What it is
Canonical name: BacDive. 1
BacDive is a bacterial strain database providing strain isolation-environment and phenotype metadata. 1 The BERDL Data Atlas inventories 97,334 BacDive strain phenotype profiles, supporting the corpus size used by the metal-tolerance studies. 2
Known aliases: No aliases were reported in the source documents. 1
Stable external identifier: No stable external identifier was reported in the source documents. 1
BacDive taxonomy must be reconciled with gtdb naming before joins: its species names can differ from GTDB names, and the pitfalls report identifies BacDive–GTDB naming differences as a source of unmatched species concepts rather than evidence of biological absence. 3
The plant-microbiome analysis identified 2,482 plant-related BacDive strains, but these records contributed zero additional genome upgrades beyond ncbi_env; this refines BacDive’s role in plant-association analyses as a phenotype and metadata resource rather than an independent source of genome assignments. 4
Key facts from BacDive–metal-tolerance studies
BacDive strain records have been linked to gtdb pangenome metal-tolerance scores from the metal-fitness-atlas to test whether isolation environments and classical phenotypes validate or independently predict genome-based tolerance. 1,5
The isolation-environment validation matched 42,227 of 97,334 BacDive strains (43.4%) to 6,426 GTDB species, including 25,089 matched strains with isolation-source metadata. 1 Exact species-name agreement accounted for 33,535 strains (34.5%), while GTDB suffix removal produced 8,692 additional matches (8.9%); 55,107 strains (56.6%) remained unmatched. 1
A later phenotype bridge used the same 97,334-strain BacDive corpus but a different matching and filtering procedure: 37,368 strains (38.4%) matched a pangenome species and metal score, representing 5,647 GTDB species, including 3,994 species with at least 5 phenotype features. This refines the earlier coverage estimate by showing that usable matching depends on the analysis endpoint and phenotype requirements. 5
Bacteria isolated from heavy-metal contamination sites had a median metal score of 0.240, a mean of 0.236, a median delta of +0.053, and Cohen’s d of +1.00 relative to the environmental baseline; the comparison had Mann-Whitney p=0.006 and n=10. 1 The environmental baseline had a median metal score of 0.187 and a mean of 0.195. 1 The reported contamination gradient was heavy metal (+1.00) > waste/sludge (+0.57) > all contamination (+0.43) > industrial (+0.20). 1 Waste/sludge isolates had n=305 and d=+0.57, all-contamination isolates had n=176 and d=+0.43, and industrial isolates had n=796 and d=+0.20. 1
Within Pseudomonadota, contamination isolates had metal scores delta=+0.040 above environmental isolates (p<0.001; n=85 contamination and n=5,655 environmental), while within Actinomycetota the delta was +0.035 (p<0.001; n=62 and n=772, respectively). 1 These environmental associations support BacDive metadata as an ecological validation signal, but the phenotype bridge refines that interpretation by showing that curated phenotypes often track taxonomy rather than adding independent predictive power. 1,5
The phenotype bridge tested 10 features; seven passed FDR (false discovery rate) correction at q < 0.05: Gram stain (d = -0.610, q = 4.0e-60), oxidase (d = +0.530, q = 1.3e-24), motility (d = +0.345, q = 7.2e-23), urease (d = -0.175, q = 9.1e-06), enzyme breadth (rho = -0.058, q = 8.2e-04), nitrate reduction (d = +0.100, q = 7.4e-03), and catalase (d = +0.104, q = 4.1e-02). H₂S production did not pass correction (d = -0.867, q = 7.3e-02), while metabolite breadth (rho = -0.013, q = 4.7e-01) and acetate utilization (d = +0.005, q = 7.9e-01) were not significant. 5
Gram-negative species had higher metal tolerance scores than Gram-positive species (Cohen’s d = -0.61, p < 1e-60, n = 3,272 species), but the signal was almost entirely between lineages, especially Actinomycetes and Proteobacteria, so the association is phylogenetically confounded. 5 Urease-positive species likewise had lower scores (d = -0.18, p < 1e-5), driven by Actinomycetes (d = -0.59, p < 1e-16) and disappearing within Gammaproteobacteria (d = +0.08, ns) and Bacilli (d = +0.06, ns); this contradicts a general prediction that urease-associated nickel handling confers higher tolerance. 5 Catalase showed a Simpson’s-paradox pattern: the overall association was positive (d = +0.10), but catalase-negative species scored higher within Actinomycetes (d = -0.62, p < 1e-5), Gammaproteobacteria (d = -0.49, p = 0.004), and Betaproteobacteria (d = -0.51, p = 0.006). 5
The anaerobe-aerobe difference was negligible (d = -0.016, p = 0.55) among 3,751 species; facultative anaerobes had mean score 0.221, compared with 0.216 for aerobes and 0.215 for anaerobes, although the three-group Kruskal-Wallis test was marginally significant (H = 8.53, p = 0.014). 5 Thus the new analysis contradicts a broad anaerobe advantage and finds the effect biologically trivial. 5
The Gene Function Ecological Agora analysis supports BacDive as a phenotype anchor for genome-derived ecological interpretations: among 318 Mycobacteriaceae species, 95% of recorded cell shapes were rod-shaped, 99% were non-motile, and 89.4% were aerobic-leaning; among 577 Bacteroidota species, anaerobic phenotypes occurred at 33.2% versus 22.0% atlas-wide and the profile was saccharolytic and glycoside-hydrolase-rich. 6 These observations are consistent with, but do not independently establish, the corresponding atlas ecological patterns. Cyanobacteriia coverage was too thin for interpretation at n = 4, which refines the scope of phenotype-based validation. 6
Taxonomy alone explained 35.4% of metal tolerance variance, phenotype features alone explained 16.3%, and taxonomy plus phenotype yielded R² = 0.345, slightly below taxonomy alone (delta R² = -0.009). Adding n_metal_clusters increased the full model to R² = 0.633, making genome-encoded resistance repertoire the strongest predictor in this analysis. 5 In 5-fold phylogenetic-blocked cross-validation, gene-count-only, phenotype-only, taxonomy-only, taxonomy-plus-phenotype, and full models had R² values of 0.063, 0.163, 0.354, 0.345, and 0.633, respectively, with RMSE values 0.045, 0.043, 0.038, 0.038, and 0.028; every model used n = 3,994 species. 5 SHAP (Shapley additive explanation) importance placed taxonomic class/order codes and n_metal_clusters among the top predictors, while phenotype features contributed minimally after taxonomy was included. 5
Only 24 matched strains had metal-utilization records covering iron, manganese, arsenate, chromate, cobalt, or zinc; 8 results were positive and 16 were negative. 1 The exploratory comparison of positive utilizers with other strains had Mann-Whitney p=0.14 and Cohen’s d=-0.57, so the direction was inconclusive. 1
BacDive utilization is four-valued rather than binary: +, -, produced, and +/-; only explicit + and - observations should enter utilization percentages. 3 For Pseudomonas fluorescens, indole had 60 “produced” entries and 1 actual utilization test, illustrating why production records must not be counted as utilization positives. 3
The newer gene-resolution cross-resistance study contradicts a strong interpretation of BacDive as an independently validated predictor at species scale: across 28 organisms, its multi-metal tolerance scores showed no correlation with BacDive metal-associated isolation at the Fitness Browser species scale (Spearman rho approximately -0.02, p > 0.8). 7 After excluding 2 organisms without tier-classification data and collapsing multiple Fitness Browser strains from the same species, the effective sample size was 20 independent species, which the study considered too small for a meaningful correlation test. 7 The report also notes that genus-plus-species substring matching was imprecise for organisms identified only to genus level, such as Acidovorax sp. 7 This refines the earlier positive environmental validation by distinguishing genome-level cross-resistance structure from the underpowered species-scale isolation-environment test. 7
The study’s heavy-metal estimate was imprecise because only n=10 matched heavy-metal isolates were available; at n=10 versus approximately 5,000 environmental-baseline strains, the minimum detectable effect size at 80% power was approximately d=0.93, compared with the observed d=1.00. 1 The phenotype study likewise had sparse feature coverage, ranging from 43,378 strains with isolation-source data and 29,784 with metabolite-breadth data to 5,254 with H₂S-production data and 1,980 with acetate-utilization data. 5 The new validation reports n = 26 at Fitness Browser organism scale but n = 20 after matching and collapsing to independent species, a discrepancy that further refines the documented power limitation. 7
The direct Fitness Browser–BacDive validation contained 12 organisms representing 6 unique species: Cupriavidus basilensis, Methanococcus maripaludis, Ralstonia solanacearum, Pseudomonas simiae, azospirillum-brasilense, and Pseudomonas fluorescens. 5 All Gram-typed organisms were Gram-negative, all urease-typed organisms were urease-negative yet routinely tested against nickel, and the single anaerobe had only 1 metal tested versus 4–5 for aerobes, preventing interpretable within-set tests. 5
Plant microbiome and pathway-validation evidence
The plant-microbiome analysis supports BacDive as an independent check on GapMind pathway predictions: BacDive metabolite-utilization data agreed with GapMind predictions at 83.1% consistency. 4 However, the GapMind core-level completeness score was 0% across compartments and may be too stringent for broad taxonomic comparisons, so this agreement does not establish complete pathway activity. 4
This analysis also refines interpretation of BacDive’s plant metadata: 2,482 plant-related strains were identified, but they produced zero additional genome upgrades beyond ncbi_env; plant compartment assignments therefore remained primarily dependent on NCBI isolation-source metadata. 4 The broader plant-associated census contained 7,995 genomes among 293,059 GTDB r214 genomes, and only 2.7% had plant-associated annotations, underscoring the limited coverage of isolation metadata. 4
The corrected analysis found weakly negative rather than complementary pathway profiles among co-occurring plant-associated genera, with Cohen’s d approximately −0.39 for prevalence weighting; the BacDive–GapMind agreement supports the pathway measurements but does not independently validate the ecological complementarity inference. 4
FW300-N2E3 metabolic utilization evidence
A cross-database study of Pseudomonas FW300-N2E3 supports BacDive as a complementary species-level utilization resource: 8 of 58 WoM exometabolites (14%) could be matched to P. fluorescens BacDive utilization data, compared with 28 of 58 (48%) matched to Fitness Browser conditions and 13 of 58 (22%) to GapMind predictions. 8 Among 7 WoM-produced metabolites with BacDive matches, 3/7 (43%) were utilized; this was not higher than the overall P. fluorescens baseline of 22/80 (27.5%; p = 0.40). 8
BacDive was the variable component of the comparison: 3/7 matched metabolites were utilized, while tryptophan had 0+/50- utilization, trehalose 1+/5-, lysine 0+/3-, and glycine 0+/1-. 8 The tryptophan result is the strongest discordance between production and species-level utilization, but its interpretation is limited because BacDive aggregates P. fluorescens strains and does not establish that FW300-N2E3 itself cannot grow on tryptophan. 8 The sample-size gradient refines interpretation of BacDive negatives: tryptophan is high confidence at n=50, trehalose is moderate confidence at n=6, lysine is moderate confidence at n=3, and glycine is low confidence at n=1. 8
Three metabolites provided four-way consistency across WoM production, Fitness Browser growth, BacDive utilization, and GapMind pathway prediction: malate was utilized by 49/49 strains (100%), arginine by 40/48 strains (83%), and valine by 1/1 strain. 8 These findings support using BacDive utilization records alongside, rather than as a substitute for, organism-specific fitness and pathway evidence; production and non-utilization are not intrinsically contradictory because exometabolomics can reflect overflow, byproduct release, or ecological secretion. 8
FW300-N2E3 was isolated from groundwater at the Oak Ridge Field Research Center in an ENIGMA SFA context, and the study notes that P. fluorescens BacDive records span a broad clade identified under GTDB reclassification as Pseudomonas_E fluorescens_E. 8 BacDive counts use per-strain consensus, applying majority vote among duplicate records per strain, and compound coverage varies from 1 to 51 strains; raw records can therefore exceed the number of strains. 8
BacDive represents culturable, described strains rather than the full diversity of environmental bacteria, and culture-collection bias may under-represent metal-tolerant extremophiles. 1 It is also biased toward well-studied organisms, including Pseudomonas and escherichia-coli, while poorly studied species have sparse data. 5 The Metal Fitness Atlas scores are genome-based predictions rather than direct metal-tolerance measurements, so these results remain phenotype-to-genome correlations rather than phenotype-to-phenotype measurements. 5 The FW300-N2E3 comparison refines this limitation by showing that strain aggregation, taxonomy breadth, and sample size constrain species-level utilization inference. 8 The Agora analysis further refines it by showing that phenotype coverage is uneven across clades, with Cyanobacteriia represented by only n = 4 species. 6 The plant analysis likewise shows that BacDive pathway validation is useful but that plant-linked strain coverage does not necessarily add genome assignments. 4
The cross-resistance analysis identifies metal concentration differences, unequal numbers of metal experiments, lack of non-metal stress controls, and non-independent organism sampling as limitations requiring concentration normalization, controls, and phylogenetic methods such as PGLS (phylogenetic generalized least squares). 7
BacDive–GTDB and cross-database analyses should therefore use live schema and identifier inspection rather than copied historical joins, because naming conventions and database contents can change; unmatched records, orphan pangenomes, and naming differences must not be interpreted as biological absence. 3
Related pages
- BacDive Isolation Environment × Metal Tolerance Prediction — summary of the BacDive isolation-environment and metal-tolerance validation study. 1
- BacDive Phenotype Signatures of Metal Tolerance — summary of the BacDive phenotype-signature and phylogenetic-confounding study. 5
- BERDL Data Atlas — Inventory, Topic Map, and Cross-Reference Synergies — inventory and cross-reference analysis placing BacDive among KBase Data Lakehouse phenotype resources. 2
- Metabolic Consistency of Pseudomonas FW300-N2E3 Across Four KBase Data Lakehouse Databases — cross-database metabolic consistency analysis for Pseudomonas FW300-N2E3. 8
- Gene Function Ecological Agora — GTDB-scale integration of BacDive phenotype profiles with clade-function and ecological analyses. 6
- Gene-Resolution Metal Cross-Resistance Across Diverse Bacteria — gene-resolution cross-resistance study and its underpowered BacDive isolation-environment validation. 7
- Plant Microbiome Ecotypes — Compartment-Specific Functional Guilds and Their Genetic Architecture — plant-associated ecotypes, compartment-specific functions, and BacDive–GapMind pathway validation. 4
- pitfalls — database naming, utilization coding, identifier reconciliation, and interpretation safeguards relevant to BacDive integration. 3
- environmental-resistome — ecological validation and genome-encoded predictors of metal tolerance. 1,5,7
- pangenome-integration — BacDive–GTDB species matching and pangenome coverage. 1,5
- multi-omics-integration — integration of genome-derived scores with curated isolation and utilization phenotypes. 1
- ecotype-environment-gene-content — phenotype anchors for interpreting clade-specific gene-function and environmental patterns. 6,4
- condition-specific-fitness — need for per-metal analyses and matched-strain experiments to test metal-specific tolerance. 5
- metabolic-model-gapfilling — BacDive utilization records and GapMind pathway predictions as complementary metabolic evidence. 4
- metal-fitness-atlas — source of the genome-based metal-tolerance scores used in the comparisons. 1,5