Testing whether laboratory fitness phenotypes predict environmental gene distributions

Laboratory fitness phenotypes can be tested as predictors of where genes occur in natural environments by linking within-species carrier-versus-non-carrier comparisons, environmental metadata, and independent metagenomic observations. The evidence supports partial lab–field concordance, but it does not establish that laboratory fitness effects alone determine environmental gene distributions. ₁

Core evidence

The study tested 151 accessory dark-gene clusters across 31 species using carrier-versus-non-carrier environmental comparisons. Of 137 clusters tested for environmental-category enrichment, 10 were significant at FDR < 0.05, where FDR means the false-discovery rate, and 1 of 67 showed significant AlphaEarth embedding divergence. ₁

The 10 clusters with significant environmental-category enrichment (FDR < 0.05) are listed below. Each entry gives the gene, the lab condition class with |fit|, the environment category, the odds ratio and the FDR. An odds ratio below 1 means carriers are less often found in that category, not more. ₁

• P. putida PP_0025: stress, 4.8; human_clinical; OR 27.5; FDR 7e-6
• P. putida PP_3434: nitrogen, 3.1; human_clinical; OR 28.6; FDR 7e-6
• P. putida PP_0642: nitrogen, 2.8; human_clinical; OR 11.6; FDR 0.001
• P. putida PP_3105: stress, 3.7; human_assoc; OR inf; FDR 0.004
• B. thetaiotaomicron 354052: stress, 2.4; human_assoc; OR 0.17; FDR 0.005
• P. syringae B728a Psyr_0167: in planta, 4.6; plant_assoc; OR 11.9; FDR 0.005
• B. thetaiotaomicron 350920: stress, 2.2; human_assoc; OR 0.37; FDR 0.031
• P. putida N2C3 AO356_11255: nitrogen, 3.4; freshwater; OR inf; FDR 0.031; module prediction D-Ala-D-Ala carboxypeptidase
• K. oxytoca BWI76_RS15640: carbon, 2.1; human_assoc; OR 0.14; FDR 0.031; module prediction phage tail tape-measure
• P. syringae B728a Psyr_2830: stress, 3.4; plant_assoc; OR 10.9; FDR 0.031

The report sees two patterns in these significant clusters. First, P. putida dark genes with stress or nitrogen phenotypes are enriched in clinical isolates (human_clinical or human_associated). The report suggests this reflects roles in host-associated niche adaptation. Second, P. syringae dark genes with in-planta or stress phenotypes are enriched in plant-associated genomes, consistent with their laboratory phenotypes. The host-adaptation reading is the report’s hypothesis, not a tested mechanism. The report also identifies Pseudomonas and P. syringae as carrying the strongest within-species signals, and it treats NMDC correlations across 6,365 metagenomic samples as community-level corroboration. It describes carriers of its top-ranked candidate, AO356_11255, as occurring exclusively in freshwater/soil environments, matching a nitrogen-utilization laboratory phenotype. That wording is stronger than the concordance table below, which shows 80.0% of AO356_11255 carriers versus 8.3% of non-carriers from soil, freshwater, or wastewater environments. ₁

A pre-registered mapping between laboratory fitness conditions and environmental associations found 29 of 47 testable clusters concordant, corresponding to 61.7% concordance. Concordance was 100% across 4 pH tests and 78% across 9 nitrogen-source tests. Six clusters reached FDR < 0.2. The authors describe this lab–field concordance approach, which tests whether lab fitness conditions predict the field environments of gene carriers, as novel to their knowledge in its systematic application across multiple organisms and condition classes; this is a self-described novelty claim. ₁

The six concordant clusters at FDR < 0.2 are listed below. Each entry gives the gene, the lab condition class with |fit|, the expected environments, the share of carriers versus non-carriers found in those environments, the odds ratio and the FDR. ₁

• K. oxytoca BWI76_RS15525: carbon, 2.3; soil, freshwater, plant; 14.3% vs 2.5%; OR 6.6; FDR 0.069
• K. oxytoca BWI76_RS15535: carbon, 2.2; soil, freshwater, plant; 14.3% vs 2.5%; OR 6.6; FDR 0.069
• P. putida N2C3 AO356_12450: carbon, 2.1; soil, freshwater, plant; 62.5% vs 0%; OR inf; FDR 0.093
• P. putida N2C3 AO356_11255: nitrogen, 3.4; soil, freshwater, wastewater; 80.0% vs 8.3%; OR 44.0; FDR 0.093
• P. putida N2C3 AO356_25185: anaerobic, 2.7; soil, freshwater, animal; 55.6% vs 0%; OR inf; FDR 0.178
• P. putida N2C3 AO356_24150: nitrogen, 3.0; soil, freshwater, wastewater; 55.6% vs 0%; OR inf; FDR 0.178

The 29/47 concordance rate provides suggestive rather than decisive evidence: a one-sided binomial test gave p = 0.072 with Wilson score 95% CI [0.474, 0.742], whereas Fisher’s combined probability across 47 individual tests gave p = 0.031. These statistics answer different questions and should not be treated as a single resolved significance claim. The report’s own conclusions nonetheless state that the overall 61.7% concordance rate (29/47 testable clusters) exceeds the 50% chance level. They cite binomial p = 0.072 and Fisher’s combined p = 0.031 together. They also cite significant environmental enrichment in 10/137 clusters, AlphaEarth divergence in 1/67, and NMDC confirmation of all 4 testable pre-registered predictions, with 76/105 abiotic correlations reaching significance. The conclusions acknowledge that compositional coupling inflates the NMDC signal. They still state that directional concordance across independent datasets supports the inference that lab fitness phenotypes reflect real ecological function. The report also lists systematic lab–field concordance testing among its contributions: 61.7% concordance across 47 testable clusters, with 4/4 pre-registered NMDC predictions confirmed. The binomial test is not significant at 0.05, so that conclusion is stronger than the binomial result alone supports. ₁

The report’s figures cover these tests: fig09_carrier_tests.png shows carrier vs non-carrier environmental enrichment tests, fig11_concordance_matrix.png shows lab–field concordance by condition class, fig12_nmdc_correlations.png is the NMDC abiotic correlation volcano plot, and fig17_concordance_comparison.png compares dark vs annotated gene concordance distributions. ₁

The central discoveries digest reports the same directional concordance for 29 of 47 testable dark-gene clusters (61.7%; Fisher’s combined p=0.031). It also reports that NMDC confirmed all 4 pre-registered abiotic predictions and all 7 pre-registered trait predictions. The digest calls this the first systematic evidence that Fitness Browser laboratory phenotypes reflect real ecological function, validating fitness data as a proxy for in situ importance. ₂ The digest passage does not mention the one-sided binomial p = 0.072 or the compositional-coupling caveat, so its framing is stronger than the project report’s own statistics support; see Tensions. _2,1

The strongest reported biogeographic associations occurred in Pseudomonas putida: PP_0025 had odds ratio 27.5 and FDR 7e-6, while PP_3434 had odds ratio 28.6 and FDR 7e-6, both as enrichment in human_clinical genomes. AO356_11255, the project’s top-ranked candidate, had odds ratio 44.0 and FDR 0.093 in the concordance test. In that test, 80.0% of carrier genomes versus 8.3% of non-carriers came from soil, freshwater, and wastewater environments. Its environmental pattern matched a nitrogen-utilization fitness phenotype, but this match remains an association rather than direct evidence of ecological causation. The report notes that AO356_11255 is the only accessory gene among its top 10 candidates and has the strongest biogeographic signal (lab–field OR = 44, plus an NMDC nitrogen correlation). For the other top candidates, the report proposes two functional readings. Three MR-1 genes (199738, 203545 and 202450) carry K03306 module predictions with different domain architectures, which the report reads as paralogous members of a conserved nitrogen-responsive system. MR-1 genes 202463 and 203247 both predict PF01145 under stress, which the report reads as pointing to a stress-responsive membrane protein family (YGGT/GBBH-like). These functional readings are hypotheses. ₁

A direct test linking laboratory metal-tolerance phenotypes to field abundance refines this evidence toward conditional rather than universal concordance. Across 108 Oak Ridge groundwater sites, the aggregate laboratory tolerance score had a positive but non-significant association with the high-uranium/low-uranium abundance ratio (Spearman rho = 0.503, p = 0.095, n = 12 genera), while five of 11 tested genera had significant uranium associations after BH-FDR correction, with effects in both directions. ₃ This supports the interpretation that environmental structure can be detected at particular taxa or conditions without showing that a general laboratory tolerance ranking predicts field distributions. ₃

An ecotype analysis refines this evidence by showing that whole-genome gene-content similarity was generally more strongly associated with phylogenetic than environmental similarity: across 172 species, the median partial correlation was 0.0025 for environment versus 0.0143 for phylogeny, with phylogeny dominant in 60.5% of species. A significant positive environmental effect occurred in 12 species (7.0%), a significant negative effect in 4 species (2.3%), and no significant effect in 156 species (90.7%). ₄ This supports interpreting laboratory–environment concordance as conditional and potentially locus-specific rather than as evidence that environmental conditions broadly explain accessory gene distributions. The analysis also suggests the hypothesis that environmental effects may be more detectable in specific functional subsets than in whole-genome gene content. ₄

Oak Ridge genus-level field test in detail

In the Oak Ridge test, 14 of the 26 unique genera represented in the Fitness Browser were detected in oak-ridge-field-research-center groundwater communities via 16S amplicon sequencing. The most prevalent were Sphingomonas (93% of 108 sites), Pseudomonas (91%), and Caulobacter (82%). Desulfovibrio, the primary ENIGMA model organism, was detected at only 34% of sites at very low abundance (max 0.09% relative abundance). ₃

Of the 14 Fitness Browser genera detected at Oak Ridge, 11 had sufficient prevalence (>=10 sites) for correlation analysis. After BH-FDR correction (Benjamini–Hochberg false-discovery-rate control), 5 showed significant correlations with uranium, in both directions (Spearman rho; p; FDR q): ₃

• herbaspirillum: +0.336; 3.8e-4; q 0.001; more abundant at high-U sites ₃
• bacteroides: +0.264; 0.006; q 0.013; more abundant at high-U sites ₃
• Caulobacter: -0.411; 1.0e-5; q 1.1e-4; less abundant at high-U sites ₃
• sphingomonas: -0.382; 4.5e-5; q 2.5e-4; less abundant at high-U sites ₃
• pedobacter: -0.266; 0.005; q 0.013; less abundant at high-U sites ₃

Several expected signals were null. Azospirillum had a nominal positive correlation (rho=+0.20, p=0.042) that was only marginal after FDR correction (q=0.077). Desulfovibrio showed no correlation with uranium (rho=0.022, p=0.82), and neither did Pseudomonas (rho=-0.059, p=0.55), despite both being ENIGMA model organisms. ₃

The aggregate test, which correlated a lab-derived metal tolerance score (negative mean fitness under stress, higher = more tolerant) with the high-uranium/low-uranium field abundance ratio, was suggestive but not significant (Spearman rho=0.503, p=0.095, n=12 genera). The trend is in the predicted direction, with genera of higher lab tolerance tending to have higher abundance at contaminated sites, but statistical power is limited by the small number of genera. The report’s hypothesis-outcomes section calls H1 not supported but suggestive, whereas its novelty section says the overall metal tolerance metric doesn’t predict field success (‘H1 rejected’); see Tensions. ₃

The report cautions that the uranium-associated community shift is not simply ‘more metal-tolerant organisms’. It reflects broader ecological restructuring, including changes in rare taxa, redox conditions, and carbon/energy sources. The report proposes Herbaspirillum and Bacteroides as possible tolerant colonizers, which is a hypothesis rather than an established result. ₃

The report names several mechanisms that could decouple laboratory tolerance from field abundance. Field sites vary in pH, redox potential, carbon sources, sulfate, nitrate, and dozens of other parameters beyond uranium, so an organism may tolerate metals in the lab but lack the metabolic capabilities for a particular site. Field communities involve competition, cross-feeding, and syntrophy, whereas lab fitness measures single-organism performance in isolation. Geochemistry measurements are snapshots, but community composition reflects historical conditions and colonization history. The 108 overlapping samples may not capture the full range of Oak Ridge geochemistry, and confounding variables (pH, dissolved oxygen, carbon sources) were not controlled. The report cites LaSarre et al. (2020), who found microbial interactions in coculture are ‘not necessarily predictable a priori’ from single-organism fitness data, as a direct parallel. ₃

The report describes itself as the first study to directly link Fitness Browser lab fitness data with enigma-coral field community composition across a geochemical gradient; this is a self-described novelty claim. Its genus-level, bidirectional results support this page’s conditional-concordance reading rather than a general tolerance-ranking model. ₃

Isolation-ecology test of genome-predicted metal tolerance

A separate validation refines the lab-to-field question. Instead of testing laboratory fitness phenotypes directly, it tests genome-based metal-tolerance predictions against where organisms were isolated in bacdive. The report describes the result as validating the metal-fitness-atlas genome-based prediction method against real-world isolation ecology: organisms isolated from heavy-metal contamination sites had metal-tolerance scores a full standard deviation above the environmental baseline (Cohen’s d, a standardized mean difference, = +1.00; Mann-Whitney p=0.006, n=10). ₅ The report describes the effect as dose-dependent across contamination intensity: heavy metal (+1.00) > waste/sludge (+0.57) > all contamination (+0.43) > industrial (+0.20). ₅ The heavy-metal group is small and only just powered (see below). This result therefore supports the hypothesis that genome-predicted metal tolerance tracks isolation ecology, but it does not establish it. ₅

Per-category scores against the environmental baseline (n; median; mean; median delta; Cohen’s d; p) were: ₅

• Heavy metal contamination: n=10; median 0.240; mean 0.236; delta +0.053; d +1.00; p 0.006 ₅
• Waste/sludge: n=305; median 0.219; mean 0.218; delta +0.032; d +0.57; p<0.0001 ₅
• All contamination: n=176; median 0.215; mean 0.211; delta +0.028; d +0.43; p<0.0001 ₅
• Industrial: n=796; median 0.199; mean 0.202; delta +0.012; d +0.20; p<0.0001 ₅
• Host-associated: n=12,086; median 0.194; mean 0.201; delta +0.007; d +0.14; p<0.0001 ₅

One directional prediction failed. Contrary to the report’s hypothesis H1d, host-associated bacteria scored slightly but significantly higher than the environmental baseline (d=+0.14, p<0.0001). ₅ This contradicts a simple expectation that host association lowers predicted metal tolerance. ₅

The contamination signal held in some phyla but not others, which partly addresses phylogenetic confounding. Within Pseudomonadota (85 contamination vs 5,655 environmental isolates), contamination isolates had significantly higher scores (delta=+0.040, p<0.001). The same was true within Actinomycetota (62 vs 772; delta=+0.035, p<0.001). The signal was not significant within Bacillota (19 vs 492; delta=-0.012, p=0.285) or Bacteroidota (5 vs 156; delta=-0.008, p=0.456), where contamination-isolate sample sizes were small. ₅

The heavy-metal result sits at the detection limit. With n=10 heavy-metal isolates versus ~5,000 environmental baseline isolates, the minimum detectable effect at 80% power was approximately d=0.93, and the observed d=1.00 barely exceeds this threshold. The report notes that n=10 after matching leaves the effect-size estimate imprecise. ₅

A direct phenotype comparison was null. Metal-utilization records gave 8 positive and 16 negative results, an underpowered and nonsignificant comparison (Mann-Whitney p=0.14, d=-0.57). Positive utilizers had lower predicted metal scores, but the report cautions that this surprising direction is not significant and should not be over-interpreted given the tiny sample size. ₅

A species-scale BacDive check in the metal_cross_resistance project gave a null result. Multi-metal tolerance scores did not correlate with BacDive isolation from metal environments at the Fitness Browser species scale (Spearman rho ~ -0.02, p > 0.8). The validation set excluded 2 organisms without tier classification data (azobra, BFirm with <3 metals) and collapsed multiple Fitness Browser strains of the same species into one species-level entry (e.g., 5 P. fluorescens strains sharing the same BacDive pool). After matching and collapsing, the effective sample size was 20 independent species, which the report calls far too small for a meaningful correlation test. Its limitations section instead gives n = 26 at Fitness Browser organism scale; the two denominators are reported separately and are not reconciled here. The figure table labels the result only as ‘H3 BacDive validation (null at FB scale)’, with no test statistic. This underpowered null refines rather than contradicts the isolation-ecology result. ₆

The metal_cross_resistance report treats this null as expected. It states that the prior bacdive_metal_validation project demonstrated that metal tolerance prediction requires pangenome-scale analysis (42K strains, Cohen’s d = +1.0) to achieve statistical power (see Tensions for how this framing compares with the BacDive report’s own n=10 heavy-metal group). As future work, not a reported result, it proposes a pangenome-scale H3 validation. That validation would apply the cross-resistance gene signatures to predict multi-metal tolerance across 27K species, using the KEGG/PFAM mapping approach from the Metal Fitness Atlas, and validate the predictions against BacDive polymetallic isolation environments. _6,5

The metal-fitness-atlas adds a specialist-level null. A metal functional signature of 1,286 KEGG KO terms was used to score all 27,702 pangenome species. After genome-size normalization, bioleaching genera as a group were not significantly enriched over background (Mann-Whitney p=0.17). The report reads this as indicating that metal tolerance genes are broadly distributed across bacteria rather than concentrated in specialists. Without normalization, species with large open pangenomes (K. pneumoniae, P. aeruginosa) dominate, reflecting genome size rather than metal biology. This refines the isolation-ecology evidence: genome-predicted metal tolerance separated contamination isolates in BacDive, but it did not single out known metal specialists as a group. _7,5

Independent metagenomic validation

The National Microbiome Data Collaborative (NMDC) analysis independently mapped 5 of 6 carrier genera to 47 taxon columns across 6,365 metagenomic samples. It confirmed all 4 testable pre-registered abiotic predictions: nitrogen-source carriers correlated with total nitrogen at ρ = +0.109, n = 1,231, FDR = 2.3e-4, and ammonium nitrogen at ρ = +0.231, n = 1,230, FDR = 8.0e-16; pH carriers correlated with pH at ρ = +0.157, n = 4,366, FDR = 7.4e-25; and anaerobic carriers correlated negatively with dissolved oxygen at ρ = -0.298, n = 272, FDR = 1.5e-6. The genus mapping used a two-tier taxonomy bridge: GTDB metadata NCBI taxids first, with a taxonomy-dimension fallback. The report stresses that the NMDC metagenomic samples are entirely separate from the pangenome-based carrier analysis, so it treats these nitrogen, pH and oxygen matches as independent confirmations. ₁

NMDC trait-feature analysis also confirmed all 7 pre-registered trait-condition predictions at FDR < 10⁻²¹, including associations of nitrogen-source carriers with nitrogen fixation at ρ = 0.60 and nitrate denitrification at ρ = 0.52, and of carbon-source carriers with aerobic chemoheterotrophy at ρ = 0.73 and fermentation at ρ = 0.59. Anaerobic carriers correlated with fermentation (ρ=0.59) and iron respiration (ρ=0.45). The analysis correlated carrier-genus abundance with community trait scores from 76 functional-group columns across 6,365 samples. All 7 pre-registered predictions had the expected direction, and a binomial sign test on these 7/7 gave p = 0.0078. An additional 441/449 exploratory tests reached FDR < 0.05. The report cautions that these high trait-condition correlations, including the pre-registered trait associations, likely reflect compositional coupling rather than independent gene-phenotype validation, because shared genera drive both carrier abundance and community trait scores. ₁

The independent validation strengthens the hypothesis that some laboratory phenotypes predict environmental distributions, especially for pre-specified abiotic directions. However, 441 of 449 exploratory trait tests also reached FDR < 0.05. The report attributes the trait results to compositional coupling rather than to independent gene-phenotype validation, because abundant genera contribute to both carrier abundance and community trait scores. This caveat covers the high trait-condition correlations in general, the pre-registered trait associations included, not only the exploratory significance rate. The report states that compositional coupling inflates the NMDC signal. The abiotic side shows a similar pattern: 76 of 105 total Spearman correlation tests reached FDR < 0.05 (72.4%). The report says this likely reflects confounding, because the carrier genera (Pseudomonas, Klebsiella, Bacteroides) are among the most abundant and ubiquitous taxa in NMDC samples. It therefore treats the 4/4 pre-registered prediction rate as the more meaningful metric, because that rate tests specific directional hypotheses. The report also states that these correlations demonstrate ecological co-occurrence rather than causal gene-phenotype links. ₁

Interpretation and limits

The strongest interpretation is conditional concordance: laboratory fitness can identify environmentally structured genes or carriers when the laboratory condition corresponds to an ecologically relevant resource or stressor, but laboratory phenotypes should not be assumed to predict natural distributions universally. ₁ The Oak Ridge result supports this qualification: genus-level field associations were bidirectional, and the aggregate metal-tolerance relationship was only suggestive rather than statistically significant. ₃ The ecotype result supports this qualification: environmental effects were usually weak or non-significant for whole-genome gene content, and environmental and host-associated bacteria did not differ significantly in environmental effects (p=0.66). ₄ It also refines the interpretation by raising the hypothesis that collection coordinates and AlphaEarth embeddings may fail to represent the relevant microenvironment, particularly for host-associated organisms. ₄

A cross-species Fitness Browser–pangenome analysis supports this caution while clarifying the distinction between predicting conservation and predicting environmental distribution. Across approximately 194,000 genes from 43 bacteria, essential genes were 82% core whereas always-neutral genes were 66% core; fitness breadth was positively but weakly associated with core status (Spearman rho=0.086, p=8.1e-230). ₈ Thus, laboratory fitness importance can track broad evolutionary conservation, but only weakly, and this result does not show that fitness predicts where genes occur environmentally. The same analysis refines the interpretation of condition specificity: genes with strong condition-specific effects were more often core (77.3% core) than genes without specific phenotypes (70.3% core), and 4,450 genes fit an ephemeral-niche pattern of overall neutrality but criticality in one condition. ₈ This supports treating condition-specific laboratory effects as potentially informative about environmental specialization without assuming they are confined to accessory genes.

A DvH comparison refines this interpretation by separating environmental transfer from conservation prediction. Across 757 Fitness Browser experiments, classified into six condition categories, condition class explained less of pangenome conservation than fitness importance: field-stress-important genes were 83.6% core and field-core-important genes were 82.4% core, while the overall non-essential baseline was 76.3%; logistic models had cross-validated AUCs of 0.517 for field fitness alone, 0.531 for lab fitness alone, and 0.548 for both, versus 0.645 after adding gene length. These results are from a single organism (DvH). The report says this limits generalizability, because results may differ in organisms with larger accessory genomes. The results also predict core status rather than environmental distribution, so they support caution against treating laboratory fitness as a sufficient environmental predictor. ₉ The pathway_capability_dependency report cites the same contrast (field-important genes 83.6% core vs 76.3% baseline) and, on that basis, adopted condition-type stratification (carbon/nitrogen/stress) for fitness aggregation. That downstream reuse extends the DvH result into method design; it is not an independent replication. ₁₀

The condition-class breakdown refines this result. Compared with the non-essential baseline, three classes were significantly enriched after BH-FDR correction: field-stress-important genes (n=298) were 83.6% core (odds ratio, OR, 1.58; FDR q 0.026), field-core-important genes (n=376) were 82.4% core (OR 1.46; q 0.026), and lab-nutrient-important genes (n=452) were 81.4% core (OR 1.36; q 0.037). None of the remaining classes was significant: lab-other genes (n=292) were 81.5% core (OR 1.37; q 0.073), lab-antibiotic genes (n=109) were 73.4% core (OR 0.86; q 0.49), and heavy-metal genes (n=198) were 71.2% core (OR 0.77; q 0.14). Heavy-metal and lab-antibiotic genes trended below baseline without reaching significance. The report also notes a contrast among field-relevant genes: heavy-metal-important genes had low conservation (71.2%), whereas uranium/mercury-important genes, which sit in the field-stress class, had high conservation (83.6%). ₉

The report reads that below-baseline trend as suggesting that resistance functions are disproportionately in the accessory genome, consistent with recently acquired adaptive traits carried on mobile genetic elements. Because the heavy-metal and antibiotic deficits were not significant, this remains a hypothesis rather than a demonstrated mechanism (see environmental-resistome). ₉

Condition specificity did not cleanly separate field from laboratory relevance. Lab-specific fitness-defect genes (n=50) were 96.0% core, field-specific genes (n=52) were 88.5% core, and field-biased genes (n=89) were 83.1% core. The field-specific versus lab-specific contrast was not significant (Fisher exact OR=0.32, p=0.27), whereas universal genes were significantly more core than neutral genes (OR=1.35, p=0.033). ₉ The report notes that the significant lab-nutrient enrichment and the nonsignificant field-versus-lab-specific comparison (n=50-52 per group) limit the hypothesis that ecologically relevant genes are specifically conserved. The report also flags two limits on this comparison. The field-specific and lab-specific gene sets are small (n=50-52), which limits statistical power for the key comparison. Field and lab fitness effects are also correlated (r ~ 0.7 from the scatter plot), meaning most genes that are sick in one context are sick in the other. This supports this page’s conditional-concordance reading. ₉

Module-level analysis gave a null result for field activity. Across 52 independent component analysis (ICA) fitness modules, the mean core fraction was 0.886 and the median was 1.000, and field condition activity did not significantly correlate with module conservation (Spearman rho=0.071, p=0.62). ₉ Under a threshold-based classification, 9 lab modules (low field activity and less conserved) had a mean core fraction of 0.516. The 21 ecological modules contained 239 member genes, 52 of them unannotated, which the report proposes as candidates for novel environmental adaptation functions. ₉

The DvH analysis had no gene-level field measurements to draw on. enigma-coral holds 4,346 field samples with geochemistry data (uranium, metals) across 596 Oak Ridge locations and 213,044 ASVs (amplicon sequence variants) for community composition. These do not provide gene-level fitness, and the report flags them only as potentially useful for future analyses. Its proposed follow-up, which has not been carried out, would use the 213K ASVs and 4,346 field samples in ENIGMA CORAL to ask whether Desulfovibrio abundance at Oak Ridge sites correlates with the geochemistry conditions tested in the Fitness Browser. ₉

The environmental analysis is limited by sparse metadata, inconsistent NCBI isolation-source records, and AlphaEarth coverage of only 28% of genomes, or 83K/293K. ₁ The ecotype analysis reports AlphaEarth coverage of only 28.4% of genomes and likewise identifies missing or imprecise geographic coordinates as a limitation. ₄ NMDC validation was conducted at genus level, mapping NMDC taxon columns to pangenome genera via ncbi_taxid. The report notes this may miss species-specific dark gene signals. The validation matched only 5 of 6 carrier genera, and it may be influenced by broad correlations involving common genera such as Pseudomonas, Klebsiella, and Bacteroides. The report states that the high significance rate (76/105) likely reflects the dominance of common genera (e.g., Pseudomonas, Klebsiella) in both datasets. ₁ A separate coverage table lists Fitness Browser BBH (bidirectional best hit) coverage for gene-fitness as 0.18% (35 species), rated point-validation only. This limits how far laboratory fitness can be projected onto environmental gene content. ₁₁ The condition–environment mapping also departed from the research plan. The plan specified 7 mappings, including osmotic, temperature, and oxidative stress. The implementation used 6: stress (consolidating metal, osmotic, and oxidative), carbon source, nitrogen source, pH, motility, and anaerobic. The stress classes were merged because the Fitness Browser expGroup field uses a broad “stress” category. Motility and anaerobic were added because they emerged as prominent condition classes among dark genes with strong fitness effects. ₁

The Oak Ridge analysis further refines these limitations: 16S amplicon sequencing could not match laboratory organisms at species or strain level, and pH, dissolved oxygen, carbon sources, temporal history, and other geochemical factors were not controlled. ₃ The ecotype analysis further refines the environmental limitation: its partial correlations assume linear relationships between distance matrices, and geographic coordinates may describe collection sites rather than organisms’ actual microenvironments. ₄ The study did not run equivalent annotated-gene null controls for either the biogeographic analysis or the lab–field concordance analysis, so the reported dark-gene associations cannot yet be separated fully from effects of accessory-gene prevalence, taxonomic composition, or analysis design. ₁

The NMDC trait results therefore support environment-embedding-geography and environmental-resistome as evidence of testable environmental structure, while also reinforcing taxonomic-resolution-dependent-functional-inference and environment-embedding-geography as safeguards against causal overinterpretation. ₁

Relation to fitness experiments

The laboratory component used Fitness Browser phenotypes to classify gene-condition relationships, while the environmental component tested whether carrier distributions tracked metadata or embedding dimensions. This cross-domain design connects condition-specific-fitness with lab-field-fitness-concordance and environment-embedding-geography. ₁

The result refines a simple laboratory-to-field transfer model: agreement was observed for 29 of 47 testable clusters, but the one-sided binomial result was p = 0.072, metadata coverage was incomplete, and compositional coupling affected exploratory trait associations. ₁ The DvH analysis supports this refinement: field and laboratory fitness dimensions were only weak predictors of core status individually, with field-plus-lab CV-AUC 0.548, while gene length increased the full-model CV-AUC to 0.645. ₉ The Oak Ridge test supports the same refinement by showing that a field abundance gradient can contain significant taxon-specific signals even when the aggregate laboratory tolerance score is not significant. ₃ The cross-species conservation analysis further supports separating these questions: fitness importance showed a weak association with core status across approximately 194,000 genes, but the analysis did not test environmental distributions directly. ₈

The ecotype analysis supports testing this model at the level of functional subsets rather than only whole-genome gene content: it proposes evaluating COG categories, including V-Defense and L-Mobile, for stronger environmental effects. ₄ The DvH evidence also refines the scope of the transfer question: its 678 essential genes lacked transposon fitness data, and the analysis therefore compared conditionally important non-essential genes with pangenome status rather than testing all genes’ environmental distributions. ₉

Cross-organism evidence from the discoveries digest refines the dark-gene side of this question. In the digest, 65 dark-gene ortholog groups show conserved fitness effects across 3+ organisms, and dark-gene concordance levels are statistically indistinguishable from those of annotated genes (Mann-Whitney p=0.17). Motility-related dark genes show the strongest cross-organism concordance, which the digest reads as suggesting conserved but incompletely annotated chemotaxis machinery. This is a hypothesis, not an established function (see cross-species-fitness-transferability). ₂

Tensions

Oak Ridge genus count. The lab_field_ecology report gives the aggregate laboratory tolerance–field abundance association as Spearman rho = 0.503, p = 0.095, n = 12 genera, with five of 11 tested genera significant individually. ₃ The bacdive_metal_validation report cites the same result as a suggestive but non-significant correlation of rho=0.50, p=0.095, n=11 genera. ₅ Both reports agree the aggregate association is suggestive but nonsignificant. The genus count differs between them and is not reconciled here. _3,5 Within lab_field_ecology itself, 11 of the 14 detected Fitness Browser genera met the >=10-site prevalence threshold for the per-genus uranium correlations, while the aggregate tolerance test reports n=12 genera. These two counts are also not reconciled here. ₃

Wording of the Oak Ridge H1 outcome. The lab_field_ecology hypothesis-outcomes section calls the laboratory-tolerance prediction not supported but suggestive (rho=0.503, p=0.095, n=12 genera). Its novelty section instead says the overall metal tolerance metric doesn’t predict field success (‘H1 rejected’). Its literature comparison likewise describes lab metal tolerance as not predicting field abundance. The statistic itself is a nonsignificant positive trend, and both wordings are recorded here without choosing between them. ₃

Power of the BacDive metal validation. The metal_cross_resistance report characterizes the prior bacdive_metal_validation result as pangenome-scale (42K strains, Cohen’s d = +1.0), with the statistical power that its own 20-species test lacked. ₆ The bacdive_metal_validation report places its heavy-metal effect (d=+1.00) on n=10 isolates, barely above a minimum detectable effect of approximately d=0.93 at 80% power. ₅ The two descriptions of the same result’s power are not reconciled here. _6,5

Strength of the validation claims. The discoveries digest presents the 29/47 dark-gene concordance (Fisher’s combined p=0.031) plus the NMDC confirmations as validating Fitness Browser data as a proxy for in situ importance. ₂ The project report pairs the same 29/47 with a one-sided binomial p = 0.072. It also attributes the high NMDC trait-condition correlations, including the 441 of 449 exploratory tests reaching FDR < 0.05, largely to compositional coupling. Its statistics therefore support only suggestive, conditional concordance. Even so, the report’s own conclusions describe the concordance as exceeding chance and as supporting the inference that lab fitness phenotypes reflect real ecological function. ₁ Likewise, the BacDive report describes its heavy-metal result as validating the Metal Fitness Atlas while also noting that the effect barely exceeds its detection threshold. ₅

Open Directions

• Use full NMDC sample labels with within-sample permutation tests that preserve taxonomic structure to determine whether the 441 of 449 exploratory associations remain after compositional coupling is controlled. ₁
• Run the complete carrier-versus-non-carrier biogeographic pipeline on matched annotated-accessory-gene controls to test whether dark-gene associations exceed the background rate for accessory genes. ₁
• Expand environmental validation beyond the 5 of 6 matched carrier genera by integrating species-resolved or strain-resolved observations and asking whether laboratory phenotype–environment concordance persists below genus level. ₁
• Increase environmental metadata coverage beyond the 28% AlphaEarth genome coverage and harmonize isolation-source fields to test whether the 29/47 concordance estimate changes with improved environmental resolution. ₁
• Replicate the pre-registered pH, nitrogen-source, carbon-source, and anaerobic predictions in independent cohorts using partial-correlation or matched-sample models to distinguish ecological signal from study and taxonomic composition. ₁
• Compare environmental carrier predictions with condition-specific fitness across additional organisms, using continuous fitness scores and quantitative gene-cluster prevalence rather than binary core/auxiliary status to test whether the DvH pattern generalizes. ₉
• Test COG functional subsets, alternative embedding distances, and direct environmental metadata in the 172-species ecotype dataset to determine whether locus-specific environmental effects are obscured by whole-genome analyses. ₄
• Match species- or strain-resolved Oak Ridge metagenomic observations to uranium-specific laboratory fitness scores, while controlling for pH, redox, and carbon sources, to test whether the genus-level bidirectional associations persist below 16S resolution. ₃
• Test whether the weak fitness–conservation relationship generalizes to environmental prevalence by modeling continuous gene-cluster abundance across matched habitats, while controlling for gene length, transposon coverage, phylogeny, and taxonomic composition. ₈
• Map the accessory AMR genes seen in isolate genomes onto NMDC/MGnify environmental metagenome communities to test whether they are also prevalent in community DNA; this is proposed, not an established result. ₁₂
• Connect laboratory fitness to AlphaEarth environmental data to test whether organisms from more variable environments carry more trade-off genes in their core genome, a question the conservation synthesis leaves open. ₁₃
• Enlarge the heavy-metal isolate group beyond n=10 and add contamination isolates for Bacillota and Bacteroidota, to test whether the d=+1.00 effect and its within-phylum pattern hold once the comparison is no longer at the detection limit. ₅
• Re-derive the Oak Ridge genus set from the lab_field_ecology outputs to settle whether the aggregate tolerance–abundance test used 11 or 12 genera. _3,5
• Cross-reference carbon pathway predictions with RB-TnSeq (random barcode transposon sequencing, the pooled mutant-fitness assay behind the Fitness Browser) fitness data from kescience_fitnessbrowser. The aim is to experimentally validate which pathways are functionally important on specific carbon sources. The pseudomonas_carbon_ecology report proposes this test but does not report it. ₁₄
• Run the proposed pangenome-scale H3 test, predicting multi-metal tolerance across 27K species from cross-resistance signatures and validating the predictions against BacDive polymetallic isolation environments. The aim is to determine whether the null at 20 independent species reflects low power or a true absence of signal. ₆

See Field vs Lab Gene Importance in *Desulfovibrio vulgaris* Hildenborough for the DvH analysis details, Ecotype Correlation Analysis for the ecotype correlation analysis, Lab Fitness Predicts Field Ecology at Oak Ridge for the Oak Ridge laboratory-to-field test, and Fitness Effects vs Conservation — Quantitative Analysis for the cross-species fitness–conservation analysis.