What this entity is
Bakta is the canonical name for a bacterial genome annotation method used in these reports for antimicrobial-resistance (AMR) gene classification. 1 The pan-bacterial atlas supports this role, applying uniform Bakta and AMRFinderPlus annotations across 27,690 pangenome species before integrating conservation, function, environmental, and fitness data. 2 The lanthanide methylotrophy atlas supports and broadens Bakta’s role as a product-level annotation source for metabolic-marker detection across a 293,059-genome collection. 3
Known aliases: Bakta bacterial genome annotation; bakta_amr; bakta_annotations. 1
Stable external identifier: No stable external identifier is specified in these reports. 1
Annotation data and integration
The fitness-cost analysis used the kbase-ke-pangenome tables bakta_amr and bakta_annotations to identify and classify AMR genes before linking them to kescience-fitnessbrowser fitness measurements. 1 The broader Bakta-based assembly contained 178 Tier 1 AMR genes and 1,174 Tier 2 keyword-annotated genes; the fitness-linked analysis included 1,352 AMR genes across 43 organisms. 1 Tier 1 comprised 110 genes in the fitness-linked analysis, while Tier 2 comprised 691 genes. 1 Tier 2 represented 86% of the AMR genes and could include non-AMR genes such as general efflux transporters; a Tier 1-versus-Tier 2 sensitivity comparison found indistinguishable fitness distributions (KS p = 0.17). 1
The atlas refines this workflow at larger scale: Bakta product annotations and eggNOG hits were present for 93.0% of AMR clusters, while 7.0% had Bakta annotations only. 2 It identified 83,008 AMRFinderPlus hits on gene-cluster representatives, spanning 82,908 distinct clusters, 1,939 AMR gene families, 2,079 AMR products, and 14,723 species. 2 Its mechanism categories were assigned by keyword matching against AMRFinderPlus product descriptions rather than the CARD Antibiotic Resistance Ontology (ARO), making this a related but distinct classification layer from the Bakta-derived classifier used in the fitness-cost analysis. 2
The new prophage analysis refines interpretation of Bakta annotation products: keyword and Pfam matching in bakta_annotations was used to identify prophage markers, but may include false positives such as phage-defense systems and miss divergent prophages because dedicated prophage predictors were not used. 4 Across the GTDB pangenome, this annotation-derived inventory contained 3,465,244 broad prophage marker clusters and 1,261,929 strict prophage-marker clusters; 83.8% of broad markers were accessory and 53.8% singleton. 4 Of 27,702 species, 14,669 (52.9%) carried both AMR and prophage markers. 4 In a sampled set of 36,041 AMR gene instances from 1,953 genomes across 100 species, 20,073 (55.7%) were on contigs carrying strict prophage markers, although only 1,991 (5.5%) were within 5 genes and distances were based on ordinal gene positions rather than base-pair coordinates. 4 This supports Bakta’s utility for large-scale candidate discovery but does not establish prophage-mediated AMR mobilization; the species-level association between prophage-marker density and AMR repertoire breadth was Spearman rho 0.572 across 4,770 species (p<10^-300), with partial rho 0.464 after controlling for genome count. 4
The new pitfalls guidance refines interpretation and reproducibility of these annotation tables: Bakta and InterProScan annotations in the pangenome are joined at the gene-cluster level, with eggnog_mapper_annotations.query_name joining to gene_cluster.gene_cluster_id, not gene.gene_id. 5 Because KBase Data Lakehouse schemas and namespace addresses can change during Delta-to-Iceberg migration, live catalog and schema discovery should precede reuse of historical Bakta queries. 5 Numeric fields in relevant pangenome metadata may be strings and must be explicitly cast before arithmetic, ordering, comparisons, or aggregation. 5
In the annotation-gap study, Bakta was repurposed beyond AMR classification: its annotations were queried for alternative EC numbers and product-name matches against gapfilled reactions. 6 This extends the method’s role to recovering metabolic gene candidates; the Bakta stream produced 1,459 EC candidate entries and added 22 newly resolved reaction-organism pairs (10.9% of 201 pairs). 6 These assignments were integrated with diamond/BLAST homology, fitness, and pangenome evidence rather than treated as sufficient alone. 6
The lanthanide atlas refines this evidence-integration principle through marker-specific calibration: Bakta was preferred for lanmodulin because it identified 62 product-level Lanmodulin calls, whereas eggNOG produced 505 non-specific or false-positive lanM calls; Bakta was also preferred for xoxJ, eggNOG K00114 was primary for xoxF with Bakta used as a union source, and eggNOG K14028 was primary for mxaF. 3 Among 2,185 xoxF genomes lacking eggNOG PQQ annotations, 1,288 (59%) had at least one Bakta PQQ product, while 897 lacked PQQ evidence from either source. 3 Thus Bakta can recover metabolic-marker evidence missed by eggNOG, but product calls do not establish complete pathways or enzyme activity. 3
The PGP pangenome analysis extends Bakta’s metabolic-marker use to plant-growth-promoting genes: exact Bakta gene-name matches detected 13 markers across 27,702 species, including 11,272 species carrying at least one marker. 7 Soil/rhizosphere enrichment was found for acdS, pqqC, and hcnC, while nifH was depleted; these are gene-distribution results rather than functional validation. 7 PGP genes were more core than the 46.8% genome-wide baseline: pqqC was 81.5% core and acdS 70.4% core, while pqqD was a partial exception at 55.5% core and 27.5% singleton. 7 Product-only or variant names may be missed, and truncations, frameshifts, and pseudogenization were not filtered. 7
Pan-bacterial annotation and dark genes
The discoveries log supports Bakta’s broader annotation utility but refines its coverage interpretation: across 132.5M gene clusters, eggNOG had higher COG, KEGG, and Pfam coverage (51%, 38.5%, and 63%) than Bakta (8.2%, 17.3%, and 7.7%), whereas Bakta had higher GO coverage (15% versus 7.4%), product descriptions (71.2% versus 70.4%), and unique UniRef50 links covering 79.2%. 8 Their union increased any-functional-annotation coverage to 77.3%, and Bakta rescued 11.2M clusters among 39.2M missed by eggNOG. 8 Only 33.3% of Bakta’s 17.6M distinct UniRef50 IDs existed in the KBase Data Lakehouse UniProt identifier table, so downstream identifier integration remains incomplete. 8
Bakta v1.12.0 supports and sharpens dark-gene reclassification: among 39,532 Fitness Browser dark genes with pangenome links, 33,105 (83.7%) were reclassified and 6,427 remained truly dark. 8 The functional-dark-matter analysis used Bakta v1.12.0 with DB v6.0 and represented 58.1% of all dark genes. 9 All 100 top-ranked dark-gene candidates received Bakta product descriptions, and 5 dark genes also received AMRFinderPlus annotations. 9 Among the 6,427 genes remaining hypothetical in both Fitness Browser and Bakta, UniRef50 links covered 79.4%, UniParc/UniRef100 links 69.1%, and RefSeq links 62.4%. 9
The new truly-dark-gene analysis refines this result by separating the 33,105 annotation-lag genes from 6,427 genes hypothetical in both pipelines and 17,479 dark genes without pangenome links, whose Bakta status could not be assessed. 10 Among truly dark genes, only 4.0% had Pfam hits, 4.6% had KEGG KOs, and 43.5% had partial eggNOG-mapper signal; nevertheless, 79.4% had UniRef50 links and 84.7% had database cross-references. 10 Only 246 (3.8%) had zero annotation clues, while 2,314 had partial functional or phenotype clues suitable for prioritization. 10 Thus the new analysis supports the existing conclusion that dark status often reflects incomplete or insufficiently integrated annotation rather than complete absence of evidence. 10
Truly dark genes were shorter (121 versus 194 aa), less conserved (43% versus 73% core), had fewer orthologs (29% versus 64%), and had higher GC deviation (d=0.247); 41% of their neighbors were hypothetical and 12% were within 2 genes of mobile elements. 8 The new analysis refines the interpretation of these differences: the corresponding core-genome fractions were 43.1% versus 72.7%, ortholog presence was 29.3% versus 63.7%, and median ortholog breadth was 1 versus 4 organisms; truly dark genes also had an essential fraction of 18.0% versus 13.4%. 10 These patterns are consistent with biological novelty, accessory-genome localization, or horizontal-transfer-associated evolution, but GC deviation and mobile-element proximity do not establish horizontal gene transfer, and short genes are intrinsically harder to annotate and measure by transposon insertion. 10
These results refine the AMR, gapfilling, metabolic-marker, and prophage-marker uses: Bakta annotations are valuable candidate-recovery evidence, but are strongest when combined with eggNOG, homology, pangenome, dedicated mobile-element prediction, and fitness evidence rather than used as standalone functional assignments. 8 9 4 The lanthanide atlas supports this caution because its Bakta-validated lanmodulin calls were clade-restricted and only partly co-occurred with xoxF, while sequence-level evidence for pseudogenes, truncated ORFs, or assembly fragmentation was out of scope. 3 The truly-dark-gene report likewise notes that Bakta may produce false-negative functional calls when a known function is absent from the tested PSC database. 10
Mechanism classification and conservation
The fitness-cost report classified resistance mechanisms using Bakta-derived AMR annotations, including efflux, enzymatic inactivation, metal resistance, and unknown categories. 1 The baseline-fitness comparison included 254 efflux, 304 enzymatic-inactivation, 144 metal-resistance, and 74 unknown genes. 1 Mechanism assignments were strongly associated with pangenome conservation status (χ² = 69.3, p = 1.4×10⁻¹³): 44% of metal-resistance genes were accessory, compared with 13% of efflux genes and 16% of enzymatic-inactivation genes. 1
The atlas supports and broadens this conservation analysis: AMR genes were 30.3% core versus 46.8% for the pangenome baseline, while the auxiliary genome was 2.2x enriched for AMR (33.6% versus 15.3%). 2 Its mechanism-level results showed 54.9% core beta-lactamases and 6.5% core regulatory genes, although these categories came from AMRFinderPlus product keywords rather than the earlier Bakta-derived scheme. 2
The prophage analysis refines the interpretation of accessory status: AMR genes within 10 genes of a prophage marker were 67.6% accessory versus 65.5% for distal AMR genes (odds ratio 1.10, one-sided p=0.005, bootstrap 95% CI [1.024, 1.185]), a modest 2.1 percentage point difference. 4 The threshold dependence and heterogeneity—odds ratios 0.78, 0.92, 1.10, 1.19, and 1.28 at 3, 5, 10, 15, and 50 genes, respectively; only 33 of 74 testable species above 1 and median species-level odds ratio 0.85—contradict any simple claim that immediate prophage proximity uniformly marks recently acquired AMR cargo. 4 The proposed reversal at very close range and broader genomic-island interpretation are hypotheses, not direct mobilization tests. 4
The product classifier did not handle fosfomycin or tellurite resistance annotations, leaving approximately 25 genes in the unknown mechanism category; reclassification would move these genes to enzymatic inactivation and metal resistance, respectively. 1 The atlas similarly found 22.2% of hits Other/Unclassified and proposed systematic CARD ARO mapping. 2
Related pages
Bakta annotation supported integration of resistance classification with environmental-resistome and pangenome-integration. 1 The atlas extends these connections through amrfinderplus, kbase-ke-pangenome, environmental metadata, and Fitness Browser links across the pan-bacterial collection. 2 Marker calibration connects Bakta to gene-function-acquisition-depth through source-specific interpretation of metabolic annotations. 3 The annotation-gap study connects it to metabolic-model-gapfilling, multi-omics-integration, and condition-specific-fitness through evidence triangulation. 6 The PGP analysis extends these links to ecotype-environment-gene-content and gene-function-acquisition-depth. 7 The prophage analysis additionally connects Bakta-derived marker interpretation to phage-defense-syndromes-and-arms-race through the distinction between prophage markers and possible phage-defense-system false positives. 4
The method is discussed in Fitness Cost of Antimicrobial Resistance Genes, Pan-Bacterial AMR Gene Landscape, Annotation-Gap Discovery via Phenotype-Fitness-Pangenome-Gapfilling Integration, Functional Dark Matter — Experimentally Prioritized Novel Genetic Systems, discoveries, Lanthanide Methylotrophy Atlas — Summary, PGP Gene Distribution Across Environments & Pangenomes, pitfalls, and Prophage-AMR Co-mobilization Atlas. 1 2 6 9 8 3 7 5 4
The new dark-gene report extends the method’s documented scope to residual annotation uncertainty and candidate prioritization; it is discussed in Truly Dark Genes — What Remains Unknown After Modern Annotation?. 10
amrfinderplus is a related AMR annotation resource. 1