Laboratory Fitness as an Imperfect Proxy for Natural Selection

Laboratory fitness measures how mutations affect growth under tested conditions, whereas natural selection acts across environments such as soil, biofilms, host tissue, and other settings not represented by a single laboratory regime. ₁ This distinction is central to interpreting Gene Conservation, Fitness, and the Architecture of Bacterial Genomes and Fitness Effects vs Conservation — Quantitative Analysis. The direct Desulfovibrio vulgaris Hildenborough (DvH) comparison in Field vs Lab Gene Importance in *Desulfovibrio vulgaris* Hildenborough refines the proxy problem by testing field-labelled and laboratory conditions within one organism. ₂ The Conservation vs Fitness — Linking FB Genes to Pangenome Clusters analysis independently supports the interpretation by linking Fitness Browser essentiality calls to pangenome conservation across 33 organisms. ₃

The integrated evidence supports a narrower claim than “field fitness predicts natural selection”: fitness importance is more consistently associated with genome conservation than whether a condition is labelled field or laboratory, but the association is weak, confounded, and not a direct measurement of selection in nature. _{2 4} The fitness_effects_conservation report reads its fitness-conservation gradient as empirical support for selection as a major driver of core-gene maintenance. It sets this against the McInerney et al. (2017) view that pangenome gene frequencies reflect a balance of selection, drift, and HGT. This is an interpretation of a correlation, not a direct test of selection. ₄ The evidence connects condition-specific-fitness, gene-essentiality, pangenome-integration, and adaptive-versus-housekeeping-functional-differentiation.

The Proxy Problem

Random barcode transposon sequencing (RB-TnSeq), a method that estimates mutant fitness from barcode abundance after pooled growth, measures the cost or benefit of gene disruption under particular laboratory conditions. ₁ Pangenome conservation records retention across genomes and therefore reflects evolutionary pressures accumulated across environments and histories rather than direct natural-environment fitness. ₁ The measurements are complementary but not interchangeable. ₁

The broad synthesis integrated RB-TnSeq fitness with pangenome conservation for 194,216 protein-coding genes from 43 diverse bacteria. ₁ Essential genes were 82% core, while genes always neutral in available experiments were 66% core; the synthesis describes this gradient as clear but modest. ₁ The separate analysis of approximately 194,000 genes supports this gradient. By fitness category, 27,693 essential (no viable mutants) genes were 82% core, 15,989 often-sick (>10% experiments) genes 78%, 25,201 sometimes-sick genes 72%, 20,739 mixed (sick + beneficial) genes 70%, 9,705 sometimes-beneficial genes 70%, and 94,889 always-neutral genes 66%. By breadth, essential genes were 82% core, genes important in 20+ experiments 79%, in 6-20 experiments 73%, in 1-5 experiments 71%, and in 0 experiments 66%. Binned by strongest effect, essential genes were 82.2% core, min_fit < -3 genes 77.7%, and min_fit -1 to 0 genes 66.4%. ₄ The report describes this as a 16-percentage-point gradient spanning ~194,000 genes across 43 diverse bacteria. It adds that the gradient, while statistically robust, means fitness importance only weakly predicts conservation. ₄ Fitness breadth correlated positively but weakly with core status (Spearman rho=0.086, p=8.1e-230). ₄ Thus, laboratory neutrality does not imply evolutionary dispensability.

The conservation_vs_fitness analysis refines these results in a separate 33-organism linkage: 27,693 putative essential genes were 86.1% core versus 81.2% of non-essential genes, with a median odds ratio of 1.56; 18 of 33 organisms showed significant enrichment after Benjamini-Hochberg false discovery rate correction (BH-FDR, a multiple-testing correction) at q < 0.05. ₃ The 82% and 86.1% estimates must not be averaged: they come from distinct cohorts and analytical definitions, creating a measurement-scope tension. _{1 3}

In DvH, the same importance-versus-context distinction was tested using 757 Fitness Browser experiments, 2,725 non-essential genes with fitness and pangenome links, and 52 independent component analysis (ICA, a decomposition method for coordinated fitness modules) modules. ₂ Among genes with fitness below -2, field-stress genes were 83.6% core across 298 genes (OR=1.58 versus baseline, FDR q=0.026), field-core genes were 82.4% core across 376 genes (OR=1.46, q=0.026), lab-other genes were 81.5% core across 292 genes (OR=1.37, q=0.073, so this enrichment was not significant after correction), and lab-nutrient genes were 81.4% core across 452 genes (OR=1.36, q=0.037). ₂ FDR means false discovery rate, and these q values were calculated after Benjamini–Hochberg correction. ₂

This refines condition-specific-fitness rather than showing that field classification is generally superior. Lab-antibiotic-important genes were 73.4% core across 109 genes (OR=0.86, q=0.49), and heavy-metal-important genes were 71.2% core across 198 genes (OR=0.77, q=0.14), below the 76.3% baseline for analyzed non-essential genes but not significantly depleted after correction. ₂ Across thresholds -3.0, -2.0, -1.5, and -1.0, field-stress genes were 89.4%, 83.6%, 84.0%, and 82.1% core, lab-antibiotic genes were 82.2%, 73.4%, 77.3%, and 76.2%, and heavy-metal genes were 70.6%, 71.2%, 76.7%, and 76.7%. ₂ Field-stress was highest at every listed threshold. The report says heavy-metals was consistently the lowest, but its own table contradicts this at -1.0, where lab-antibiotic genes (76.2%) fell below heavy-metal genes (76.7%). The report attributes the lab-antibiotic dip at -2 to a small sample (n=109 at -2 vs n=45 at -3). ₂

Specificity results support the importance component but not a simple ecological-label effect: 50 lab-specific genes were 96.0% core, 52 field-specific genes 88.5% core, 89 field-biased genes 83.1% core, 352 universal genes 79.8% core, and 2,083 neutral genes 74.5% core. ₂ Universal genes were significantly more core than neutral genes (OR=1.35, p=0.033), whereas lab-specific and field-specific genes did not differ significantly (Fisher exact OR=0.32, p=0.27). ₂ The field- and lab-specific sets were small (n=50–52), limiting comparison power. ₂

Importance, Conservation, and Prediction

The DvH ranking was not a strong predictive model. Logistic regression with 10-fold cross-validated area under the receiver operating characteristic curve (CV-AUC) gave 0.517 (standard deviation 0.052) for field fitness, 0.531 (0.052) for laboratory fitness, and 0.548 (0.053) for combined field and laboratory fitness. ₂ Adding gene length increased performance to 0.645 (0.068). ₂ The report concludes that gene length is a much stronger predictor of core status than fitness effects from either field or lab conditions, and that neither fitness dimension alone is informative (CV-AUC near 0.5). ₂ This refines pangenome-integration: fitness and pangenome status can be integrated, but ecological context alone is weak and genomic architecture must be modelled.

The broader cross-bacterial result supports the direction but qualifies its strength: strong or broad fitness effects were more conserved, yet the fitness-breadth association was weak; genes with strong condition-specific annotations were 77.3% core versus 70.3% without those annotations (OR=1.78, p=1.8e-97). ₄ This refines the DvH result because condition-specific effects need not be predominantly accessory: core genes can also have conditional effects. ₄

In DvH, field fitness alone had CV-AUC 0.517 (0.052), laboratory fitness alone 0.531 (0.052), and their combination 0.548 (0.053), while the full model including gene length reached 0.645 (0.068). ₂ The broader analyses report a real but weak gradient and did not test gene length as a covariate. _{4 1} These are not contradictory estimates of one model: organisms, predictors, outcomes, and cohorts differ. The unresolved question is how much of the broader gradient survives adjustment for length, insertion callability, phylogeny, and pangenome coverage.

The conservation_vs_fitness linkage supports a modest conservation–essentiality association but also limits its interpretation. Essentiality was inferred from RB-TnSeq under library-construction and growth conditions, and the essential fraction ranged from 12.9-28.9% per organism. ₃ There were 3,683 essential-auxiliary genes and 1,259 essential-unmapped genes, showing that laboratory viability requirements can occur outside confidently identified core clusters. ₃ Essential-core genes were 87% functionally known, whereas 38.2% of essential-auxiliary genes and 44.7% of essential-unmapped genes were hypothetical. ₃ Divergent ribosomal proteins, translation factors, transposases, and DNA-binding proteins in the latter category suggest—but do not establish—that some unmapped essential genes are divergent or recently acquired variants of core functions. ₃ This refines genomic-under-representation: unknown function is not equivalent to low importance, but the evidence remains conditional on callability and pangenome coverage.

Gene Length, Callability, and Missing Fitness

Gene length is a major confounder because it affects transposon-fitness reliability and the probability that a gene is classified as core. ₂ Short genes receive fewer insertions, reducing fitness-estimate quality and stability, while core genes tend to be longer. ₂ Length therefore associates with both measurement quality and the predicted outcome; its CV-AUC contribution may reflect better callability or correlated genomic properties rather than an independent causal role in conservation. ₂

The conservation_vs_fitness analysis supports this measurement-bias interpretation: essential genes were slightly shorter on average, consistent with insertion bias, but no length-adjusted conservation model was estimated. ₃ The DvH models excluded 678 essential genes because no transposon mutants were recovered; those genes were 80.1% core, compared with 76.3% among the 2,725 non-essential genes analysed with fitness and pangenome data. ₂ Consequently, the AUC values describe genes with usable non-essential fitness measurements, not the complete pangenome. ₂

The broader analysis supports the same concern: singleton and novel genes had near-zero mean fitness under tested laboratory conditions, but poor transposon coverage may contribute to apparent neutrality. ₄ Its essential genes were 82% core, directionally consistent with the 80.1% DvH essential-gene fraction, although the populations and analyses differ. _4,2 The conservation_vs_fitness report likewise treats its essential-gene definition as an upper bound because missing fitness data can reflect short genes, low-complexity regions, or scaffold edges rather than essentiality. ₃

Coverage limitations extend to pangenomes. Of 48 Fitness Browser organisms, 44 mapped to pangenome clades; four were unmatched, ten were excluded from Phase 2 for <90% DIAMOND coverage, and Dyella79 was excluded because a locus-tag mismatch produced a 0% join rate. ₃ Clades containing only 2 genomes also had poorly discriminating categories because presence in both genomes yields 100% core. ₃ These facts support cohort- and callability-dependent interpretation rather than universal estimates of natural selection. ₃

The Core-Genome Burden Paradox

Core genes were more likely than accessory genes to be burdensome under laboratory conditions: 24.4% of core genes showed positive fitness when deleted, compared with 19.9% of accessory genes. ₁ A positive deletion-fitness value means the mutant grew better under tested laboratory conditions, so the intact gene imposed a measurable context-specific burden. ₁ Core genes were also 1.78x more likely to have strong condition-specific phenotypes and 1.29x more likely to be trade-offs important in some conditions but burdensome in others. ₁

The fitness_effects_conservation analysis supports and refines this paradox: core genes were more likely than auxiliary genes to show positive deletion fitness, but also had heavier tails in both negative and positive directions. ₄ Core genes can therefore be more functionally active and conditionally costly, rather than simply less burdensome because they are conserved. ₄ The cross-organism result is consistent with the broader 24.4% versus 19.9% burden comparison and the 1.78-fold condition-specific enrichment. _4,1

The apparent paradox is expected if genes retained by natural selection carry costs in nutrient-rich or simplified laboratory environments while providing benefits in unassayed environments; this remains a hypothesis about environmental value, not a direct measurement of natural selection. ₁ Conservation is therefore not evidence that a gene is inert or universally beneficial in the laboratory, and laboratory burden is not evidence that it lacks value in nature. ₁ The synthesis calls laboratory conditions an impoverished proxy for nature. It proposes that a gene with positive deletion fitness in rich media may be essential for survival in soil, biofilm, host tissue, or other natural conditions; this is offered as a possibility, not a field measurement. ₁ The core_gene_tradeoffs report resolves the burden paradox with the same impoverished-proxy argument and notes that laboratory conditions capture only a fraction of the environmental conditions bacteria face in nature. ₅

The joint classifications provide indirect evidence: 28,017 genes were both costly in the laboratory and conserved, while 5,526 were costly and dispensable. ₁ The first group is consistent with purifying selection in natural environments, but does not directly measure it; the second comprises candidates for ongoing gene loss, without establishing that loss is occurring. ₁ The core_gene_tradeoffs analysis reports the same groups as signatures consistent with selection rather than direct measurements of selection in nature. ₅ Elsewhere the same interpretation is stated more strongly than the evidence allows. Core_gene_tradeoffs describes the 28,017 costly-and-conserved genes as maintained by natural selection because they are essential in natural environments that laboratory experiments do not capture ₅. Costly_dispensable_genes describes costly+conserved genes as having their laboratory burden offset by natural selection in the environment ₆. The central digest calls them the strongest evidence for purifying selection in natural environments. It also lists 5,526 costly-and-dispensable genes (candidates for ongoing gene loss) and 21,886 neutral-and-dispensable genes that it labels niche-specific ₇. None of these reports measured natural-environment essentiality directly, so these statements are extrapolations from laboratory fitness and pangenome presence. _5,6,7 The pathway_capability_dependency report cites the same 28,017 genes as costly in the lab but conserved in nature. It notes that condition-type stratification partially addresses this, but laboratory RB-TnSeq does not capture the full range of natural selective pressures. ₈ This supports pangenome-integration by combining conservation and laboratory cost instead of treating either measurement as sufficient.

Accessory genes were not systematically burdensome, contradicting a simple streamlining expectation that flexible genes are generally costly to maintain. ₁ Costs may instead depend on environment, genetic background, or accessory function. ₁ Singleton genes appeared near-neutral, but the result may reflect poor coverage rather than true dispensability. ₄

The fitness_effects_conservation analysis also identified an ephemeral-niche pattern in 2.7% of genes: these genes were neutral overall but critical in one tested condition, and they were more common among core genes than auxiliary or singleton genes. ₄ This supports the hypothesis that apparent laboratory neutrality can conceal conditional importance, while not demonstrating importance in an untested natural niche. ₄

Functional and Module-Level Structure

The proxy limitation applies to coordinated functions as well as isolated genes. ICA identified 1,116 co-regulated fitness modules across 32 organisms; modules contained 86% core genes versus an 81.5% baseline, with odds ratio 1.46 and p=1.6e-87, and 59% of modules were more than 90% core. ₁ This supports cofitness-network-architecture: conserved architecture is organized into coordinated units whose laboratory phenotypes may depend on conditions and gene interactions.

In DvH, the 52 ICA modules had mean core fraction 0.886 and median 1.000, but module conservation did not significantly correlate with field-condition activity (Spearman rho=0.071, p=0.62). ₂ Using 0.886 as the classification threshold, 21 ecological modules had mean core fraction 0.980, 17 conserved-quiet modules 0.983, 5 field-variable modules 0.829, and 9 lab modules 0.516. ₂ These results did not support field activity as a predictor of module conservation. ₂ They distinguish field-active conserved modules from lower-conservation lab modules, but do not establish that the 52 unannotated genes among 239 ecological-module members mediate environmental adaptation. ₂

The burden pattern was function-specific: motility and chemotaxis showed a +7.8pp core-burden excess, RNA metabolism a +12.9pp excess, and protein metabolism a +6.2pp excess; cell-wall genes showed the reverse, with non-core cell-wall genes more burdensome. ₁ These patterns are consistent with flagella being energetically expensive but useful for chemotaxis in nature, and ribosomal components being costly but required for rapid growth responses, but those interpretations are hypotheses rather than direct measurements of natural selection. ₁ The core_gene_tradeoffs report presents motility genes as the exemplar of the burden paradox. It states that energetically expensive flagellar machinery is conserved because it is essential for chemotaxis in natural environments; that ecological explanation is also interpretive. ₅

The core_gene_tradeoffs analysis refines this functional interpretation: the burden–conservation relationship was not uniform, with core genes more burdensome in Protein Metabolism, Motility, and RNA Metabolism, while non-core Cell Wall genes were more burdensome. ₅ It identified 25,271 condition-dependent trade-off genes, or 17.8% of genes examined, which were 1.29 times more likely to be core than non-core (odds ratio 1.29, p=1.2e-44). ₅

Condition-specific fitness did not establish niche-specific fitness. Strong condition-specific effects were more common among core genes, which may simply have more detectable effects under tested conditions; module-family breadth did not predict conservation (rho=-0.01, p=0.91), possibly because high baseline conservation left little room for a gradient. ₁ The newer analysis supports this pattern: genes with strong condition-specific annotations were 77.3% core, versus 70.3% of genes without those annotations (OR=1.78, p=1.8e-97). ₄

Limits on Ecological Generalization

In DvH, ecological labels exposed contrasts but did not provide a reliable natural-selection proxy. Field-stress genes were 83.6% core and field-core genes 82.4% core, whereas heavy-metal and lab-antibiotic-important genes were 71.2% and 73.4% core; field-stress, field-core, and lab-nutrient enrichments were significant after correction, but heavy-metal and lab-antibiotic deficits were not. ₂ This refines rather than contradicts the broader condition-specific result. ₂

Field and laboratory fitness effects were correlated at approximately r ~ 0.7 from the scatter plot, suggesting that most genes sick in one context were also sick in the other. ₂ The DvH result is single-organism evidence from a pangenome with relatively few genomes and a 76.3% baseline core fraction among analysed non-essential genes; generalization to organisms with larger accessory genomes has not been established. ₂

The broader analyses were biased toward rich media and standard stresses, covered 43 bacteria primarily from Proteobacteria, used single-gene knockouts that did not capture epistasis, and may have treated poor transposon coverage as neutrality in singleton genes. ₄ Uneven pangenome coverage, unmatched organisms, <90% DIAMOND joins, the Dyella79 locus-tag failure, and two-genome clades further limit inference. ₃ These constraints refine the claim that fitness importance predicts conservation by requiring control for lineage, genomic architecture, callability, and pangenome integration. ₃

The lower conservation of lab-antibiotic and heavy-metal-important genes is consistent with resistance functions being disproportionately accessory and potentially mobile, but mobile-element association was not measured. ₂ It therefore suggests, without establishing, that specific metal-resistance mechanisms such as efflux pumps and metal-binding proteins may be accessory, whereas uranium and mercury responses may involve more fundamental stress pathways, including DNA repair and sulfate reduction. ₂ This connects to environmental-resistome and metal-cross-resistance.

Community-level lab-to-field comparisons support the proxy limitation and extend lab-field-fitness-concordance. Aggregate lab-derived metal tolerance (negative mean fitness under stress) did not significantly predict the high-uranium/low-uranium field abundance ratio (Spearman rho=0.503, p=0.095, n=12 genera); the report calls the correlation suggestive but not significant. ₉ The report notes that field sites vary in pH, redox potential, carbon sources, sulfate, nitrate, and other parameters beyond uranium. An organism may therefore tolerate metals in the lab but lack the metabolic capabilities needed at a particular site. Field competition, cross-feeding, and syntrophy are also absent from single-organism laboratory fitness. ₉

Antimicrobial-resistance cost estimates carry parallel caveats (antimicrobial-resistance-fitness-cost). The amr_fitness_cost report notes that all 25 organisms are lab-adapted strains. Compensatory evolution during laboratory maintenance may have reduced measurable costs below those of wild strains. In the other direction, resistance genes, particularly efflux pumps, can be beneficial in vivo by exporting host antimicrobials (Roux et al. 2015), so the in vitro cost may overestimate ecological cost. These caveats push the estimate in opposite directions. ₁₀ The report also separates its +0.086 relative difference between AMR and non-AMR transposon knockouts in pooled libraries from Melnyk et al. (2015). That meta-analysis of ~600 resistance cost measurements found costs in ~70% of cases, with mean relative fitness 0.91–0.95 (5–9% cost), from isogenic resistant and sensitive strain pairs; the two are different measurement scales. ₁₀ In amr_pangenome_atlas, only 37/48 Fitness Browser organisms had AMR genes, and these were mostly environmental strains in which intrinsic resistance predominates. The fitness analysis therefore does not capture the cost of recently acquired mobile resistance in clinical pathogens. Its 100% identity link threshold avoids paralog confusion but may undercount fitness effects for closely related variants. ₁₁

The pathway_capability_dependency report refines how laboratory fitness is summarized. It treated fitness breadth as a better conservation predictor than fitness magnitude and gave breadth 30% of a composite importance score. It also proposes the hypothesis that the breadth of environmental conditions a species encounters, rather than mean fitness under any single condition, may constrain pathway gene loss, with implications for Black Queen Hypothesis predictions. ₈ This is consistent with the weak breadth-conservation association reported across 43 bacteria, but neither report establishes environmental breadth as a cause of retention; see condition-dependent-gene-tradeoffs. _8,4

Tensions

The central tension is that 24.4% of core genes showed positive fitness when deleted in the laboratory, meaning the intact genes imposed a laboratory burden, yet core genes were more conserved than always-neutral genes, which were 66% core. ₁ The evidence supports an environmental mismatch interpretation: laboratory conditions measure a restricted component of fitness, whereas pangenome conservation integrates selection across broader contexts. ₁

A second tension is that core genes were 1.78x more likely to show strong condition-specific phenotypes, while condition-specific fitness was not equivalent to niche-specific fitness. ₁ Resolving this requires environmental experiments or validated environmental proxies, not further interpretation of laboratory measurements alone. ₁

A related tension is that core genes were more likely to show beneficial deletion effects and strong fitness effects overall, while singleton genes appeared largely neutral. ₄ This supports the role of assay coverage and network context, but does not establish whether singleton neutrality reflects true dispensability or missing measurements. ₄

The DvH results add a scale tension: field-stress genes were more core than heavy-metal-important genes, although both categories were treated as ecologically relevant. ₂ The report hypothesizes that specific metal-resistance mechanisms may be accessory while uranium- and mercury-related responses may involve fundamental stress pathways; this is not a direct mobile-element or natural-selection measurement. ₂ The DvH heavy-metal result also contrasts with the Metal Fitness Atlas finding, cited in counter_ion_effects, that metal-important genes were 87.4% core (OR=2.08). Counter_ion_effects raised possible inflation by a shared-stress component as a pre-correction concern, because osmotic stress genes are core cellular machinery. After correction, the core enrichment persisted and increased for 7 of 14 metals (Mo, W, Hg, Se, Ni, Cr, U), and the report judged the atlas conclusions robust. ₁₂ Organism scope and definitions differ, so the DvH 71.2% and atlas 87.4% figures are not reconciled here. _12,2 Within the DvH report itself, the prose claim that heavy-metals was lowest at every threshold is contradicted by its table at -1.0 (lab-antibiotic 76.2% versus heavy-metals 76.7%). ₂

The 82% essential-core estimate in the broad synthesis and the 86.1% estimate in the 33-organism linkage are an unresolved scope tension because their cohorts and definitions differ. _{1 3} Likewise, DvH shows that adding gene length raised CV-AUC to 0.645, whereas fitness alone reached 0.517–0.548, while broader analyses report a real but weak gradient without length adjustment. _{2 4,1} These results are not contradictory estimates of one model; the unresolved issue is how much of the broad gradient survives explicit adjustment for length, insertion callability, phylogeny, and pangenome coverage.

Open Directions

• Use gene_fitness_conservation.csv for the 2,725 analysed genes, with insertion counts or other callability measures, in nested cross-validated models to test how much of gene length’s CV-AUC 0.645 contribution remains after measurement quality is modelled. ₂
• Add the 678 essential genes through an essentiality-aware missing-fitness or two-stage model, testing whether gene length predicts core status similarly among essential genes and genes with recovered transposon mutants. ₂
• Replace binary core/auxiliary labels with quantitative gene-cluster prevalence and fit length-adjusted models to ask whether length predicts the core boundary or intermediate prevalence. ₂
• Reanalyse thresholds from -1 to -3 with gene length, insertion coverage, and fitness uncertainty as covariates. ₂
• Extend length- and callability-adjusted analysis across the 194,216-gene, 43-bacterium comparison and test whether the weak gradient persists after measurement quality and lineage are controlled. _1,4
• Reanalyse the 33-organism essential-core enrichment with gene length, insertion coverage, pangenome clade size, and quantitative cluster prevalence to test whether the median odds ratio of 1.56 persists. ₃
• Test whether the 28,017 costly-and-conserved and 5,526 costly-and-dispensable genes differ in length, callability, mobile-element association, or prevalence before interpreting them as natural-selection signatures. _{1 5}
• Stratify the 25,271 trade-off genes and both costly groups by length, callability, functional category, and pangenome prevalence to test whether core enrichment persists. ₅
• Combine the 5,526 costly-and-dispensable genes with mobile-element annotations and genome-loss histories to test recent acquisition or ongoing loss. ₁
• Link core trade-off genes to AlphaEarth environmental data and use comparative analysis to ask whether organisms from more variable environments have more trade-off genes in their core. ₁
• Compare RB-TnSeq results across the 43 organisms, identify gene families essential in every organism, and test whether conservation persists across additional environmental conditions. ₁
• Characterize the 48 accessory modules containing co-regulated functions exclusively in the flexible genome across environmental conditions and genetic backgrounds. ₁
• Pair direct fitness assays in soil, biofilms, or host-associated conditions with pangenome conservation to identify laboratory-burdened core genes with context-dependent natural-environment benefits. ₁
• Reassess apparently neutral singleton genes with improved transposon coverage and environmental assays. ₄
• Use Fitness Browser experiments and quantitative gene-cluster prevalence across additional environmentally relevant organisms to test whether the weak field-versus-lab effect generalizes beyond DvH. ₂
• Link DvH resistance-gene genomic context to environmental geochemistry and acquisition histories, testing whether the heavy-metal versus field-stress contrast reflects accessory acquisition or distinct physiological requirements. ₂
• Combine genomic-context analyses with resistance-gene calls to test whether the 73.4% core fraction for lab-antibiotic genes and 71.2% for heavy-metal genes is associated with mobile genetic elements. ₂
• Reanalyse the 52 ICA modules with functional annotation and environmental metadata to test whether the 52 unannotated genes among 239 ecological-module members mediate adaptation. ₂
• Characterize the 3,683 essential-auxiliary and 1,259 essential-unmapped genes with orthology, gene-context, and multi-condition fitness data to test whether they are strain-specific replacements for core functions or poorly annotated mobile and divergent genes. ₃
• Link the 4,346 ENIGMA CORAL field samples with geochemistry and 213,044 ASVs to test ecological associations for DvH or other organisms with environmental data and gene-level fitness measurements. ₂
• Replace binary important/not-important calls with continuous fitness scores (e.g., mean or minimum fitness per condition class) as logistic-regression predictors, as the DvH report proposes. No results from that alternative analysis are reported, so whether it improves on the 0.548 combined-fitness CV-AUC remains open. ₂
• Extend the n=12-genus uranium comparison with site pH, redox potential, carbon sources, sulfate, and nitrate as covariates to test whether lab metal tolerance predicts field abundance once non-metal site variables are modelled. ₉
• Measure resistance costs in wild or minimally passaged isolates alongside the 25 lab-adapted strains to test whether laboratory compensatory evolution lowered the measured costs. ₁₀