the photic-aquatic biome by 2.77× photic enrichment (p<10⁻⁵²) and by a depth signature of 2.05% ancient versus 14.9% atlas-wide, which the report reads as consistent with donor origin per Cardona 2018. ₁₅ The hypothesis table records H3 as SUPPORTED at class rank, with genus rank too sparse. Phenotype support is limited because n=4 BacDive coverage is too thin. ₁₅ This refines the dilution reading above but does not settle it. The near-universality rationale is an interpretation offered alongside the dilution rationale, and the report still states that the rank-gradient requires a clearer explanation. ₁₅

A Phase 1A pilot calibration finding showed a parallel rank gradient in consumer scores computed against a parent-phylum dispersion null. For UniRef50 clusters present in at least 3 clades (the informative subset), consumer z means were −3.77 at genus (n=423), −3.95 at family (329), −4.10 at order (267, peak vertical clumping) and −1.36 at class (172), where clumping weakens. ₁₅ The pilot read class-rank UniRef50 dispersion approaching the random-permutation null as consistent with cross-class HGT becoming detectable above noise, while vertical inheritance dominated from genus through order; this is a consistency argument, not direct proof of transfer. ₁₅ This pilot diagnostic should not be presented as a current atlas-wide result: the within-phylum dispersion metric was later judged wrong because it tracked class coverage rather than HGT activity, with per-phylum Sankoff parsimony identified as the right within-phylum analog and so a proposed rather than an implemented replacement, and the parent-phylum anchor behind these consumer scores was itself flagged as masking intra-phylum transfer, as the next paragraph records. ₁₅

The reference frame used for the null can also manufacture apparent clumping. AMR — antimicrobial resistance — showed strongly negative consumer z (−4.4 to −4.8) at parent-phylum dispersion across genus, family and order ranks. The report attributes this to the parent-phylum anchor masking documented intra-phylum HGT, not to an absence of HGT, because the null measures cross-phylum dispersion. ₁₅ A within-Pseudomonadota dispersion diagnostic found lower dispersion for positive controls than for negative housekeeping genes (pos median 0.25, neg median 0.33). However, the metric was biased by class coverage: housekeeping is pan-Pseudomonadota, whereas HGT-active classes are clade-specific within phylum. The metric was therefore dropped, and per-phylum Sankoff parsimony was named as the right within-phylum analog. ₁₅ A further diagnostic found only AMR strongly Pseudomonadota-biased (median 50 %; 49 % of AMR UniRefs >50 % Pseudomonadota; all other classes ≤ 4 % median). This is an AMR detection-bias confound, because bakta_amr is sourced from AMRFinderPlus’s Pseudomonadota-heavy reference set. ₁₅ These cases support the page’s coverage argument at the level of reference nulls: the taxonomic frame that defines ‘dispersed’ determines which transfer signals are visible. ₁₅

At genus rank, a concordance atlas of 8.64M rows joined to tree-proxy quadrants flagged 169K agreeing, 2.55M conflicting and 5.92M insufficient comparisons. ₁₅ The report classifies most of the 2.55M per-tuple disagreements (“deep-stable + genus-active”) as an expected rank relationship rather than real conflict, and it estimates ~22K real rank-dependent conflicts. That classification is the report’s own interpretation, not an independent test. ₁₅

The mycobacteriaceae × mycolic acid case shows a family-rank effect behaving as a mixture. The producer-side effect was d=0.31 at family rank, supporting H2 Innovator-Isolated, but median leaf consistency (LC, per-rank × clade × KO species prevalence) for mycolic-acid KOs in Mycobacteriaceae was 0.15, lower than the atlas reference of 0.20. ₁₃ Restricting to the mycolic-positive sub-clade (10 of 13 genera with mean LC ≥ 0.5) gave producer d=+0.394 versus family-rank d=+0.309, which the project report states as d=0.39 versus d=0.31, a modest amplification. _13,15 The supported claim therefore narrows from Mycobacteriaceae innovating mycolic acid to sub-clades of Mycobacteriaceae (M. tuberculosis / M. leprae complexes etc.) being mycolic-acid Innovator-Isolated, with the family-rank d=0.31 effect described as a population-mixture average. ₁₅ The discoveries digest generalizes this: when a clade-level supported finding has surprisingly low within-clade leaf consistency, the family-rank effect is a population mixture across positive sub-clade and negative members, and sub-clade-restricted recomputation sharpens the supported claim. ₁₃ This supports the ENIGMA masking concern from a different direction: aggregation over heterogeneous members attenuated, rather than erased, the effect here. _13,1

The project report gives the leaf-consistency basis in detail. For Mycobacteriaceae × mycolic (focal n=33,643), median LC was 0.15 against an atlas reference of 0.20, or 0.75× lower. This means mycolic gains land in clades where the mycolic KO is present in only 15% of family members. The same leaf-consistency figure reports PSII LC=0.88, read as confirming a class-defining innovation, and LC=0.41 for PUL — polysaccharide utilization loci, gene clusters for degrading complex carbohydrates — read as a phylum tendency. ₁₅ The report describes the family-rank producer d=0.31 as an average across two kinds of members. One kind is the mycolic-positive sub-clades, for example the M. tuberculosis, M. leprae and M. avium complexes; the other is mycolic-negative or atypical members. The report notes that this matches established biology, since mycolic-acid biosynthesis is not uniformly distributed across Mycobacteriaceae. It also states that deep-rank aggregation obscured the heterogeneity and that leaf consistency surfaced it. ₁₅ It links the narrowed claim to the Mycobacteriaceae 7.88× host-pathogen biome enrichment, which is also driven primarily by the mycolic-positive pathogenic sub-clades. ₁₅

The leaf-consistency result (LC=0.15 versus atlas reference 0.20) suggested that the family-rank d=0.31 effect was a population-mixture average. Review then noted that Cohen’s d had never been recomputed on the mycolic-positive sub-clade alone. The NB29 analysis did this, using a 13-row table of per-genus mean leaf consistency on the mycolic KO panel as the basis for the sub-clade split. Results by grouping were as follows. ₁₅

• Original family-rank Mycobacteriaceae, across 13 genera: producer d=+0.309, consumer d=−0.193.
• All 13 genera at genus rank, with no LC filter: producer d=+0.326, consumer d=+0.020.
• Mycolic-positive sub-clade (LC ≥ 0.5, 10 genera): producer d=+0.394, consumer d=+0.006.
• Mycolic-low sub-clade (LC < 0.5, 3 genera): producer d=+0.211, consumer d=+0.043. ₁₅

The 10 mycolic-positive genera were Williamsia, Smaragdicoccus, Lawsonella, Hoyosella, Tomitella, Tsukamurella, Dietzia, Gordonia, Nocardia and Williamsia_A, all known mycolate-producing Mycobacteriales/Corynebacteriales genera. The 3 mycolic-low genera were Mycobacterium, Corynebacterium and Rhodococcus. ₁₅ Unexpectedly, Mycobacterium itself fell in the mycolic-low sub-clade (mean LC=0.36, median 0.08). The report attributes this to its 11-KO mycolic panel, which is calibrated to long-chain Mycobacterium-style mycolate biosynthesis: only the M. tuberculosis complex carries all panel KOs at high prevalence, and many environmental Mycobacterium species do not. ₁₅ These statements are measured at different aggregation scales and are not in conflict: a mycolic-positive species complex can sit inside a genus whose genus-mean leaf consistency is low, because that mean also averages the lower-prevalence environmental species. ₁₅ Which taxa fall on each side of the split therefore depends on the scale at which membership is averaged and on how the marker panel is calibrated, which refines rather than overturns the sub-clade framing. ₁₅

The recomputation gave three outcomes. ₁₅

• The producer effect was modestly amplified in the mycolic-positive sub-clade (d=0.39 versus family-rank d=0.31). Within-clade heterogeneity is real but did not dramatically dilute the original finding.
• The consumer signal vanished at genus rank (d=+0.006 versus d=−0.193 at family rank). The report attributes the original consumer effect to family-rank aggregation rather than to sub-clade-specific signal.
• The refined supported claim is that the mycolate-producing sub-clade of Mycobacteriaceae (10 of 13 genera) is mycolic-acid Innovator-Isolated, with producer d=0.39 at genus rank. The consumer-side component is weaker than the family-rank framing implied. ₁₅

This supports both halves of the page’s masking argument. On the producer side, aggregation attenuated a real effect. On the consumer side, aggregation produced an effect that disappeared at finer rank. ₁₅

Scale also changed a published point estimate: the Alm 2006 r ≈ 0.74 was not reproduced at full GTDB scale (r=0.10–0.29), although methodology generalization held (architectural concordance r=0.67); the report proposes substrate-scale heterogeneity, tree-aware versus paralog-count operationalization, and tree-rank granularity as mechanisms for the point estimate not surviving the scaling from 207 to 18,989 genomes. ₁₅ Tree-rank granularity is thus a candidate explanation, not an established one. ₁₅

Ortholog-tier and clade-size resolution impose similar limits in subsurface bacillota-b analyses. The Subsurface Bacillota_B Specialization — What Distinguishes Deep-Clay Lineages from Soil Congeners? used Firmicutes-level eggNOG orthologous groups (OGs, gene families defined at a chosen taxonomic level; eggNOG tax 1239) where present, falling back to Bacteria/root; some enriched OGs are at root level, which is coarse, and no Bacillota_B-specific eggNOG OG tier exists because Bacillota_B is GTDB-defined whereas eggNOG uses NCBI taxonomy, where Bacillota_B genomes are spread across Firmicutes / Negativicutes / Tissierellia at the legacy class level. ₁₆ The Self-Sufficiency, Anaerobic Toolkit, and Cultivation Bias in Clay-Confined Cultured Bacterial Genomes reports a 5/5 Bacillota_B sulfate-reduction (SR) enrichment but cautions that the n=5 vs 19 structure may hide further granularity, because desulfosporosinus, BRH-c8a, and BRH-c4a may differ in which non-SR features they carry. ₁₇ Both support the taxonomy-mismatch component of resolution: GTDB-defined clades and NCBI-based annotation tiers do not align, and small clade-level counts can conceal genus-level differences. _16,17

Aggregation inside validation datasets refines the functional-dark-matter evidence cited above. Its NMDC validation operates at genus level, mapping NMDC taxon columns to pangenome genera via ncbi_taxid, which may miss species-specific dark-gene signals; only 5 of 6 carrier genera were matched, and the high significance rate (76/105) likely reflects the dominance of common genera such as Pseudomonas and Klebsiella in both datasets. ₂ Likewise, the Metabolic Consistency of Pseudomonas FW300-N2E3 Across Four KBase Data Lakehouse Databases notes that bacdive aggregates across P. fluorescens strains, which under GTDB reclassification span a broad clade (Pseudomonas_E fluorescens_E); per-strain majority-vote consensus deduplication is applied before computing species-level utilization rates, but strain-level variation (e.g., trehalose: 1/6 positive) means the species consensus may not reflect FW300-N2E3’s specific capabilities. ₁₈

Not every taxon shows a below-species functional signal. A discoveries-digest entry reports zero strain-adaptation gene signal in Kumbhari and interprets the effect as species-abundance-mediated rather than strain-content-mediated, with phage targeting producing predictable activity depletion and no within-species strain-content escape route; the underlying project locates that null specifically in F. plautii, as the per-species evidence below records. _13,19 Across the broader Kumbhari panel the same project records strain-adaptation gene content as discriminating disease- from health-adapted strains, with IBD-biased adaptation enrichment OR=1.38 (p=2.4e-6) and housekeeping depletion OR=0.62 (p=6.4e-20), so neither the species-abundance reading nor the absence of a strain-content escape route is a dataset-wide conclusion. ₁₉ This per-taxon null qualifies the hypothesis that finer-than-species resolution will generally recover hidden functional differences: in some taxa species abundance, not strain gene content, may carry the functional effect. ₁₃

Other projects report the same limit on attribution below genus or community level. In the genotype-to-phenotype analysis, Microbial Atlas biogeography was at genus level (16S resolution), and species-level biogeography was available for only 20 pangenome-linked strains with verified GTDB matches. ₂₀ The Harvard Forest warming report leaves open which MAGs (MAG, metagenome-assembled genome) in its 298-MAG cohort carry the upregulated pmoA/pmoB. It also leaves open whether those carriers are USCα methanotrophs or are distributed across phyla, so the functional signal is not yet resolved to specific genomes. ₂₁

In the IBD phage-targeting analysis, read-level enrichment of CB-ORFs — open reading frames (ORFs) drawn from the project’s 5,157-entry ref_cborf_enrichment catalog, whose CB prefix the report does not expand — was resolved per Tier-A core species in Crohn’s disease (CD) versus healthy controls (HC). The table below gives CB-ORFs matched, CD-up counts at FDR<0.10, CD-down counts and mean effects. ₁₉

| Species ₁₉ | CB-ORFs matched | CD-up at FDR<0.10 | CD-down | Mean effect | |---|---:|---:|---:|---:| | E. bolteae | 11 | 9 (82 %) | 0 | +2.81 | | F. plautii | 5 | 2 (40 %) | 0 | +0.77 | | E. coli | 51 | 14 (27 %) | 0 | +1.56 | | M. gnavus | 19 | 5 (26 %) | 0 | +1.43 | | H. hathewayi | 7 | 1 (14 %) | 0 | +0.52 | | E. lenta | 13 | 0 | 7 (54 %) | -0.28 | | Background catalog | 5,052 | 2.5 % | — | -0.18 |

Thus 5 of 6 Tier-A core species had CD-up CB-ORF rates above the 2.5 % catalog background (range 14–82 %). E. lenta was the exception, with 54 % of its CB-ORFs CD-down. ₁₉

Other IBD signals could not be resolved to species. The ebf/ecf RPKM signal (RPKM, reads per kilobase per million mapped reads, a sequencing-depth-normalized abundance measure) and the polyamine and PUFA elevations are sample-level CD biomarkers without per-species attribution. The report positions them for treatment-response monitoring, not for per-species cocktail target selection. ₁₉ Five Tier-A species (M. gnavus, E. lenta, E. bolteae, H. hathewayi, F. plautii) are BGC-mechanism dark (BGC, biosynthetic gene cluster). Their design therefore relies on Tier-A scoring, NB06 module membership and the NB07b within-carrier metabolic signature, not on BGC presence. ₁₉ This supports the page’s argument in a translational setting: a signal measured only at sample level cannot drive decisions that require taxon-level attribution. ₁₉

Pathway-level attribution in the same project concentrated on one species. The top 25 pathway-pathobiont pairs are all Escherichia coli pathways, which the report describes as recapitulating known AIEC (adherent-invasive E. coli) biology. ₁₉ Across Tier-A species the pair-count distribution was E. coli 105 pairs > 0.4 (76% of total signal), H. hathewayi 16, M. gnavus 8, F. plautii 7, E. lenta 1 and E. bolteae 0; the report attributes E. coli’s domination to a highly specialized AIEC functional repertoire well-matched to MetaCyc, higher relative-abundance variance, and E. coli genome content being especially well-represented in HUMAnN3 compared with more obligate-anaerobe Tier-A members, and reads E. bolteae’s zero pathway-attribution despite being NB05-actionable as a sign that its CD-up signal may sit at the BGC level or the Kumbhari strain-frequency level rather than the HUMAnN3-pathway level. ₁₉

The rank at which pathways were tested also decided whether a category test was possible at all. The unstratified clause-(b) test failed because only 3 of 52 CD-up unstratified MetaCyc pathways landed in the 7 a-priori IBD categories, leaving the test structurally underpowered, so the project retested the same question at the species-resolved HUMAnN3 stratified-pathway form PWY-XXX|g__species, asking whether per-species CD-up pathways concentrate in those categories. ₁₉ Species resolution did not rescue the test:

| Species ₁₉ | CD-up | CD-down | H3a (b) test | Reason | |---|---:|---:|---|---| | Hungatella hathewayi | 16 | 17 | underpowered | Only 1 of 16 CD-up pathways in 7-cat set | | Mediterraneibacter gnavus | 0 | 0 | untestable | Background only 4 pathways in 7-cat; no CD-up | | Escherichia coli | 2 | 18 | underpowered | 0 of 2 CD-up in 7-cat | | Eggerthella lenta | 0 | 0 | untestable | Background 3 pathways in 7-cat | | Flavonifractor plautii | 0 | 2 | underpowered | 0 of 0 CD-up in 7-cat | | Enterocloster bolteae | 1 | 3 | untestable | Background 3 pathways in 7-cat |

These verdicts are schema-specific, and their chronology matters. At the earlier NB07a stage the project positioned the roughly 42K stratified pathway-species combinations as the resource to query directly and called clause (b) a follow-up question, not a current refutation — a historical statement about that stage. ₁₉ The combined NB07a + NB07b conclusion then states that the H3a (b) refutation is real for the v1.7-stated category set, while explicitly not meaning that no compositional themes exist: it means a different category schema, broader and programmatically derived from MetaCyc taxonomy, is needed. ₁₉ A subsequent v1.8 analysis built on a ModelSEED class hierarchy completely reverses the v1.7 H3a (b) “FAIL” verdict, driven entirely by category-schema choice and not by any change in the underlying data. ₁₉ This supports the page’s distinction between a null result and an untestable configuration, and adds that the category schema, not only the taxonomic rank, can decide the verdict. ₁₉

Under the revised v1.8 schema, and where species-rank counts were sufficient, themes did emerge. H. hathewayi, with 16 CD-up pathways, had two themes supported: purine/pyrimidine recycling with OR (odds ratio) = 4.86 and FDR = 0.048 over 7/16 CD-up pathways, and TMA/choline with OR = 9.33 and FDR = 0.048 over 4/16 CD-up pathways. ₁₉ The report links these to the species’ own CD-up pathway list — pentose phosphate, purine nucleobases degradation and PWY-6803 phosphatidylcholine acyl editing — and notes that H. hathewayi is a known TMA (trimethylamine) producer with CutC/D activity, calling the choline-metabolism enrichment mechanistically coherent. ₁₉ Other Tier-A core species had insufficient species-level CD-up pathway counts for per-theme power: M. gnavus, E. lenta and F. plautii had 0 CD-up, while E. coli and E. bolteae had 1-2. ₁₉

A cohort-level functional “theme” can likewise be concentrated in one taxon. Mean correlations across 15 iron pathways were E. coli +0.45, H. hathewayi +0.20, F. plautii +0.17, M. gnavus +0.16, E. bolteae +0.09 and E. lenta +0.03. ₁₉ Those 15 pathways include ENTBACSYN-PWY (enterobactin biosynthesis, E. coli-canonical), HEMESYN2-PWY (heme biosynthesis II) and 8 menaquinol-biosynthesis pathways, and the E. coli figure is a Spearman rank correlation, ρ = +0.45 across 2,674 CMD_IBD samples, so iron-pathway abundance and E. coli abundance covary in rank more strongly than for any other Tier-A core species, which show weak iron-pathway coupling. ₁₉ A rank correlation of this kind does not establish proportional scaling or pathway ownership; reading the iron-pathway signal as dominantly carried by E. coli is the report’s own attribution, and on that basis the report narrows its framing from “all CD pathobionts have iron specialization” to E. coli driving the iron-acquisition theme, with the AIEC-subset and strain-specialization account offered as interpretation rather than measurement. ₁₉

The project then checked that narrowing at a different evidence layer. The iron-theme narrowing to E. coli is correlation-based, derived from species × pathway Spearman ρ across CMD_IBD samples, and 8 of 15 iron pathways are menaquinol biosynthesis — canonical bacterial respiratory quinones, not strictly E. coli-specific — while the strongest signature pathways (ENTBACSYN-PWY, HEMESYN2-PWY) are E. coli-canonical. ₁₉ Direct genomic content evidence then showed that E. coli alone among Tier-A core carries iron-siderophore matches to the MIBiG reference biosynthetic-gene-cluster collection, 54 BGCs versus 0 for the other 5 actionable species. ₁₉ Finer HUMAnN3 species-stratified per-pathway attribution would extend this but requires raw reads and was dropped, leaving genomic BGC content as the project’s equivalent fine-grained attribution path within a no-raw-reads scope. ₁₉ This supports the page’s claim that the attainable attribution rank is fixed by available data rather than freely chosen. ₁₉

Stratifying within carriers reversed the direction of per-species signals. A within-carrier CD-versus-nonIBD meta-analysis with single-substudy fallback found H. hathewayi biosynthesis-up and sugar-degradation-down, while E. coli pathways were CD-DOWN within carriers, which the report reads as AIEC strain-level specialization. ₁₉ It offers two non-mutually-exclusive interpretations of the E. coli pattern: that CD’s E. coli are an AIEC-specialized subset that has shed peripheral metabolic capabilities, or that they face metabolic competition from co-abundant Klebsiella and other Enterobacteriaceae so that per-cell read-mapping share is lower across pathways. ₁₉ The figure NB07b_stratified_pathway_DA.png presents the per-Tier-A-core-species stratified pathway differential-abundance effect heatmap with per-species CD-up/down counts. ₁₉

Below species rank the project reached opposite answers for two taxa. For F. plautii it found zero FDR<0.10 strain-adaptation genes in the Kumbhari analysis out of 3,245 genes total tested, and the deep-dive output data/nb10a_f_plautii_strain_adaptation.tsv was empty, with no FDR-passing genes. ₁₉ The report interprets this as F. plautii CD-association operating through species-level abundance rather than strain-level genomic adaptation, reasoning that the 7α-dehydroxylation activity is encoded by core bai-operon genes that are presumably present in essentially all F. plautii strains — a presumption the report marks as such rather than measures. ₁₉

For E. coli the same report concludes the opposite, treating the CD-association as strain-content-mediated through the AIEC subset, and it proposes an AIEC strain-resolution diagnostic for the 8/23 E. coli-positive UC Davis patients. ₁₉ It cites prior work reporting high prevalence of adherent-invasive Escherichia coli associated with ileal mucosa in Crohn’s disease. ₁₉ The contrast refines the Kumbhari strain null recorded above: whether a functional effect sits at species abundance or at strain gene content is a per-taxon empirical question, not a general property of a resolution. _19,13

Two further limits attach to the taxon label itself rather than to data volume. A species-level enrichment need not carry the phenotype implied by the name: S. thermophilus appears in anti-inflammatory multi-strain probiotic formulations, so its species-level CD-up signal does not automatically imply pathobiont status, and strain-level literature and engraftment evidence were used as the disambiguation step. ₁₉ Serology attributes no more cleanly: OmpC and ANCA cross-react with related antigens beyond their named targets, so per-species attributions drawn from them are correlative rather than exclusive. ₁₉

Aggregation on the viral side of the same project removed species-level signal outright. H. hathewayi and M. gnavus correlate negatively with the “Unknown” phage family, which captures 80% of HMP2 viromics observations that VirMAP could not classify, a pattern the report reads as pathobiont blooms displacing some unclassified phages, consistent with reduced phage diversity in CD dysbiosis. ₁₉ Because VirMAP left 80% of phages classified as “Unknown” family, per-family differential-abundance and correlation power is limited. ₁₉ The report concludes that “Unknown”-family aggregation hides species-level signal and that per-virus differential abundance is more informative than per-family for IBD-relevant phages, which supports the page’s central claim in a non-bacterial dataset. ₁₉

Tensions

The relaxed coverage-adjusted model yielded FDR q = 0.0462, whereas the strict coverage-adjusted model yielded FDR q = 0.130, and both confirmatory genus-level modes were null after predeclared testing. ₁ This is a methodological tension between a coverage-sensitive exploratory association and a resolution- and specification-robust association, not evidence that the estimates should be averaged. ₁ The difference could reflect retained abundance, bridge ambiguity, covariate sensitivity, taxonomic aggregation, or a combination of these factors. ₁

The functional-dark-matter report adds a related coverage tension: 29 of 47 testable lab–field clusters were concordant (61.7%), but the one-sided binomial test gave p = 0.072, whereas Fisher’s combined probability across 47 individual tests gave p = 0.031. ₂ NMDC confirmed all 4 testable pre-registered abiotic predictions, but the report cautions that 441 of 449 exploratory trait tests reached FDR < 0.05, largely because of compositional coupling. ₂ These results refine rather than resolve the ENIGMA tension: finer-resolution validation can produce coherent directional evidence, while broad or exploratory associations remain sensitive to coverage and compositional structure.

The within-species ecotype evidence supports the premise that finer functional structure exists, but it does not resolve whether that structure is ecological or phylogenetic because the ecotype analysis lacked within-species phylogenetic controls. ₆ This limits direct extrapolation from gene-content ecotypes to contamination-associated community functions: the ENIGMA analysis is null or mode-sensitive at genus resolution, whereas the ecotype analysis detects functional differentiation within species. _{1 6}

The plant-association evidence refines rather than resolves this tension: compartment effects remained statistically significant but explained only 0.060 of location-only variance, and the earlier large effect was reduced to R² = 0.072 after removal of genome-rich species. ₃ The same report found 78.7% of 1,115 plant-associated species classified as dual-nature under a refined marker scheme, but categorical assignments misclassified all four neutral controls in a curated 18-organism panel. ₃ These results show that finer or richer marker resolution can produce extensive apparent functional structure while still being sensitive to taxonomic composition, genome size, and classifier definitions. ₃

The COG comparison found consistent core-versus-novel functional partitioning across 32 species, whereas the ENIGMA analysis did not find a robust contamination-associated shift in coarse COG proxies at genus resolution. _{7 1} This is not a direct contradiction because the studies test different signals: broad evolutionary partitioning by gene novelty versus a site-level ecological association after taxonomic bridging. _{7 1}

The multi-rank results do not agree on which rank is informative: the PSII effect was absent at genus, family, and order and present only at class, attributed to dilution when pooling genera lacking the full paralog repertoire, whereas the mycolic-acid family-rank effect was attenuated relative to a sub-clade restriction. _15,13 ENIGMA and SSO analyses, by contrast, frame genus-level aggregation as potentially masking finer-scale effects. _1,14 These should not be averaged into a single preferred rank; the project report itself states that the PSII rank-gradient requires a clearer explanation. ₁₅

The report offers two accounts of the PSII STABLE verdicts at genus, family and order. One is dilution, because most individual genera do not carry the full PSII paralog repertoire. The other is expected near-uniformity, because PSII is a near-universal, deeply ancestral feature within Cyanobacteria, with PSII leaf consistency of 0.88. ₁₅ Gene presence and paralog expansion are distinct measurements, so these accounts are not necessarily incompatible; what remains unresolved is that the corpus contains no analysis separating them, and the report itself states that the rank-gradient requires a clearer explanation. ₁₅ Separately, the mycolic recomputation places Mycobacterium in the mycolic-low group by genus-mean LC=0.36 while the leaf-consistency narrative names the M. tuberculosis complex as mycolic-positive; these are different aggregation scales rather than opposing claims, and no analysis in the corpus compares them directly. ₁₅ Neither open comparison should be settled by preferring one framing. ₁₅

A per-species functional attribution in the same project did not survive a change of data level, and the two results should not be merged. The H. hathewayi TMA/choline finding from curated-metagenome (cMD) pathway-level analysis does not strongly replicate at HMP2 sample level, where H. hathewayi × choline gave ρ=−0.07 (non-significant) and E. coli dominates, narrowing the TMA/choline signal to a combined Enterobacteriaceae + Lachnospiraceae signal rather than a specifically H. hathewayi one. ₁₉ The report’s possible reasons are that HMP2 has lower H. hathewayi prevalence (25% of paired samples) than the cMD studies that drove the earlier finding, and that at sample level E. coli’s high prevalence (50%) and abundance variance dominate the cohort-correlation signal. ₁₉ This sits in tension with the species-rank theme result above (OR = 9.33, FDR = 0.048): the two were computed on different cohorts at different aggregation levels, and the corpus contains no analysis reconciling them, so neither should be preferred by default. ₁₉

Open Directions

• Obtain species- or strain-level ENIGMA or metagenomic data and repeat strict, relaxed, and unique-clade bridge analyses to test whether contamination-linked functional associations emerge below Genus resolution. ₁
• Replace broad COG-fraction proxies with curated metal-stress gene sets and pathway-level summaries, then test whether higher-resolution functional definitions distinguish taxonomic turnover from functional redundancy and whether associations persist after mapped-coverage adjustment. ₁
• Quantify the contribution of the 862 unmapped genera and expand the genus-to-clade bridge, evaluating whether missing abundance explains differences between mapping modes, weak broad functional signals, and score instability. ₁
• Fit mixed-effects or hierarchical models including depth, location cluster, sampling date, and compositional controls to test whether the relaxed-mode defense association persists under richer site structure. ₁
• Compare aggregate, within-fraction, and high-coverage analyses using the same resolution-specific samples and bootstrap procedure to determine whether coverage or taxonomic ambiguity is the dominant source of sensitivity. ₁
• Overlay core-genome phylogenetic trees on the 12-species ecotype assignments and repeat COG enrichment tests to separate within-species ecological differentiation from phylogenetic structure. ₆
• Extend the species- and strain-resolution bridge with the ecotype assignments, then test whether functional differences detected in gene-content ecotypes predict contamination-associated community scores after coverage adjustment. ₆
• Repeat the taxonomic-resolution comparison with COG L, COG V, and composite LV features separately, testing whether mobile-element and defense signals remain detectable as clade specificity increases and mapped abundance decreases. ₇
• Add plant-compartment and host metadata to species- and strain-level bridge analyses, then test whether the small compartment signal and the 11.7% pangenome–MGnify genus overlap are explained by taxonomic sampling, detection process, or genuine functional differentiation. ₃
• Expand KBase Data Lakehouse phylogenetic-tree coverage and apply within-species phylogenetic controls to the plant-associated ecotypes, especially the 47 of 65 candidate species lacking tree data, before interpreting subclade functional associations. ₃
• Use the functional-dark-matter covering sets and top candidate organisms to design cross-taxonomic validation experiments, testing whether gene-level fitness phenotypes and pathway predictions remain concordant when reference taxa, condition coverage, and conservation breadth are varied. ₂
• Compare detailed results with Contamination Gradient vs Functional Potential in ENIGMA Communities and integrate the workflow with multi-omics-integration to quantify how geochemistry, community composition, pangenome mapping, and functional annotation jointly constrain inference. ₁
• Recompute PSII producer and consumer scores restricted to Cyanobacteria genera that carry the full PSII paralog repertoire and compare them with the 137-genus pooled result, testing whether dilution explains the STABLE verdicts at genus, family, and order. ₁₅
• Screen other family-rank supported findings for low within-clade leaf consistency and recompute effects on sub-clades with mean LC ≥ 0.5, testing how general the population-mixture rule is. ₁₃
• Run the formal SSO genus-versus-class sensitivity comparison and test both against spatially matched shotgun metagenomes, determining whether trait-based process maps survive direct genomic evidence. ₁₄
• Split the Bacillota_B sulfate-reduction enrichment into Desulfosporosinus, BRH-c8a, and BRH-c4a and test their non-SR accessory features separately, checking whether the 5/5 clade-level result hides genus-level differences. ₁₇
• Implement the proposed per-phylum Sankoff parsimony analog and recompute AMR consumer scores with it instead of the parent-phylum anchor, testing whether intra-phylum HGT becomes visible once the cross-phylum null is removed. ₁₅
• Re-score the Mycobacteriaceae sub-clade split with a mycolic KO panel not calibrated to long-chain Mycobacterium-style mycolates, testing whether Mycobacterium leaves the mycolic-low group and whether the producer d=0.39 result holds. ₁₅
• Tabulate per-genus PSII KO prevalence and per-genus paralog counts within Cyanobacteria to separate the dilution account from the near-universality account of the finer-rank STABLE verdicts. ₁₅
• Map pmoA/pmoB carriage across the 298-MAG Harvard Forest cohort to attribute the upregulated signal to USCα methanotrophs or to other phyla. ₂₁
• Move the Pseudomonas environment prediction from species level to genome level, using the full 789K genome-pathway matrix with genome-specific isolation sources, to test whether statistical power for environment prediction increases. ₂₂
• Run the proposed AIEC strain-resolution diagnostic on the 8/23 E. coli-positive UC Davis patients, testing whether the E. coli CD-association is strain-content-mediated as the report interprets. ₁₉
• Re-run the IBD-category clause at species rank with the broader ModelSEED-derived class schema across the roughly 42K HUMAnN3 stratified pathway-species combinations, testing whether per-species tests that were underpowered or untestable under the v1.7 seven-category set become decidable. ₁₉
• Replace “Unknown”-family phage aggregation with per-virus differential abundance in the HMP2 viromics data, recovering species-level phage signal lost to the 80% unclassified fraction. ₁₉
• Independently validate the genus-level conservation assumption behind taxonomy-inferred prophage burden before reading the 57 module-abiotic correlations as prophage ecology rather than taxonomic composition. ₁₂