BERIL Knowledge Wiki

Dataset in Data Infrastructure, Provenance, and Research Practice, Ecotypes and Environmental Gene-Content Differentiation, Laboratory Fitness Versus Natural Selection and Gene Conservation, Lanthanide-Dependent Methylotrophy, Metabolic Traits, Auxotrophy, and Community Dependence, Pangenome Structure and Genome Evolution, Resistance, Defense Systems, and Mobile Elements, Sampling Coverage and Statistical Confounding

Genome Taxonomy Database

Well corroborated. AI-generated wiki from AI-conducted research reports. Partially reviewed by scientists. See About This Wiki.

Sources cited
+1424 project reports and 2 cross-project digests, listed in the margin
Relations
supports 2refines 2
Linked from
41 pages12 topics

What this entity is

  • Canonical name: Genome Taxonomy Database.
  • Known alias: GTDB.
  • Stable external identifier: Not specified in the source document. 1

GTDB is a genome-taxonomy and pangenome framework for comparative microbial analysis. 2 It supports pangenome-integration by estimating gene prevalence across sampled genomes and classifying AMR genes as core or accessory. 1 The Within-Species AMR Strain Variation extends this use to within-species AMR variation across 1,305 species and 180,025 genomes in the GTDB/NCBI collection. 3

The BERDL Data Atlas — Inventory, Topic Map, and Cross-Reference Synergies reports 1,011,650,903 KBase pangenome gene rows spanning 293,059 genomes, 27,690 GTDB species clades or pangenomes, and 132.5M gene clusters; these are available-data-layer counts rather than independently deduplicated records, with genome_id and ncbi_taxon_id serving as cross-reference anchors. 4

The Plant Microbiome Ecotypes — Compartment-Specific Functional Guilds and Their Genetic Architecture extends the GTDB-r214 comparative-ecology census to plant association: across 293,059 GTDB r214 genomes it identified 1,136 plant-associated species and 7,995 plant-associated genomes spanning root, rhizosphere, phyllosphere, endophyte, and other plant contexts. 5 The analysis used GTDB pangenomes with marker surveys, eggNOG, GapMind, NMDC, MGnify, and within-species phylogenetic data; compartment and complementarity effects were small, categorical cohort labels were poor species-level discriminators, and phylogenetic-tree coverage limited subclade inference. 5

The Carbon Source Utilization Predicts Ecology and Lifestyle in Pseudomonas extends this atlas to standardized GapMind carbon-pathway profiles across 12,732 genomes and 433 Pseudomonas species clades in GTDB r214. 6 Pseudomonas s.s. contained 19 species and 6,905 genomes, whereas Pseudomonas_E contained 398 species and 5,687 genomes; the genome-level carbon-pathway table contained 789,012 rows. 6 The dominant carbon-profile axis separated these deep subgenera, while lifestyle categories substantially overlapped within Pseudomonas_E. 6

Pangenome, pathway, and fitness integration

GTDB supports genome-scale metabolic capability analysis. The Metabolic Capability vs Dependency crossed GapMind pathway-completeness predictions with Fitness Browser RB-TnSeq (random-barcode transposon sequencing) across 161 organism-pathway combinations from 7 model bacteria and extended pathway-variation analysis to 2,810 GTDB species with at least 10 genomes. 7 Complete pathways were not equivalent to experimentally observed dependency: the 161 pairs comprised 57 (35.4%) Active Dependency, 66 (41.0%) Latent Capability, 24 (14.9%) Incomplete but Important, and 14 (8.7%) Missing pairs. 7 Active Dependency required a complete pathway containing fitness-important genes, whereas Latent Capability required completeness without significant defects under standard conditions. 7 Only 7 of 48 Fitness Browser organisms had matching GapMind genome data, so GTDB-linked fitness conclusions remain limited relative to the approximately 293,000 genomes used for broader pathway analysis. 7

The same study supports and refines GTDB’s pangenome role: variable pathway count correlated with pangenome openness at raw Spearman rho=0.327, p=7.2e-71 and partial rho=0.530, p=2.83e-203 after controlling for genome count, across 2,810 GTDB species with at least 10 genomes. 7 The signal was positive in 13 of 18 genera and significant at p<0.05 in 5 of 18: Clostridium, Eubacterium, Mesorhizobium, Pseudomonas, and Streptomyces; the strongest within-genus correlations were rho=0.534 for Clostridium and rho=0.506 for Eubacterium. 7 The absence of full phylogenetic independent contrasts and the association between genome count and observed variation constrain causal interpretation. 7

The Metabolic Capability vs Metabolic Dependency extends and refines GTDB integration from gene prevalence to genome-scale metabolic capability, using GTDB taxonomy with GapMind pathway predictions and pangenome data covering 7,334 species while comparing 48 Fitness Browser organisms and 1,695 complete pathway-organism pairs. 8 Only 48 organisms were fitness-tested among 293,000 genomes with pathway predictions. 8 At 22 species-clade level, latent capability rate correlated positively with pangenome openness (Spearman ρ = 0.69, p = 0.0004, n = 22 clades), whereas pathway-level conservation did not distinguish latent capabilities from active dependencies (Mann–Whitney U p = 0.94; rank-biserial r = 0.052). 8

Across 80 GapMind pathways—18 amino-acid biosynthesis and 62 carbon-source utilization pathways—core-only versus all-gene completeness gaps were 0.146 for leucine and valine, 0.141 for arginine, 0.140 for lysine, and 0.140 for threonine biosynthesis. 7 These results suggest accessory contributions to pathway completeness and are consistent with community-level sharing hypotheses, but do not demonstrate metabolite exchange. 7 Among 225 species with at least 50 genomes and at least 3 variable pathways, binary pathway profiles yielded a median of 4 metabolic ecotypes per species and a maximum of 8 in Alistipes onderdonkii and Barnesiella intestinihominis; ecotype count correlated with pangenome openness at raw Spearman rho=0.262, p=6.8e-05 and partial rho=0.322, p=8.0e-07 after controlling for genome count. 7

The pathway-capability analysis supports condition-specific interpretation: all 66 aggregate Latent Capability pairs became fitness-important under at least one condition type, with nitrogen limitation, stress, and carbon limitation the most frequent triggers. 7 It also refines this result because the median-based importance threshold can reclassify pairs by construction when applied to condition-specific subsets; independent calibration against known essentials is needed. 7 Active Dependencies had mean core gene completeness of 0.986 versus 0.975 for Latent Capabilities. 7

The Community Metabolic Ecology via NMDC × Pangenome Integration extends and scales this role to community metabolic ecology: GTDB pangenome species were bridged to GapMind pathway records and NMDC community data across 220 samples, 27,690 GTDB species, and 80 pathways, with 92% of samples mapping at least 85% of community abundance to GTDB pangenome species. 2 Community-weighted pathway completeness was compared with metabolomics in a merged analysis-ready matrix of 174 samples, supporting a test of community-scale Black Queen Hypothesis dynamics, while GapMind measures genomic potential rather than expression and abiotic covariates were unavailable. 2 Soil and Freshwater communities separated in pathway-completeness PCA, and 17 of 18 amino-acid pathways differed across ecosystem types, although habitat physicochemistry and taxonomic composition could not be separated. 2

The plant study supports and qualifies this pathway-integration role. BacDive metabolite-utilization data agreed with GapMind predictions at 83.1% consistency, but the GapMind core-level completeness score was 0% across plant compartments and may be too stringent for broad taxonomic comparisons. 5 Corrected pair-level analyses found weak redundancy rather than complementarity among co-occurring plant-associated genera: Cohen’s d ≈ −0.4 with permutation p < 0.001, d ≈ −0.43 under corrected max aggregation, and d = −0.39 with prevalence weighting. 5 The NMDC co-occurrence test remained structurally underpowered, with 0 PGP-dominant, 3 pathogen-dominant, and 66 dual-or-mixed genera among 69 co-occurring genera. 5

The Pseudomonas analysis supports and qualifies use of GTDB-linked GapMind profiles for metabolic-ecotype analysis. Among 54 free-living and plant-associated species with at least 5 genomes and at least 60% majority-environment agreement, carbon profiles were associated with isolation environment by a 999-permutation test (p = 0.006); between-group mean distance was 2.054 versus within-group mean distance 1.890, with ratio 1.087. 6 PCA of the 62-pathway profiles captured 74.9% of variance in the first 5 components, with PC1 explaining 31.2% and PC2 17.9%. 6 A Random Forest classifier—an ensemble method for classification—achieved balanced accuracy 0.408 +/- 0.169 in 5-fold stratified cross-validation versus a 0.250 chance baseline, indicating a real but modest ecological signal rather than reliable fine-grained environment prediction. 6

Free-living and plant-associated Pseudomonas species had a median of 57 pathways complete in more than 50% of genomes versus 55 in host-associated species; within Pseudomonas_E, plant-associated species averaged 56.7 pathways, free-living species 56.1, and host-associated species 55.2. 6 The dominant comparison was instead the subgenus split: host-associated Pseudomonas s.s. showed strong loss of plant-derived sugar and sugar-alcohol pathways relative to Pseudomonas_E, while amino-acid and core organic-acid pathways remained near-universal. 6 Among 7 Pseudomonas s.s. and 189 Pseudomonas_E species with at least 5 genomes, 43 of 62 pathways differed significantly by Mann–Whitney U testing with Benjamini–Hochberg false-discovery-rate correction (q < 0.05); xylose completeness was 0.0% versus 74.4%, ribose 27.9% versus 92.0%, arabinose 0.0% versus 62.6%, galacturonate 28.6% versus 88.4%, myo-inositol 0.0% versus 58.8%, mannitol 25.9% versus 77.5%, and sorbitol 25.9% versus 77.4%, respectively comparing the P. aeruginosa and P. fluorescens groups. 6 The rhamnose and fucose pattern—more complete in P. aeruginosa at 66.8% than in the P. fluorescens group at 41.3% and 45.1%—was not significant after correction and may reflect rhamnolipid-associated virulence biology. 6

This result extends prior metabolic-ecotype findings but refines their ecological scope: the filtered dataset contained 51 species (soil 13, freshwater 13, plant_surface 19, and rhizosphere 6), and the most discriminating pathways were D-serine (importance 0.132), arabinose (0.094), rhamnose (0.086), fucose (0.085), and xylose (0.070). 6 Proposed follow-up includes within-species analysis of P. fluorescens and P. putida ecotypes and cross-reference with RB-TnSeq fitness data; the current 62 pathways omit important aromatic capabilities such as toluene, naphthalene, and benzoate degradation. 6

Pangenome atlas extensions

The PGP Gene Distribution Across Environments & Pangenomes extends and refines GTDB pangenome analysis from AMR and metabolic capability to plant-growth-promoting (PGP) gene ecology. Among 27,702 species, 11,272 carried at least one of 13 PGP markers; the analysis included 32,736 PGP gene clusters, 27,690 species with GapMind scores, and 291,279 genomes with isolation-source metadata. 9 Eight of 10 focal-gene pairs were significant after Benjamini–Hochberg false-discovery-rate correction: five positive and three negative associations. 9 The strongest positive association was pqqC × acdS (OR = 7.24, n = 286 co-occurring species, q = 1.2e-83); pqqC also co-occurred with hcnC (OR = 1.91) and ipdC (OR = 1.55). 9

This refines ecological-guild interpretation: nifH was negatively associated with hcnC (OR = 0.23, q = 5.8e-29) and pqqC (OR = 0.57, q = 2.9e-19), but not ipdC (OR = 1.13, q = 0.54); only 157 species (1.4%) carried at least three focal traits. 9 In soil/rhizosphere species, acdS, pqqC, and hcnC were enriched: acdS 15.8% versus 2.6% (OR = 7.02, q = 5.1e-62), pqqC 43.8% versus 21.2% (OR = 2.90, q = 2.8e-53), and hcnC 11.3% versus 6.4% (OR = 1.85, q = 6.1e-08); nifH was depleted at 12.9% versus 19.7% (OR = 0.60, q = 5.5e-08), while ipdC was not significantly enriched (OR = 0.79, q = 0.47). 9 acdS enrichment remained strong after phylum-level fixed effects (OR = 6.98, p = 4.8e-61) and in a strict rhizosphere-only sensitivity analysis (OR = 10.6, q = 7.6e-38). 9

The PGP results contradict a simple generalization from accessory AMR patterns: all 13 PGP genes had higher core fractions than the genome-wide baseline of 46.8%, and the mean PGP accessory fraction was 29.7% versus 53.2% genome-wide. 9 Highlighted core fractions were pqqC 81.5%, pqqB 78.1%, hcnA 78.5%, ipdC 76.5%, acdS 70.4%, nifH 63.8%, and pqqD 55.5%; all had BH-FDR q < 0.05. 9 PGP gene richness correlated negatively with pangenome openness (Spearman ρ = −0.195, p = 2.0e-97, n = 11,272 species with ≥2 genomes). 9 pqqD was an exception, with 27.5% singleton content, suggesting the hypothesis that it can sometimes spread as a standalone gene; the functional pqqB–pqqC unit was predominantly core. 9

The plant-ecotype analysis supports and extends this core-retention pattern beyond the PGP marker set. Beneficial PGP gene clusters had a 64.6% core fraction, compared with 45.2% for pathogenic clusters and 46.8% genome-wide; the per-species Mann–Whitney U comparison gave U = 83,567,419 and p = 3.38e-125, with a bootstrap 95% confidence interval of [0.089, 0.106] for the beneficial-minus-pathogenic difference. 5 Biofilm genes had an 83.3% core fraction, IAA biosynthesis 78.1%, nitrogen fixation 72.3%, and phosphate solubilization 70.9%; T4SS was 48.5% singleton, coronate toxin 45.2% singleton, and effectors 39.8% singleton. 5 This supports more stable inheritance of beneficial functions after acquisition, but has a taxonomic-scope tension with a study of ten Pseudomonas strains in which biocontrol traits were largely accessory. 5

A refined 17-marker panel produced PERMANOVA pseudo-F = 23.2, p = 0.001, and R² = 0.071 across 607 root, rhizosphere, and phyllosphere species; db-RDA (distance-based redundancy analysis) location-only R² = 0.060, p = 0.001, while PERMDISP gave F = 15.6 and p = 0.001. 5 Compartment-associated functional separation is therefore real but weak, with approximately 6% of variance attributable to location shifts. 5 The earlier R² = 0.527 effect fell to R² = 0.072 after excluding three genome-rich species per compartment, indicating domination by taxonomic sampling and major rhizobial or Pseudomonas clades. 5

Of 5,671 eggNOG ortholog groups passing the 5% prevalence filter, 5,341 (94.2%) were significant in raw plant-versus-non-plant tests and 3,840 had OR > 2, which was interpreted mainly as broad taxonomic composition. 5 After phylum-level control, 50 families retained significance; COG3569 had Fisher OR = 8.92 and phylogenetically controlled OR = 6.01, q = 7.7e-242, and occurred in 54% of plant-associated species versus 12% of non-plant species. 5 All 50 had eggNOG descriptions and InterProScan domains, 48/50 had GO terms, and 39/50 had MetaCyc pathways; 19/50 were unknown-function COG category S, 7/50 involved electron transport and energy metabolism, 6/50 inorganic ion transport, 4/50 carbohydrate metabolism, and 3/50 secondary metabolite biosynthesis. 5 A per-species logistic regression converged for 48/50 families, with all 48 retaining positive plant-association coefficients, absolute coefficient > 0.2, and BH-FDR q < 0.05; controlled odds ratios for the top ten ranged from 4.4 to 6.6, with q-values reaching 10⁻⁹⁶. 5

The plant study refines GTDB-based gene-acquisition interpretation. Plant-associated species carrying singleton marker clusters were 16 times more likely to carry transposase or integrase singletons (Fisher OR = 15.95, p = 8.8e-20), but marker singletons had an overall enrichment ratio of 0.78 relative to the genomic average, with Wilcoxon p < 1e-300. 5 MGnify showed higher genus-level mobilome burden in plant-associated genera than non-plant genera, with median 3.7 versus 2.8 mobile elements per genome and Mann–Whitney p = 1.49×10⁻⁵. 5 This contrast suggests a scale-dependent pattern in which diverse accessory elements occur across plant-associated genera while useful functions may stabilize in adapted strains; direct GeNomad annotations are needed to test the hypothesis. 5

The Polyhydroxybutyrate Granule Formation Pathways — Distribution Across Clades and Environmental Selection supports and refines these atlas interpretations with a non-AMR storage pathway. Across 27,690 GTDB species, 21.9% carried phaC and 21.7% carried a complete PHB pathway defined as phaC plus phaA/phaB; 6,067 phaC-positive species spanned 95 of 142 GTDB phyla. 10 PHB prevalence was 60.9% in Pseudomonadota (4,544/7,456 species), 52.9% in Myxococcota (83/157), 34.4% in Halobacteriota (89/259), 27.9% in Thermoproteota (65/233), 21.9% in Desulfobacterota (70/319), and 18.5% in Actinomycetota (587/3,172); it was absent in Campylobacterota (0/271), Gemmatimonadota (0/102), Nanoarchaeota (0/135), and Marinisomatota (0/90). 10

PHB prevalence was 44.0% in plant-associated species, 43.6% in soil, 34.5% in wastewater/engineered environments, 25.5% in freshwater, 20.1% in sediment, 18.7% in marine environments, 11.1% in human-associated species, 7.4% in human clinical species, and 3.3% in animal-associated species. 10 A chi-squared test gave chi2 = 1,656.36, p ~ 0, dof = 2; after genome-size stratification, high- versus low-variability environments showed 4.4x, 4.6x, 3.1x, and 1.4x enrichment across genome-size quartiles Q1–Q4, with p = 1.18 x 10^-11, p = 3.25 x 10^-47, p = 4.17 x 10^-70, and p = 3.21 x 10^-14. 10 This supports phb-granule-ecology, but point-in-time NMDC correlations were modest (all |rho| < 0.12) and do not directly measure temporal variability. 10

Genome size, rather than PHB specifically, was the primary correlate of apparent niche breadth: PHB-positive and PHB-negative species had median AlphaEarth embedding variance of 0.3295 (n = 531) and 0.2472 (n = 1,477), with raw rho = 0.106 and p = 1.77 x 10^-06, but median genome sizes of 4.34 Mbp and 2.44 Mbp; genome size correlated with niche breadth at rho = 0.302 with p = 1.5e-43, becoming partial rho = -0.047 with p = 0.037 after control. 10 PHB core/accessory status included 5,371 species with phaC core and 1,959 with phaC accessory; 60.1% of 311 phylogenetically discordant phaC-positive species carried it as accessory versus 32.3% of all phaC-carrying species. 10 The inferred 311 potential acquisition and 278 potential loss events, including 38 putative acquisitions in Lachnospiraceae, remain hypotheses because no phaC gene tree was reconstructed. 10

Environmental HGT and transfer-context extension

The T4SS–CAZy Environmental HGT extends GTDB’s pangenome role from gene prevalence and ecological distribution to environmental horizontal gene transfer (HGT) involving type IV secretion systems (T4SS) and carbohydrate-active enzyme (CAZy) neighborhoods. Among 30,497 high-quality environmental metagenome-assembled genomes (MAGs), 6,652 (21.8%) carried T4SS or conjugative machinery under a multi-marker definition comprising VirB4/6/8/9/10/11, VirD4, TraI/D, TrwB, and TraG. 11 Ninety-two CAZy families showed elevated co-occurrence with T4SS loci at distances of ≤10 kb, although threshold validation remains pending; GT2 glycosyltransferases were the top hit, occurring in 767 genomes with an average length of 5,041 bp. 11

This supports and extends GTDB-based pangenome integration by associating gene neighborhoods with environmental contexts, but it refines the interpretation: the associations are observational and the synteny threshold still requires validation. T4SS–CAZy associations were enriched in marine sediment (OR=5.5, q<10⁻⁹⁸), barley rhizosphere (OR=10.4), and maize rhizosphere (OR=4.1). 11 A 70/30 discovery/validation split reproduced the enrichment patterns across both sets. 11

The GT2 gene tree contained 77 detected HGT events, including 32 normalized high-confidence cross-phylum events. 11 The strongest event, Node_4915, contained 35 genes, 82.9% syntenic, spanned 8 phyla—WOR-3, Desulfobacterota, Patescibacteria, Bacteroidota, Firmicutes_A, Methanobacteriota, Bdellovibrionota, and Acidobacteriota—and had Max_Divergence = 4.843. 11 Among all events, 65 spanned 2 phyla and 12 spanned ≥3 phyla (15.6%); divergence and synteny were negatively correlated (Spearman ρ = −0.615, p<0.001), with more distant events having lower syntenic percentage. 11 The most-involved phyla were Firmicutes_A (27 events), Pseudomonadota (22), Bacillota_A (19), and Actinomycetota (10). 11

These findings support the existing GTDB view that pangenome comparisons can reveal gene acquisition across phylogenetic backgrounds, while refining it by adding neighborhood-level evidence rather than species-level prevalence alone. In 376 genomes, the GT2 neighborhood was parsed as a list-format region: T4SS occurred in 503 neighborhood entries and GT2 in 495, confirming syntenic co-localization at the contig level. 11 GH23, a murein lytic transglycosylase family, was the second most common CAZy family in GT2 neighborhoods, with 106 occurrences, suggesting that cell-wall-remodeling genes cluster with GT2–T4SS syntenic loci. 11

The transfer-context result refines interpretation of GTDB mobile-element and resistance associations. CAZy genes were not detected on plasmids by ICEfinder; 12 integrative mobilizable elements (IMEs) occurred among the top 100 accumulators. 11 T4SS-positive genomes had 10× higher mobile genetic element (MGE) density than other genomes (p<0.001), consistent with chromosomal or integrative transfer rather than plasmid mobilization. 11 This supports the hypothesis that T4SS machinery mediates environmental dissemination of CAZy diversity across phylogenetically distant bacteria, but does not establish a mechanism experimentally. 11

The same analysis supports and extends environmental-resistome and metal-cross-resistance: GT2-neighborhood MAGs (n=376) had a mean of 0.045 metal-resistance types, compared with 0.004 for non-GT2 MAGs (n=260,276; Mann–Whitney p=8.6e-27), and genomes with GT2 in T4SS-proximal neighborhoods carried 11× more metal-resistance genes. 11 This links CAZy–T4SS synteny with metal-resistance breadth, but the report treats the relationship as a hypothesis about HGT hubs and niche breadth rather than a demonstrated causal pathway. 11

The HGT results also refine the existing pangenome caution that accessory status alone does not identify transfer route. The absence of plasmid-borne CAZy genes, together with elevated MGE density and IME observations, is consistent with chromosomal or integrative mechanisms involving IMEs or integrative conjugative elements rather than plasmid mobilization. 11 Pending validation includes a synteny-threshold permutation test requiring unfiltered Spark data, BLAST validation of Node_4915 against NCBI nr, a housekeeping-gene null baseline, and biome-enrichment factorization using θ = OR(T4SS-CAZy) / [OR(T4SS) × OR(CAZy)]. 11

Prophage and AMR integration

The Prophage-AMR Co-mobilization Atlas extends GTDB’s AMR pangenome role from prevalence and conservation to prophage-marker co-localization and species-level mobilome association. Across the full pangenome inventory, it identified 83,008 AMR gene clusters and 3,465,244 broad prophage marker clusters, including 1,261,929 strict prophage-marker clusters; prophage markers were 83.8% accessory and 53.8% singleton, while AMR genes were 69.7% accessory and 36.1% singleton. 12 Of 27,702 species, 14,669 (52.9%) carried both AMR and prophage markers. 12

Among 36,041 AMR gene instances from 1,953 genomes sampled across 100 species, 20,073 (55.7%) were on contigs carrying strict prophage markers; 12,026 (33.4%) were within 50 genes, 7,137 (19.8%) within 20 genes, 3,731 (10.4%) within 10 genes, and 1,991 (5.5%) within 5 genes of a prophage marker, with median nearest-marker distance 34 genes. 12 AMR genes within 10 genes were 67.6% accessory versus 65.5% for distal AMR; Fisher’s exact test gave odds ratio 1.10, one-sided p=0.005, and bootstrap 95% confidence interval [1.024, 1.185]. 12 Threshold-dependent odds ratios—0.78 at 3 genes, 0.92 at 5 genes, 1.10 at 10 genes, 1.19 at 15 genes, and 1.28 at 50 genes—together with only 33 of 74 testable species having species-level odds ratios above 1 and a median species-level odds ratio of 0.85, argue against a uniform immediate-neighborhood effect. 12

The strongest GTDB-scale result supports environmental-resistome: across 4,770 species, prophage marker density was positively associated with AMR repertoire breadth (Spearman rho=0.572, p<10^-300, n=4,770), with log-log slope 0.823 (SE=0.018), p<10^-300, and R²=0.30; a 10-fold increase in prophage density predicted a ~6.6-fold increase in AMR breadth. 12 After controlling for genome count, partial Spearman rho remained 0.464 with p=1.0×10^-253. 12 Associations were significant across Pseudomonadota (rho=0.54), Bacillota_A (rho=0.55), Bacillota (rho=0.40), Bacteroidota (rho=0.59), and Actinomycetota (rho=0.29). 12

This refines interpretations of both AMR mobility and phage-defense-syndromes-and-arms-race: the species-level correlation is compatible with direct phage mobilization or with high-recombination species independently accumulating prophages and AMR genes, but does not establish phage-mediated AMR transfer. 12 The attempted fitness comparison reinforces the GTDB-to-Fitness Browser coverage limitation: the KBase Data Lakehouse fitness browser contains RB-TnSeq data for only 48 model organisms, with poor overlap with GTDB pangenome species. 12

Other GTDB atlas and ecological integrations

The Lanthanide Methylotrophy Atlas — Summary supports and extends GTDB’s pangenome role beyond AMR to marker-based metabolic atlas construction. Among 293,059 genomes classified with GTDB-r214 taxonomy, there were 3,690 xoxF versus 195 mxaF genomes, a global xoxF:mxaF ratio of 18.92 : 1; equal weighting across 271 MDH-informative GTDB families gave an xoxF fraction of 0.960 [95% CI 0.941, 0.976], while a Bayesian binomial GLMM gave 0.993 [95% CI 0.992, 0.994], corresponding to a phylogeny-corrected ratio of ~143 : 1 [95% credible interval 122, 169]. 13

The BacDive Isolation Environment × Metal Tolerance Prediction extends GTDB to ecological validation of genome-derived metal-tolerance scores: species-name matching linked 42,227 of 97,334 BacDive strains (43.4%) to 6,426 GTDB species, with 25,089 matched strains retaining isolation metadata. 14 A separate phenotype-focused bridge linked 37,368 of 97,334 strains (38.4%) to a pangenome species and metal score, covering 5,647 GTDB species and 3,994 species with at least 5 phenotype features. 15

GTDB supports phylogenomics and gene-function acquisition analysis. The Gene Function Ecological Agora spans 18,989 species representatives, 13.74M rank × clade × KO producer/participation scores, 17,073,194 Sankoff-parsimony gain events, and 3.94M KO × genus mobile-genetic-element machinery records. 16 CRISPR-Cas showed a 24.5× recent-to-ancient gain ratio, TCS histidine kinases 10.3×, β-lactamases 9.0×, and clean tRNA-synthetase controls 2.3×. 16 The Mycobacteriaceae mycolic-acid producer signal was +0.31 at family rank and +0.394 in a mycolic-positive sub-clade, while mean leaf_consistency was 0.15. 16

The Conservation vs Fitness — Linking FB Genes to Pangenome Clusters extends GTDB-linked conservation analysis beyond AMR: 44 of 48 Fitness Browser organisms mapped, 33 were retained, and the Phase 1 table contained 177,863 gene-to-cluster links with 100.0% median protein identity and 94.2% median gene coverage. 17 Mapped genes were 82.0% core and 18.0% auxiliary, including 7,574 singleton genes within the auxiliary category. 17 Four organisms were unmatched because their species had too few GTDB genomes; Dyella79 was excluded for a locus-tag mismatch and ten organisms for <90% DIAMOND coverage. 17

The Fitness Modules × Pangenome Conservation supports and extends this conservation role from individual genes to co-regulated fitness modules. Across the 29/32 organism subset with pangenome links, module genes were 86.0% core versus 81.5% for all genes (OR=1.46, p=1.6e-87), a +4.5 percentage-point enrichment. 18 Among 974 modules with >=3 mapped genes, 577 (59%) were core modules, 349 (36%) mixed, and 48 (5%) accessory; the median module was 93.4% core. 18 Module-family breadth did not predict conservation (Spearman rho=-0.01, p=0.914), redirecting the signal toward gene-level conservation. 18 Thirty-eight families had <50% core genes, suggesting the hypothesis that some accessory co-regulated units may reflect horizontally transferred or niche-specific functions, although the analysis did not demonstrate either mechanism. 18

The Functional Dark Matter — Experimentally Prioritized Novel Genetic Systems extends and stress-tests GTDB-r214 integration for functional discovery. Full pangenome analysis queried 93.5M gene-cluster annotations and expanded conservation coverage from 32,791 (57.5%) to 37,997 (66.6%) of 57,011 dark genes by propagating ortholog-group identifiers, recovering 5,206 additional genes. 19 Among 11,774 root ortholog groups, 55.9% were kingdom-level, 11.0% class-level, 10.5% family-level, 6.9% genus-level, 6.5% mobile, 4.8% phylum-level, 3.9% order-level, and 0.5% species-level. 19 A conservation-weighted covering set selected 42 organisms covering 95.6% of importance-weighted priority across 28,584 high-priority dark genes. 19

GTDB also enables ecological and compound-to-genome placement. The ENIGMA Carbon Census — Tiered Knowledge Census of 83 Enrichment Compounds placed 494 strain records representing 359 distinct strains on GTDB taxonomy for 8 ENIGMA-isolate-callable compounds, using 64 high-certainty gene-complete Tier-2, 387 medium-certainty Tier-2, and 43 medium-certainty Tier-3 signature-enzyme records. 20 The contamination analysis bridged 1,392 observed genera to 530 mapped genera, producing 530 strict clades, 7,380 relaxed clades, and 150 species-proxy clades; 862 genera were unmapped and 380 mapped genera were ambiguous multi-clade genera. 21

The Genotype × Condition to Phenotype Prediction from ENIGMA Growth Curves extends GTDB’s role to genotype × condition phenotype modeling. The corpus contained 46,389 genome × condition pairs from 727 genomes and 363 conditions, with 4,293 shared kegg orthologs; GTDB-linked assignments supported 8 metabolic guilds across 20 taxonomic orders, but the guilds were aligned with rather than identical to taxonomy. 22 A short-identifier collision produced 1,751 spurious clinical genomes; checking CORAL brick 522 GTDB-Tk genus assignments reduced verified linkages from 32 to 20 and eliminated all false matches. 22 KO presence/absence predicted binary growth only moderately under genus-blocked holdout (overall AUC 0.620; AUC 0.653 with KO × condition interactions), whereas continuous growth kinetics had negative R² and could not be predicted across genera. 22

The plant analysis classified 292 species and 2,446 genomes as root-associated, 160 species and 941 genomes as rhizosphere-associated, 157 species and 1,334 genomes as phyllosphere-associated, 29 species and 291 genomes as endophytes, and 498 species and 2,983 genomes as plant-other. 5 Root communities were dominated by Rhizobium, Mesorhizobium, and Bradyrhizobium; Sphingomonas and Methylobacterium dominated the phyllosphere; and Pseudomonas_E was the only genus significantly represented across all compartments. 5 Host extraction identified 1,307 species across 19 testable hosts, including 286 rice-associated, 184 Arabidopsis thaliana-associated, 122 wheat-associated, and 98 maize-associated species. 5

Only 11.7% Jaccard overlap existed between 487 pangenome-derived plant genera and 438 MGnify rhizosphere genera, reflecting isolation metadata versus metagenomic detection rather than a direct estimate of ecological absence. 5 Among 65 plant-associated species with at least 20 genomes, only 18 had phylogenetic-tree distance data; of 17 testable species, 5 passed Bonferroni-corrected Fisher tests at α = 0.05/17 = 0.0029: Xanthomonas vasicola, p = 1.6×10⁻⁴; Mesorhizobium sp002294985, p = 1.1×10⁻⁴; Agrobacterium pusense, p = 4.5×10⁻⁴; Pseudomonas_E avellanae, p = 7.3×10⁻⁴; and Xanthomonas campestris, p = 1.0×10⁻³. 5 Xanthomonas campestris segregated toward Brassica hosts, with 46/47 genomes in one subclade and corrected p = 3.3×10⁻¹¹; Xanthomonas vasicola segregated toward Zea mays, with 47/52 genomes in one subclade and corrected p = 1.7×10⁻¹⁰. 5 These results provide weak support for within-species plant-association segregation, but the effect across all 65 candidates cannot be estimated until tree coverage improves. 5

Use in AMR pangenome and fitness studies

Core/accessory labels used a ≥95% prevalence threshold across GTDB genomes. 1 Fitness Browser species had a median of 9 GTDB genomes, ranging from 2–399; a gene present in all 9 sampled genomes could therefore be labeled core despite broader sampling potentially changing its status, so species with fewer than 20 genomes were treated cautiously. 1

The strain-variation analysis supports prevalence classes for conserved versus accessory AMR: among 37,444 AMR gene-species records, 51.3% were rare, 41.3% variable, and 7.5% fixed, with median variability index 0.526. 3 77.3% of Core AMR genes were fixed, 78.7% of Singletons rare, and Auxiliary genes 57.3% variable and 42.7% rare. 3

The conservation-versus-fitness analysis found that among 27,693 putative essential genes in 33 organisms, essential genes were 86.1% core versus 81.2% for non-essential genes, with median odds ratio 1.56; 18 of 33 organisms showed enrichment at BH-FDR q < 0.05. 17 Across 48 bacteria, 859 of 17,222 ortholog families (5.0%) were universally essential, 4,799 (27.9%) variably essential, and 11,564 (67.1%) never essential; universally essential genes were 91.7% core versus 80.7% for non-essential genes, while orphan essentials were 49.5% core. 23 Conservation and essentiality are related but not equivalent. 23

The AMR fitness study classified 638 AMR genes as core or intrinsic and 163 as accessory or acquired. 1 Core and accessory AMR had the same reported mean fitness, −0.024, Cohen’s d = 0.002, and Mann–Whitney U p = 0.33. 1 Mechanism was associated with conservation: 44% of metal-resistance genes were accessory versus 13% of efflux and 16% of enzymatic-inactivation genes (χ² = 69.3, p = 1.4×10⁻¹³). 1

The pan-bacterial atlas found that among 27,690 pangenome species, 53.2% had at least one AMRFinderPlus hit; AMR genes were less often core than the pangenome baseline (30.3% versus 46.8%, OR=0.49), and in 4,252 paired species 63.7% had AMR less core than their species baseline. 24 Beta-lactamases were 54.9% core, regulatory genes 6.5%, intrinsic emhABC efflux pumps >95%, and acquired efflux genes accessory. 24 Clinical AMR had 68% accessory AMR versus 43% in soil; resistance islands occurred in 705 species (54%), with 1,517 islands, mean phi=0.827, 88% containing multiple resistance mechanisms, and a maximum of 43 genes. 23

The prophage analysis refines these AMR conservation results: the 10-gene prophage-proximal accessory fraction was 67.6%, only 2.1 percentage points above distal AMR, and the species-level median odds ratio was 0.85, so proximity should not be treated as a universal marker of recently acquired AMR. 12 Conversely, the positive species-level prophage-density/AMR-breadth association provides a new mobilome-scale correlate of AMR repertoire breadth, while remaining correlational. 12

New ecological integration: metal-resistance diversity and niche breadth

The Metal Resistance Ecology — Phylogenetic Conservation vs. Environmental Selection extends and refines GTDB’s ecological AMR role by linking AMRFinderPlus pangenome annotations across 6,789 GTDB species to a 464,000-sample 16S amplicon atlas. 25 In 606 genera with AMR data, metal type diversity was positively associated with niche breadth after phylogenetic correction using PGLS (phylogenetic generalized least squares: β = +0.021, SE = 0.0056, p = 1.5×10⁻⁴), whereas total AMR cluster count (β = −0.002, SE = 0.0065, p = 0.789) and core AMR fraction (β = +0.006, SE = 0.0054, p = 0.288) were not significant. 25 The association remained after pangenome coverage, genome-size, and three-covariate controls, but strict 5% within-environment prevalence reduced the sample from 606 to 379 and yielded β = +0.0166, SE = 0.0099, p = 0.092. 25

This supports metal-cross-resistance by distinguishing breadth across metal types from depth represented by many genes for fewer metals, but refines causal interpretation: the raw OLS correlation was only r ≈ 0.21, and the cross-sectional analysis cannot distinguish habitat expansion caused by metal tolerance from acquisition opportunities or shared factors such as genome size, metabolic versatility, or biofilm capacity. 25 In 1,624 groundwater samples, metal type diversity correlated with groundwater prevalence (Spearman ρ = +0.112, p = 0.0019), while groundwater-specific fold-enrichment was null (ρ = +0.042, p = 0.242). 25 In the 133-sample ENIGMA well dataset, community-weighted metal-type diversity ranged from 1.01 to 1.83 and differed across 8 wells (Kruskal–Wallis H = 29.10, p = 0.0001); only 16.8% of reads were joinable to genus-level GTDB AMR data. 25

Data integrity, coverage, and interpretation limits

The pitfalls refines all GTDB-linked interpretations by showing that GTDB is not a stable identifier namespace independent of database version, access configuration, or join design. KBase Data Lakehouse is migrating collections from Delta to Iceberg: migrated tables use catalog.namespace.table, such as kbase.ke_pangenome.genome, while archived files may retain the flattened kbase_ke_pangenome.genome form. Migration remains incomplete, so live catalog discovery should prefer the dotted form when available and fall back to the underscore form only when necessary. 26 The data_lakehouse_ingest tenant is a MinIO governance-group name rather than a database prefix; the kbase_ke_pangenome database is under tenant kbase with dataset ke_pangenome, not tenant kbase_ke. 26

GTDB joins require explicit identifier validation. Short ENIGMA strain names are not globally unique: ENIGMA MT20 was Rhodanobacter glycinis, whereas GTDB’s MT20 was Streptococcus pneumoniae; the erroneous match involved 8,434 genomes, including 1,751 clinical genomes, and 12 of 32 pangenome linkages through ncbi_strain_identifiers were incorrect. 26 This supports and sharpens the genotype-to-phenotype warning: genus consistency must be checked and assembly accessions such as GCF_* used instead when possible. 26 MetaPhlAn3 cross-cohort analysis likewise requires a synonymy layer for reclassifications including Bacteroides vulgatus/Phocaeicola vulgatus, Eubacterium rectale/Agathobacter rectalis, and Ruminococcus gnavus/Mediterraneibacter gnavus; failure to reconcile names produced log₂FC approximately 28, equivalent to approximately 2.68 × 10⁸ fold, because one abundance collapsed to a pseudocount. 26

Pangenome taxonomy tables must be joined through genome_id, not gtdb_taxonomy_id, because the genome table stores a genus-level taxonomy string while gtdb_taxonomy_r214v1 stores a species-level string. 26 gapmind_pathways.clade_name uses the full gtdb_species_clade_id, including the representative accession, rather than the shorter GTDB_species value. 26 Pangenome gene-cluster identifiers are species-specific and cannot be compared directly across species; cross-species comparisons should use COG categories, KEGG orthologs, or Pfam domains. 26 eggnog_mapper_annotations.query_name, Bakta, and InterProScan annotations join to gene_cluster.gene_cluster_id, not gene.gene_id. 26

The ncbi_env table is entity-attribute-value data rather than a flat genome-metadata table: extraction requires joining genome.ncbi_biosample_id to ncbi_env.accession, filtering attributes, and pivoting the long format. 26 Per-genome NCBI environment classification yielded 52.7% unknown labels, or 94,957 of 180,025 genomes, while species-level majority-vote labels reached 91% coverage through keyword classification. 26

The Pseudomonas study adds a parallel sampling and phylogenetic caution. Isolation-source metadata were available for 64.2% of genomes (8,171/12,732 with classifiable sources), and keyword classification assigned 4,197 genomes to clinical sources, 1,659 to human (other), 566 to freshwater, 551 to soil, 503 to plant surface, 275 to rhizosphere, 141 to food/dairy, 136 to animal, 80 to industrial, and 63 to marine sources. 6 P. aeruginosa comprised 53% of all genomes (6,760/12,732), while many environmental species had fewer than 10 sequenced genomes; approximately 6.7% were unknown and approximately 29.1% other under the free-text classification. 6 Species-level majority voting can obscure genuinely generalist species, and the carbon analyses did not explicitly control for phylogenetic non-independence. 6 PGLS or phylogenetic logistic regression using the GTDB species tree is therefore needed before causal interpretation. 6

The plant study adds that only 7,995 of 293,059 genomes, or 2.7%, had plant-associated annotations, and labels were inferred primarily from variable-quality NCBI isolation_source metadata; the endophyte category contained only 29 species, and BacDive identified 2,482 plant-related strains but contributed zero additional genome upgrades beyond ncbi_env. 5 Marker presence does not establish expression or phenotype: PGP, pathogenic, secretion, motility, chemotaxis, and biofilm functions can have multiple ecological roles. 5 Genome size correlated with total marker count at r = 0.44, p < 1e-300; the refined dual-nature rate was 54% in the lowest genome-size quartile versus 87% in the top three quartiles, and approximately 33% of dual-nature species shifted to neutral after genome-size normalization in an earlier sensitivity analysis. 5

The refined marker scheme classified 878 of 1,115 plant-associated species (78.7%) as dual-nature, compared with 108 pathogenic (9.7%), 94 beneficial (8.4%), and 35 neutral (3.1%); the broader binary-presence scheme classified 15,474 of 25,660 species (60.3%) as dual-nature, while the composite scoring scheme classified 38.0% as neutral, 29.5% pathogenic, 25.2% dual-nature, and 7.3% beneficial. 5 In an 18-organism validation panel, all 14 known beneficial or pathogenic species were dual-nature and all four neutral controls were misclassified. 5 Cluster-robust GLM retained 8/14 marker associations at BH-FDR q < 0.05 with absolute coefficient > 0.2, but within-genus shuffling retained only nitrogen fixation, ACC deaminase, and T3SS; no multiple-testing correction was applied across the 15 markers, and Bonferroni α = 0.0033 would demote T3SS. 5

AlphaEarth embeddings covered 83,227 of 293,059 genomes, or 28.4%; a separate quality check found 3,838 of 83,287 genomes, or 4.6%, with at least one NaN among 64 dimensions, leaving 79,449 genomes after filtering. 26 Coverage depends on valid latitude and longitude, which are especially sparse for clinical isolates. 26 The pooled geographic distance–embedding distance curve had a 2.0x near-versus-far ratio, compared with 3.4x for environmental samples and 2.0x for human-associated samples; environmental analyses should therefore stratify by environment category or exclude human-associated samples when embeddings are used as environmental proxies. 26

Existing GTDB coverage cautions remain: only 6.8% (1,872/27,690) of species had sufficient AlphaEarth coverage for niche-breadth comparisons; AlphaEarth coverage was 1,457 / 3,690 = 39.5 % of xoxF genomes versus a 28 % pangenome baseline; BacDive matching linked 42,227 strains (43.4% of 97K) to 6,426 GTDB pangenome species; and GTDB species boundaries differ from LPSN/DSMZ taxonomy. 23,13 The PHB analysis covered 2,008 species, or 7.2% of total species diversity, with 34.9% classified as “other_unknown”; its NMDC bridge mapped 3,014/3,492 taxon columns (86.3%) and used a median 87.2% of taxonomic abundance. 10

GapMind score_simplified at sequence_scope = 'core' is binary, containing only 0.0 and 1.0, and should be aggregated with MAX(score_simplified) at species level; broader GapMind data has multiple rows per genome-pathway pair and requires selecting the best category before pathway aggregation. 26 The stored categories are aa and carbon, not amino_acid. 26 Pangenome core, auxiliary, and singleton flags are integer 0/1 values: core and auxiliary counts sum to total gene clusters, while singletons are a subset of auxiliary clusters; no_core + no_aux_genome = no_gene_clusters, with no_singleton_gene_clusters contained within no_aux_genome. 26

Fitness Browser data requires exact-case orgId, organism filtering because genefitness has approximately 27M rows, and explicit casting of string-valued fit and t. 26 KO mapping is a two-hop join through besthitkegg and keggmember; kgroupdesc uses desc, and experiment grouping uses expGroup. 26 Essential genes are absent from transposon fitness records: genefitness-only analyses miss approximately 14.3% of protein-coding genes, and genes absent from genefitness provide only an upper bound on essentiality. 26

GTDB-linked analyses must account for scale and computation. The pangenome contains approximately 1B-row gene and gene_genecluster_junction tables, approximately 421M-row genome_ani, approximately 93M-row eggnog_mapper_annotations, approximately 833M-row interproscan_domains, approximately 572M-row bakta_db_xrefs, and approximately 305M-row gapmind_pathways; key filters must precede joins and results should remain in Spark until the final small output. 26 Large .toPandas() collections can exceed the 1 GB serialized driver-result cap: a filtered result involving 218K Bacteroidota contigs produced more than 30M rows and approximately 1.5 GB serialized. 26 Disabling spark.sql.autoBroadcastJoinThreshold with -1 caused an NB10 job joining 13.7M rows with 18,989 species-taxonomy rows to hang for 17+ minutes; the optimizer should generally be trusted. 26

Phylogenetic tree distance tables use bare accessions such as GCA_001038305.1, whereas other pangenome tables use GB_ or RS_ prefixes; stripping the first three characters produced 100% overlap in tested species, including 399/399 for Koxy and 287/287 for Btheta. 26 Large ANI analyses are quadratic: Klebsiella pneumoniae has 14,240 genomes and Staphylococcus aureus has 14,526, so analyses should cap samples at <=500 genomes or subsample. 26

The plant analysis provides a direct example of this identifier hazard: phylogenetic tables used bare NCBI accessions while environmental metadata used GTDB-prefixed identifiers; adding GB_ to GCA_ accessions and RS_ to GCF_ accessions recovered 1,306/1,306 genomes and 599 plant-associated genomes. 5 Its annotation audit found 12 of 22 marker Pfams absent from bakta_pfam_domains despite abundance in InterProScan, with ten non-zero Bakta Pfams returning only 10–35% of InterProScan hit counts; versioned Pfam recovery yielded 19,364 hits across 7,962 species. 5 InterProScan was therefore used for secretion-system and cell-wall-degrading-enzyme detection, and Bakta Pfam data should be treated as a supplement rather than a substitute. 5

The NMDC bridge further refines GTDB community integration: classifier and metabolomics file_id namespaces do not overlap (nmdc:dobj-11-* versus nmdc:dobj-12-*) and must be bridged through sample_id using omics_files_table, which contains 385,562 rows. 26 NMDC taxonomy_features is a wide matrix with numeric taxon-ID columns rather than tidy sample identifiers; abiotic_features stores many unmeasured variables as 0.0 rather than NULL, so zeros should become NaN before analysis. 26 Approximately 1,352 Centrifuge taxa matched multiple GTDB clades within the same genus, with alphabetical tiebreaking selecting one representative; these ambiguous taxa accounted for approximately 6.5% of mapped abundance. 2

Finally, GTDB-linked results should be treated as versioned, provenance-dependent analyses rather than automatically executable historical instructions. Queries should begin with live catalog and schema discovery because schemas, namespace availability, access permissions, API behavior, database contents, and naming conventions can change. 26 Archived reports and notebooks are historical records; committed notebook history should be verified, and long analyses should use scripts, parquet checkpoints, and explicit logging rather than relying on inactive JupyterHub kernels. 26 The plant study likewise notes that its HGT evidence is indirect, its categorical marker definitions are assumption-dependent, its complementarity analysis loses 62/322 genera (19.3%) in the NMDC taxonomy bridge, and only 18/65 candidate species have tree coverage; expanded tree coverage, accessory-gene profiling, gene-content trees, direct mobile-element annotations, and transcriptomic validation are required. 5

Sources cited

  1. Fitness Cost of Antimicrobial Resistance Genesamr_fitness_cost · cited 7×
  2. Community Metabolic Ecology via NMDC × Pangenome Integrationnmdc_community_metabolic_ecology · cited 5×
  3. Within-Species AMR Strain Variationamr_strain_variation · cited 3×
  4. BERDL Data Atlas — Inventory, Topic Map, and Cross-Reference Synergiesberdl_data_atlas · cited 1×
  5. Plant Microbiome Ecotypes — Compartment-Specific Functional Guilds and Their Genetic Architectureplant_microbiome_ecotypes · cited 35× · primary source
  6. Carbon Source Utilization Predicts Ecology and Lifestyle in Pseudomonaspseudomonas_carbon_ecology · cited 16×
  7. Metabolic Capability vs Dependencypathway_capability_dependency · cited 13×
  8. Metabolic Capability vs Metabolic Dependencymetabolic_capability_dependency · cited 3×
  9. PGP Gene Distribution Across Environments & Pangenomespgp_pangenome_ecology · cited 10×
  10. Polyhydroxybutyrate Granule Formation Pathways — Distribution Across Clades and Environmental Selectionphb_granule_ecology · cited 9×
  11. T4SS–CAZy Environmental HGTt4ss_cazy_environmental_hgt · cited 17× · supports · refines
  12. Prophage-AMR Co-mobilization Atlasprophage_amr_comobilization · cited 12×
  13. Lanthanide Methylotrophy Atlas — Summarylanthanide_methylotrophy_atlas · cited 2×
  14. BacDive Isolation Environment × Metal Tolerance Predictionbacdive_metal_validation · cited 1×
  15. BacDive Phenotype Signatures of Metal Tolerancebacdive_phenotype_metal_tolerance · cited 1×
  16. Gene Function Ecological Agoragene_function_ecological_agora · cited 3×
  17. Conservation vs Fitness — Linking FB Genes to Pangenome Clustersconservation_vs_fitness · cited 4×
  18. Fitness Modules × Pangenome Conservationmodule_conservation · cited 4×
  19. Functional Dark Matter — Experimentally Prioritized Novel Genetic Systemsfunctional_dark_matter · cited 3×
  20. ENIGMA Carbon Census — Tiered Knowledge Census of 83 Enrichment Compoundsenigma_carbon_census_1 · cited 1×
  21. Contamination Gradient vs Functional Potential in ENIGMA Communitiesenigma_contamination_functional_potential · cited 1×
  22. Genotype × Condition to Phenotype Prediction from ENIGMA Growth Curvesgenotype_to_phenotype_enigma · cited 3×
  23. Discoveries Logdiscoveries · cited 4× · cross-project digest
  24. Pan-Bacterial AMR Gene Landscapeamr_pangenome_atlas · cited 2×
  25. Metal Resistance Ecology — Phylogenetic Conservation vs. Environmental Selectionmicrobeatlas_metal_ecology · cited 6×
  26. KBase Data Lakehouse: Common Pitfalls & Gotchaspitfalls · cited 29× · cross-project digest