KBase KE Pangenome

What this is

Canonical name: KBase KE Pangenome. The database collection identifier is kbase_ke_pangenome. ₁

Known aliases: kbase_ke_pangenome; no other aliases were reported. ₁

Stable external identifier: No stable external identifier was reported. ₁

KBase KE Pangenome is a KBase Data Lakehouse pangenome dataset connected to acinetobacter-baylyi-adp1 and used with kbase-msd-biochemistry and kescience-fitnessbrowser to connect genome, pangenome, biochemical, and fitness data. ₁ AlphaFold analysis combined kbase_ke_pangenome, bakta_annotations, interproscan, and kescience-alphafold to relate pangenome gene clusters to sequence annotation and structural-representation measures. ₂

The BERDL Data Atlas inventories 1,011,650,903 KBase pangenome gene rows spanning 293,059 genomes, 27,690 GTDB species clades or pangenomes, and 132.5M gene clusters. ₃ KBase appears in 53/66 audited BERIL projects (80%), covers 10 topics, and has topic entropy 2.87; the atlas describes it as the most evenly cross-topic tenant. ₃ The realized kbase × kescience bridge accounts for 36 cross-tenant projects, mostly pangenome-by-fitness joins through genome_id and ncbi_taxon_id. ₃ These inventory counts are complementary to, rather than replacements for, ADP1-scale analyses; row totals are generally COUNT(*), and join-key presence establishes schema compatibility but not value-space overlap. ₃

The Web of Microbes (WoM) analysis supports and extends this integration role: all WoM genera had pangenome species clades, with reported top-five clade counts and genome totals of Bacillus (5 species, 2,557 genomes), Rhizobium (5, 449), Pseudomonas fluorescens (5, 139), Synechococcus (5, 88), Phenylobacterium (5, 80), Acidovorax (5, 79), Zymomonas mobilis (1, 26), and Escherichia coli (1, 2). ₄ Genus-level gene clusters, conservation, and functional annotations were available for every WoM organism genus, whereas species-level matching required strain-to-genome mapping that was not attempted. ₄ The same study proposes comparing the WoM de novo-production fraction E/(E+I) with pangenome openness and accessory genes; this is a testable hypothesis, not an established relationship. ₄

The operational evidence refines this scale assessment: KBase Data Lakehouse is migrating collections from Delta to Iceberg, so the live collection may be addressed as kbase.ke_pangenome.genome rather than the former kbase_ke_pangenome.genome; migration is incomplete and code should discover the live address, preferring the dotted form and falling back to the underscore form. ₅ The kbase tenant contains the ke_pangenome dataset; treating data_lakehouse_ingest as a database prefix or using tenant kbase_ke causes access or namespace errors. ₅ Live catalog and schema discovery should precede reuse of archived SQL, and access messages should identify the unreachable table and tenant without exposing internal service strings. ₅

Pangenome-wide functional and ecological integration

The KBase pangenome provided the linkage layer for distinguishing annotation lag from persistent hypothetical genes. Of 57,011 Fitness Browser dark genes, 39,532 had pangenome links; Bakta reclassified 33,105 (83.7%), while 6,427 remained hypothetical in both pipelines and 17,479 lacked pangenome links. ₆ This supports and refines the dataset’s role in annotation integration: pangenome linkage enables reannotation and comparative evidence, but the 17,479-gene linkage gap limits inference about the full dark-gene population. ₆ Among truly dark genes, only 29.3% had orthologs, compared with 63.7% of annotation-lag genes; their median ortholog breadth was 1 versus 4 organisms. ₆ These findings support genomic-under-representation and gene-function-acquisition-depth while showing that sequence-database recognition does not necessarily yield functional assignment: 79.4% had UniRef50 links and 84.7% had database cross-references, but only 4.0% had Pfam hits and 4.6% had KEGG KOs. ₆ See Truly Dark Genes — What Remains Unknown After Modern Annotation? for the full clue-tier census and candidate ranking.

The pan-bacterial anti-phage analysis supports and extends this scale assessment: defense markers were detected in 27,626 of 27,690 species-level pangenomes (99.8%), producing 930,573 marker hits across 825,476 unique gene clusters. Its quality-filtered analysis used 7,323 species with at least 5 genomes. Defense-system count correlated with prophage-cluster burden (Spearman ρ = 0.609, p ≈ 0, n = 7,323); after residualizing on log10 median genome size and phylum, partial ρ remained 0.301 (p = 1.6 × 10⁻¹⁵³). This is consistent with coordinated defense architecture, but did not formally establish a phylogenetically independent arms race. ₇

Six of seven defense systems were enriched in auxiliary and singleton gene clusters, with singleton fractions of 55.8–68.8% versus a 37.9% genome-wide background; singleton status is a subset of auxiliary status. DISARM was near baseline and unreliable because its PF00176 anchor is a widespread SNF2 helicase domain. CRISPR-Cas prevalence was 96% with permissive EggNOG description matching but approximately 55% using Cas1 Pfam PF01867. The primary Pfam source was the 833-million-row interproscan_domains table rather than the 18.8-million-row bakta_pfam_domains table; InterProScan returned ~500K rows after filtering approximately 20 curated accessions, while bakta_pfam_domains contained 0 Cas1 PF01867 hits compared with ~25K in InterProScan. ₇

The lanthanide methylotrophy atlas supports and extends this scale assessment by analyzing all 293,059 genomes with GTDB-r214 taxonomy, eggNOG, Bakta, and environmental metadata. Among joint marker calls, xoxF occurred in 3,690 genomes and mxaF in 195, giving an xoxF:mxaF ratio of 18.92 : 1 and an xoxF fraction of 0.9498 (95% CI [0.9425, 0.9558]); a one-sided binomial test against 10/11 gave p = 7.6 × 10⁻²². Annotation-based detection does not establish enzyme activity or methylotrophic physiology. ₈

The PGP analysis supports and broadens this ecological use: among 27,702 species, 11,272 carried at least one of 13 PGP markers, with 32,736 PGP gene clusters, 27,690 species with gapmind pathway scores, and 291,279 genomes with isolation-source metadata. A putative non-diazotrophic pqqC + acdS rhizosphere module had OR = 7.24, n = 286 co-occurring species, and q = 1.2e-83; acdS prevalence was 15.8% in soil/rhizosphere species versus 2.6% elsewhere (OR = 7.02, q = 5.1e-62). pqqC, acdS, and hcnC were soil-enriched, whereas nifH was depleted in soil-classified species. ₉

All 13 PGP genes had higher core fractions than the 46.8% genome-wide baseline; mean PGP accessory fraction was 29.7% versus 53.2% genome-wide, and PGP gene richness correlated negatively with pangenome openness (Spearman ρ = −0.195, p = 2.0e-97, n = 11,272 species with ≥2 genomes). pqqD was a partial exception, with 55.5% core and 27.5% singleton. Using GapMind completeness and SEED subsystem annotations as a pathway-membership proxy, ipdC occurred in 2.5% of species with a complete tryptophan pathway versus 0.9% with an incomplete pathway (OR = 2.81, p = 6.3e-10); a similar tyrosine association and a soil-stratified reversal make the mechanism hypothesis-generating. ₉

The Pseudomonas carbon-ecology analysis supports and refines this integration: it extracted standardized GapMind carbon-pathway predictions for 12,732 genomes across 433 Pseudomonas species clades, producing a 789,012-row genome-level carbon-pathway table. Among 7 Pseudomonas s.s. species and 189 Pseudomonas_E species with at least 5 genomes, 43 of 62 pathways differed significantly after Mann-Whitney U testing with Benjamini-Hochberg false-discovery-rate correction. The strongest differences involved xylose, ribose, arabinose, galacturonate, myo-inositol, mannitol, and sorbitol, while amino-acid and core organic-acid pathways remained near-universal. ₁₀

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 in a 999-permutation test (p = 0.006); between-group mean distance was 2.054 versus within-group mean distance of 1.890. A Random Forest classifier achieved balanced accuracy of 0.408 +/- 0.169 against a 0.250 chance baseline. The dominant PCA separation followed the division between Pseudomonas s.s. and Pseudomonas_E, so phylogenetic confounding remains central. ₁₀

Soil-metal functional genomics

The soil analysis supports and extends the environmental use of this dataset by linking 51,748 soil samples and nine metals—copper, cobalt, chromium, nickel, zinc, lead, arsenic, cadmium, and mercury—to microbial COG profiles. COG denotes a Cluster of Orthologous Groups functional category. It identified 2,355 significant COG–metal associations at FDR < 0.05 across 3,915 implied tests; chromium and lead produced the strongest signals, and transporters including ABC and RND systems plus biosynthesis genes dominated the top hits. ₁₁

The same analysis found 116 significant copper-associated COGs using 7,566 samples with nearby KBase genomes within 10 km. Positive associations included cell division and nucleotide transport categories BQ, FQ, and FK, while negative associations included energy-production categories DI and CE. These observational results refine rather than establish the hypothesis of energetic trade-offs under copper stress. A db-RDA model—distance-based redundancy analysis—gave R² = 0.799 and p = 0.005 with 999 permutations; after conditioning on batch and project effects, metal concentrations explained 80% of variance in community COG profiles. ₁₁ This result refines the dataset’s environmental-association claims because the reported R² is conditional: unconditional metal-only R² was not reported and may be substantially lower. Biome-stratified PGLS—phylogenetic generalized least squares—found distinct metal–COG relationships in soil, marine, and wastewater environments, supporting environment-specific responses rather than a universal resistance programme. ₁₁

Interpretation remains constrained by co-contamination: chromium, copper, lead, and zinc co-vary in many industrial soils, so individual associations may represent generic multi-metal stress rather than metal specificity. The 2,355 discoveries among 3,915 tests represent a 60% discovery rate, but correlated tests may make Benjamini–Hochberg FDR anti-conservative and the true FDR higher than reported. Effect sizes were not systematically reported; planned validation will audit Spearman rho values, flag rho < 0.05, test partial COG ~ Cr | Cu + Zn + Pb models, assess Moran’s I and SEVM where residual spatial autocorrelation is present, report unconditional db-RDA R², and repeat copper matching at 5 km and 20 km. These analyses require rerunning kescience_mgnify and kbase_ke_pangenome Spark tables because no local CSV of rho values or model residuals is available. ₁₁

Pathway capability and community metabolism

The metabolic-capability analysis extends dataset integration to pathway completeness, fitness dependence, pangenome openness, and environmental metadata. It used kbase_ke_pangenome.gapmind_pathways, pangenome, and gtdb_metadata with Fitness Browser genefitness, gene, and seedannotation tables across 1,695 pathway-organism pairs from 48 organisms. Outputs included 23,424,480 per-genome pathway records, 27,690 species pathway summaries, 3,065 pathway-fitness aggregates, 1,695 classifications, 1,511 conservation records, 7,334 pangenome-openness records, 800 pathway-heterogeneity records, 12,018 metabolic ecotype assignments, 10 ecotype-cluster summaries, and 370 pathway-cluster signatures. ₁₂

Using GapMind completeness and SEED subsystem annotations as a pathway-membership proxy, 267 of 1,695 complete pathway-organism pairs (15.8%) were latent capabilities, 547 (32.3%) intermediate, and 881 (51.9%) active dependencies. Carbon-source pathways were latent in 217 of 892 cases (24.3%), compared with 48 of 735 amino-acid biosynthesis cases (6.5%); the reported fractions make carbon pathways 3.7× more likely to be latent. This supports condition-specific-fitness, because genomic completeness does not establish laboratory fitness dependence. ₁₂

The Pseudomonas study refines this result: across species with at least 5 genomes, free-living and plant-associated 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. Carbon profiles alone therefore remain insufficient for fine-grained ecological prediction. ₁₀

The pitfalls evidence refines pathway interpretation: at sequence_scope = 'core', GapMind score_simplified is binary, containing only 0.0 and 1.0, and should be aggregated with MAX(score_simplified) at species level; broader GapMind data contain multiple genome–pathway rows whose best category must be selected first. Stored categories are aa and carbon, not amino_acid. Deoxyribonate and myoinositol lacked matching SEED subsystem role descriptions; full-name phenylalanine and tyrosine were recovered through phe and tyr, while alanine was excluded because fewer than 3 SEED-annotated genes met the minimum coverage threshold. SEED-proxy mapping can create false positives; direct GapMind per-step gene assignments would improve precision. _5,12

The NMDC analysis supports and broadens this use from organism records to community-weighted metabolic potential. It combined 305M GapMind pathway records with NMDC community taxonomy and metabolomics across 220 samples, 27,690 GTDB species, and 80 pathways. Mean taxonomy-bridge coverage was 94.6%; all 220 samples passed the 30% quality-control threshold, and 92% mapped at least 85% of community abundance to GTDB pangenome species. The analysis-ready merged matrix contained 174 samples, and the pathway–metabolomics matrix contained 131 samples; all 33 Freshwater samples lacked paired metabolomics. ₁₃

Across 13 testable amino-acid biosynthesis pathways, 11 of 13 (85%) had negative Spearman correlations, with binomial sign-test p = 0.011. Leucine biosynthesis had r = −0.390, q = 0.022, n = 62, and arginine biosynthesis had r = −0.297, q = 0.049, n = 80. These are weak support for community-scale Black Queen Hypothesis dynamics, not evidence that GapMind presence predicts expression or fitness. PCA of the 220-sample × 80-pathway matrix explained 49.4% of variance in PC1 and 16.6% in PC2, with PC1–5 explaining 83.0%; Soil and Freshwater separated with Mann-Whitney U = 3,674, p < 0.0001, and median PC1 values of +3.86 and −6.28. Seventeen of 18 amino-acid pathways differed by ecosystem after correction, led by glycine (H = 92.98, q = 1.2×10⁻¹⁹), asparagine (H = 66.19, q = 3.8×10⁻¹⁴), and cysteine (H = 62.66, q = 1.5×10⁻¹³); tyrosine was not differentiated (q = 0.71). ₁₃

Conservation, fitness, and modules

Across 194,216 protein-coding genes from 43 bacteria, essential genes were 82% core and always-neutral genes 66% core. Across 48 bacteria, 15 gene families were essential in every organism; of 17,222 ortholog families, 859 (5.0%) were universally essential, 4,799 (27.9%) variably essential, and 11,564 (67.1%) never essential. _14,15 A related analysis mapped 44 of 48 Fitness Browser organisms to pangenome species clades, retained 33, and found 145,821 core genes (82.0%) and 32,042 auxiliary genes (18.0%); 27,693 of 148,826 protein-coding genes (18.6%) were putatively essential, with essential genes 86.1% core versus 81.2% for non-essential genes. ₁₆

The truly dark-gene analysis refines this broad conservation–essentiality relationship: truly dark genes had a core-genome fraction of 43.1% versus 72.7% for annotation-lag genes, yet an essential fraction of 18.0% versus 13.4%; the corresponding core-fraction OR was 0.284 and essential-fraction OR was 1.420, with p-values < 1e-100 and < 1e-10, respectively. This supports prioritizing unannotated genes with fitness evidence rather than treating accessory status as evidence of dispensability. ₆

The Pan-Bacterial Metal Fitness Atlas extends these joins to metal-associated fitness across 559 experiments, 31 organisms, and 16 metals, yielding 383,349 gene × metal fitness records. Across 22 organisms and 14 metals, metal-important genes were 87.4% core versus 76.9% for baseline genes (OR=2.08, p=4.3e-162). Essential-metal tolerance genes had a mean core-fraction delta of +0.148 versus +0.081 for toxic metals (one-sided Mann-Whitney U = 39, p = 0.015). The atlas identified 1,182 ortholog groups with metal phenotypes conserved across at least two organisms and 601 across at least three; 149 conserved candidates lacked full functional annotation. ₁₇

Fitness modules extend this integration to process-level structure. Across 32 organisms, ICA (independent component analysis, a statistical decomposition method) identified 1,116 stable modules; 94.2% had elevated within-module cofitness, with mean within-module |r| 0.34 versus 0.12 in the background and 22.7x genomic-adjacency enrichment. It found 156 module families spanning at least 2 organisms, including 28 spanning 5+, 7 spanning 10+, and 1 spanning 21 organisms. Ortholog transfer achieved 95.8% strict precision, 91.2% coverage, and 0.934 F1, whereas Module-ICA and cofitness voting each had <1% strict precision. ₁₈

The truly dark-gene study supports this module-based interpretation: among 4,394 truly dark genes in organisms with integrated condition analysis data, 25.9% showed operon-like cofitness with adjacent genes using |r| ≥ 0.3, and 594 belonged to ICA fitness modules. Guilt-by-association linked modules to candidate phage integrase, metal-transporter, chemotaxis, and iron-regulation functions. These are candidate functional hypotheses, not direct functional assignments. ₆

Module genes were 86.0% core versus 81.5% for all genes (OR=1.46, p=1.6e-87); among 974 modules with >=3 mapped genes, 577 (59%) were core, 349 (36%) mixed, and 48 (5%) accessory, with median core fraction 93.4%. Family breadth did not predict conservation (Spearman rho=-0.01, p=0.914). Zero essential genes appeared in ICA modules because essential genes lack usable transposon-insertion fitness variation. ₁₉

Other analyses found 5,526 genes costly in laboratory measurements but dispensable in the pangenome; they were 7.45x more likely than costly+conserved genes to contain mobile-element keywords (OR=7.45, p=4.6e-71). Core genes were 1.78x more likely to have strong condition-specific phenotypes and 1.29x more likely to be important in some conditions yet burdensome in others. The core-tradeoff analysis identified 25,271 true trade-off genes (17.8%), with core enrichment of 1.29 (p=1.2e-44); its matrix contained 28,017 Costly + Conserved, 5,526 Costly + Dispensable, 86,761 Neutral + Conserved, and 21,886 Neutral + Dispensable genes. _20,14,21

The operational evidence refines Fitness Browser interpretation: fit and t are often strings and require explicit casting; orgId is case-sensitive, genefitness should be filtered by organism, and KO mapping is a two-hop join through besthitkegg and keggmember, with experiment grouping in expGroup. Essential genes are absent from transposon fitness records, so genefitness-only analyses miss approximately 14.3% of protein-coding genes and absence from genefitness supplies only an upper bound on essentiality. ₅

Annotation, environmental, and AMR integration

The annotation-gap study integrated pangenome conservation with metabolic-model gapfilling, GapMind, Fitness Browser phenotypes, Bakta, and BLAST across 14 organisms and 18 carbon sources. Of 201 gapfilled enzymatic reaction-organism pairs, 96 (47.8%) received candidate genes—44 (21.9%) high-confidence, 19 (9.5%) medium-confidence, and 33 (16.4%) low-confidence—while 105 (52.2%) remained unresolved. The full pipeline reached 47.8% resolution versus 42.8% without NB03 EC matching, 39.8% without NB04 Bakta, and 36.3% without NB06 BLAST. ₂₂

Using ModelSEED/RAST and COBRApy, the study constructed models for 574 organism-carbon-source combinations. Baseline FBA (flux-balance analysis, a constraint-based method for predicting growth and flux) achieved 42.5% accuracy, recall 86.5% (244 of 282 growth-positive conditions), precision 42.5% (244 of 574), and 330 false positives. Conditional gapfilling added 219 reactions to 38 false negatives; 201 were enzymatic, 14 transport, and 12 exchange. Of 201 reactions, 50 (24.9%) were EC-less dark reactions and only 8 (16%) were resolved, versus 88 of 151 (58.3%) with known EC numbers. ₂₂

Across 32 species, 9 phyla, and 357,623 genes, COG analysis found novel or singleton enrichment in COG L (+10.88%), V (+2.83%), and S (+1.64%), while core genes were depleted in J (-4.65%), F (-2.09%), H (-2.06%), E (-1.81%), and C (-1.75%). Across 132.5M clusters, eggNOG had higher COG coverage (51% versus 8.2%), KEGG coverage (38.5% versus 17.3%), and Pfam coverage (63% versus 7.7%) than Bakta, while Bakta had higher GO coverage (15% versus 7.4%), product descriptions (71.2% versus 70.4%), and unique UniRef50 links covering 79.2%. Their union reached 77.3% any-functional-annotation coverage; Bakta rescued 11.2M clusters among 39.2M missed by eggNOG. _23,15

The truly dark-gene analysis refines these aggregate annotation-coverage results by showing that sequence recognition and functional annotation diverge at the residual tail: among 5,870 unique truly dark gene clusters, Pfam provided hits for 235 clusters (4.0%), eggNOG provided annotations for 2,551 clusters (43.5%), and only 17.0% had COG categories; 55.4% of those COG assignments were category S, meaning function unknown. The result is consistent with the dataset’s use for identifying annotation gaps, but Bakta hypothetical calls can be false negatives because the tested PSC database version may lack matching known functions. ₆

The environmental embedding analysis found 83,286 genomes with coordinates, while the Pseudomonas study found statistically detectable but weak environmental pathway structure: p = 0.006, with Random Forest balanced accuracy 0.408 +/- 0.169 versus a 0.250 chance baseline. _24,10 This refines interpretation of both embedding and GDI results: ecological signal can be detectable without supporting reliable fine-grained prediction. ₁₀

The soil-metal analysis similarly supports but qualifies environmental gene-content integration: biome-specific metal–COG relationships were detected, but co-contamination, conditional model fit, spatial autocorrelation, and 10 km genome–soil matching prevent treating the associations as universal or causal. ₁₁ This directly feeds environmental-resistome, ecotype-environment-gene-content, metal-cross-resistance, and environment-embedding-geography.

The AMR cofitness analysis covered 28 organisms, with InterProScan coverage of 68% of genes, 3.6× better than old SEED annotations; 35/3,193 significant GO-enrichment tests passed FDR < 0.05 versus 0/280 with old SEED/KEGG. Of 801 AMR genes with fitness data, 136 conserved module families were connected; 208/209 (99%) assignments were in cross-organism conserved families. ₂₅

The AMR fitness-cost analysis covered 1,352 AMR genes across 43 organisms and 6,804 experiments. Non-antibiotic AMR knockouts had pooled random-effects relative-fitness shift +0.086 [95% CI: +0.074, +0.098] across 25 organisms; under antibiotics, 57.0% flipped toward greater importance (N = 797; mean flip +0.045; Wilcoxon p = 0.0001). ₂₆ The environmental-resistome analysis covered 14,723 bacterial species and 293K genomes; clinical species had median 5 AMR clusters versus 2 for soil, aquatic, and host-associated species. ₂₇ The pan-bacterial atlas identified 83,008 amrfinderplus hits, 82,908 distinct clusters, 1,939 AMR families, 2,079 products, and 14,723 species; AMR was 30.3% core versus 46.8% baseline (OR=0.49, chi-squared=23,117, p≈0), with auxiliary AMR 2.2x enriched. ₂₈

Data integrity and limitations

Pangenome gene-cluster identifiers are species-specific and require shared representations such as COG categories, KEGG orthologs, or Pfam domains; eggnog_mapper_annotations.query_name joins to gene_cluster.gene_cluster_id, as do Bakta and InterProScan annotations. ₅ Pangenome taxonomy tables should be joined through genome_id, not gtdb_taxonomy_id, and GapMind clade_name uses the full gtdb_species_clade_id, including the representative accession. ₅

Environmental analyses remain constrained by 28.4% AlphaEarth coverage, missing or imprecise coordinates, human-associated sampling bias, incomplete metadata, and absent phylogenetic controls. _29,30,31 The soil-metal analysis adds unresolved co-contamination, conditional versus unconditional variance explanation, effect-size uncertainty, spatial dependence, and proximity-matching sensitivity. ₁₁

Many KBase Data Lakehouse numeric fields are strings and require explicit casts; Spark DECIMAL values should be cast to DOUBLE or converted to float after collection. ₅ Billion-row tables include gene and gene_genecluster_junction at approximately 1B rows, genome_ani at approximately 421M, eggnog_mapper_annotations at approximately 93M, interproscan_domains at approximately 833M, bakta_db_xrefs at approximately 572M, and gapmind_pathways at approximately 305M. ₅ Key filters should precede joins and results remain in Spark until final small output; a 1 GB Spark Connect driver result-size cap makes large .toPandas() collections unsafe, with one filtered result exceeding 30M rows and approximately 1.5 GB serialized and a pandas spatial merge exceeding 9 GB of working memory. ₅

Pangenome core, auxiliary, and singleton flags are integer 0/1 values; core plus auxiliary counts equal total gene clusters, while singletons are a subset of auxiliary clusters. ₅ Phylogenetic tree distances 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. ₅ 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. ₅

The WoM integration adds a distinct coverage limitation: its 2018 snapshot contained 37 organisms across 5 ENIGMA-funded projects, including 20 experimental organisms, and genus-level pangenome representation did not resolve strain-to-genome matching. ₄ The proposed metabolic-novelty comparison therefore remains future work, and the archived WoM snapshot may not represent current WoM coverage. ₄

The truly dark-gene analysis adds several interpretation limits. Short genes are harder to annotate and measure because they contain fewer domains, fewer homologs, and fewer transposon-insertion sites; consequently, the length effect (d = −0.432) may partly explain other differences. GC deviation is an imperfect HGT proxy, and mobile-element proximity, accessory localization, and narrow taxonomic breadth support but do not establish horizontal gene transfer or gene function. Some strong fitness phenotypes may reflect polar effects on downstream genes. Ortholog concordance covered only 32 of 48 Fitness Browser organisms, and the 17,479 unlinked dark genes could not be directly assessed. ₆

Large KBase Data Lakehouse catalogs should be discovered through access-aware helpers such as get_databases(), get_tables(), and get_table_schema() rather than historical inventories. ₁₅ Direct Spark SQL is preferred for complex or large queries because REST /count loops and /schema requests can time out or return 504, 524, 503, or empty responses. ₅ The soil-metal validation specifically requires live Spark access to kescience_mgnify and kbase_ke_pangenome; no local CSV contains the needed Spearman rho values or model residuals. ₁₁

Related pages

• acinetobacter-baylyi-adp1 — organism represented in the explored pangenome database.
• eggnog — COG and orthologous-group annotation resource.
• interproscan — domain-annotation resource.
• kescience-alphafold — AlphaFold MSA-depth resource.
• metal-fitness-atlas — genome-based metal-tolerance score resource.
• pangenome-integration — cross-dataset pangenome integration and annotation coverage.
• genomic-under-representation — persistent functional annotation gaps and truly dark genes.
• gene-function-acquisition-depth — conservation, ortholog breadth, and accessory-genome signals.
• metabolic-model-gapfilling — metabolic gapfilling and evidence-triangulated reaction annotation.
• ecotype-environment-gene-content — environmental association of gene content and PGP traits.
• environment-embedding-geography — embedding, geography, and environmental association.
• environmental-resistome — ecological and within-species AMR organization.
• metal-cross-resistance — metal-specific versus multi-metal functional responses.
• phage-defense-syndromes-and-arms-race — prophage, defense-system, and phage-associated AMR context.
• cofitness-network-architecture — resistance-island and fitness-network structure.
• condition-specific-fitness — AMR fitness burden and antibiotic-dependent importance.
• multi-omics-integration — integrated pangenome, structural, domain, fitness, and environmental evidence.
• Truly Dark Genes — What Remains Unknown After Modern Annotation? — residual dark genes after Bakta reannotation, clue tiers, genomic context, and prioritized experimental candidates.
• Soil Metal Concentrations Drive Functional Gene Shifts — soil-metal concentrations, COG shifts, environmental stratification, and validation requirements.
• discoveries — cross-project discoveries log covering pangenome scale, annotation, essentiality, fitness, and data-quality findings.
• pitfalls — operational, provenance, schema, statistical-design, and reproducibility safeguards for KBase Data Lakehouse analyses.
• Pan-Bacterial Metal Fitness Atlas — cross-species metal fitness, core-genome conservation, metal-responsive modules, and pangenome-scale prediction.
• Ecotype Correlation Analysis — environmental versus phylogenetic prediction of pangenome gene-content similarity.
• AlphaEarth Embeddings, Geography & Environment Explorer — AlphaEarth embedding, geography, environment harmonization, and coordinate-quality analysis.
• Environmental Resistome at Pangenome Scale — environmental distribution of AMR functions across the pangenome.
• Pangenome Openness Analysis — pangenome openness and accessory-gene structure.
• PGP Gene Distribution Across Environments & Pangenomes — PGP gene distribution, environmental enrichment, co-occurrence, and core/accessory status.
• Community Metabolic Ecology via NMDC × Pangenome Integration — NMDC community metabolic ecology, GapMind pathway completeness, metabolomics, and ecosystem differentiation.
• Pan-bacterial Fitness Modules via Independent Component Analysis — ICA-derived pan-bacterial fitness modules, cross-organism module families, and process-level function prediction.
• Prophage-AMR Co-mobilization Atlas — pangenome-scale prophage-marker and AMR co-localization, repertoire breadth, and mobilization caveats.
• Carbon Source Utilization Predicts Ecology and Lifestyle in Pseudomonas — GapMind carbon-pathway profiles, Pseudomonas subgenus differences, and environmental prediction limits.
• Web of Microbes Data Explorer — Web of Microbes metabolite-action semantics, cross-collection links, pangenome representation, and integration limitations.