Uneven Experimental Condition Coverage Biases Fitness-Based Gene Prioritization

Fitness-based gene prioritization is biased when organisms are tested across different numbers and types of experimental conditions, because organisms with deeper condition coverage have more opportunities to reveal condition-specific phenotypes and can therefore receive higher priority than organisms with shallower coverage. ₁

The issue is distinct from condition-specific-fitness itself: a gene may be genuinely important only under a particular stress or nutrient condition, but the probability of detecting that importance depends on whether the relevant condition was assayed. ₁

Evidence from the functional-dark-matter analysis

The Fitness Browser collection contained 48 organisms, but condition coverage was uneven, with MR-1 having 121 historical conditions. ₁ The report states that organisms with deeper condition coverage can produce more specific phenotypes and receive higher prioritization scores. ₁

The prioritization framework combined six axes: fitness importance, cross-organism conservation, inference quality, pangenome distribution, biogeographic signal, and experimental tractability. ₁ Because fitness importance is derived from observed phenotypes, unequal condition coverage can influence the first axis and thereby affect the combined score even when the underlying genes have comparable biological importance.

Across the 48 organisms, 57,011 of 228,709 genes were classified as dark, and 7,787 showed strong fitness effects, defined as an absolute fitness value of at least 2 in at least one condition. ₁ The number of available conditions determines the opportunity for a dark gene to meet this at-least-one-condition criterion, so comparisons of phenotype-bearing dark genes across organisms are vulnerable to coverage bias.

The top 100 candidates came from 22 organisms, with Shewanella MR-1 contributing 25 candidates, Pseudomonas putida N2C3 contributing 18, and Marinobacter contributing 9. ₁ These counts are prioritization outcomes rather than direct estimates of the organisms’ total functional darkness, because the report also identifies uneven condition coverage as a source of ranking bias.

The recommended experimental campaign illustrates how the bias can become operational: the first three MR-1 experiments—stress, nitrogen source, and carbon source—would address 111 of the top 500 dark genes, or 20.8%, while 10 organism–condition experiments would address 242 of the top 500 dark genes, or 45.3%. ₁ These counts and percentages are copied from the report’s set-cover table, which labels the percentages as ”% of Top 500”. The percentages are not consistent with a 500-gene denominator, and the report does not explain the discrepancy, so they should be treated as internally inconsistent until rechecked. ₁ This set-cover result is useful for campaign design, but it may favor organisms with many previously characterized condition-responsive genes and extensive existing condition profiles.

The Fitness Effects vs Conservation — Quantitative Analysis analysis refines this concern by showing that fitness importance is only a weak predictor of conservation: across approximately 194,000 genes from 43 bacteria, essential genes were 82% core whereas always-neutral genes were 66% core, with breadth of fitness effects associated with core status at Spearman rho=0.086, p=8.1e-230. ₂ Thus, observed fitness activity contains real conservation signal but does not justify treating unobserved or sparsely observed genes as biologically unimportant.

The field-versus-lab analysis provides a within-organism test of this distinction: it classified 757 Desulfovibrio vulgaris Hildenborough experiments as 337 field (44.5%) and 420 lab (55.5%) experiments, and found that condition type was less predictive of conservation than fitness importance. ₃ This refines the coverage-bias claim: unequal coverage affects the opportunity to observe fitness, but the ecological label of an assayed condition is not by itself a reliable substitute for coverage-balanced evidence. Field-stress important genes were 83.6% core and field-core genes 82.4% core, whereas heavy-metal and lab-antibiotic groups were 71.2% and 73.4% core; the report interprets these differences as context-dependent conservation patterns, not as evidence that field conditions universally produce more informative genes. ₃ Field-only, lab-only, and combined fitness predicted core status weakly (cross-validated AUC 0.517, 0.531, and 0.548), while adding gene length reached 0.645, further cautioning against ranking genes from condition coverage alone. ₃

Why the bias matters

Unequal coverage can create a feedback loop in which well-profiled organisms generate more detected fitness effects, receive more candidate slots, and are then selected for additional experiments because their existing profiles make them efficient targets. ₁ Conversely, genes in poorly profiled organisms may remain unprioritized because the relevant environmental or nutritional condition has not been tested, not because the genes lack measurable phenotypes. ₁ The conservation analysis supports this caution: novel singleton genes showed near-zero mean fitness, but the report notes that this apparent neutrality may reflect poor transposon coverage rather than true neutrality. ₂

The new Core Gene Paradox — Why Are Core Genes More Burdensome? analysis supports the claim that detection depends on condition coverage while refining its relationship to pangenome position: core genes were more burdensome than non-core genes in Protein Metabolism, Motility, and RNA Metabolism, with differences of +6.2, +7.8, and +12.9 percentage points, respectively, whereas Cell Wall genes showed the reverse pattern, with a difference of -14.1 percentage points. ₄ It also identified 25,271 true trade-off genes, or 17.8% of genes examined, and found that trade-off genes were 1.29 times more likely to be core than non-core genes (odds ratio 1.29, p=1.2e-44). ₄ Thus, shallow condition panels can miss condition-dependent effects even among conserved genes; condition coverage bias is not limited to accessory or singleton genes.

The existing analysis similarly found that genes tagged with strong condition-specific effects were 77.3% core versus 70.3% without specific phenotypes, with OR=1.78 and p=1.8e-97; 4,450 genes (2.7%) fit an ephemeral-niche pattern of being neutral overall but critical in one condition, and these were more common among core genes (3.0%) than auxiliary genes (1.7%) or singleton genes (1.6%). ₂ This contradicts the expectation that condition-specific or ephemeral-niche genes are predominantly accessory, while supporting the claim that detection depends on which conditions are assayed: core genes can also have condition-specific effects, and shallow panels can miss them.

The bias also affects cross-organism comparisons. Cross-organism fitness concordance identified 65 ortholog groups present in at least three organisms whose dark-gene orthologs showed measurable effects under the same condition classes. ₁ Detectable concordance requires overlapping condition classes and measurable profiles in multiple organisms, so sparse or non-overlapping condition panels can reduce apparent transferability even when homologous genes respond similarly under untested conditions.

The problem is especially relevant to essential dark genes because standard RB-TnSeq, randomly barcoded transposon sequencing, yields no viable knockout fitness profiles for essential genes. ₁ Of 9,557 essential dark genes, prioritization therefore relied on gene-neighbor context, cross-organism conservation, phylogenetic breadth, domain annotations, and CRISPRi tractability rather than directly observed transposon fitness magnitudes. ₁ This creates a modality difference between non-essential genes ranked partly by condition-specific fitness and essential genes ranked largely by indirect evidence.

Interaction with ranking robustness

The report found that overall rank correlations remained greater than 0.93 across six alternative prioritization configurations, but only 64% of the original fitness-active top 50 remained under conservation-dominant or drop-tractability settings. ₁ These sensitivity results show that the broad ranking was stable under tested weight changes, while membership in the highest-priority set remained sensitive to how evidence axes were weighted.

The same analysis found that essential-gene top-50 retention was 36% when tractability was dropped and 48% when neighbor context was dropped. ₁ The different retention rates reinforce that prioritization stability depends on both the evidence available for a gene and the scoring dimensions used to compensate for missing direct fitness measurements.

Condition coverage should therefore be treated as an explicit source of uncertainty rather than as a neutral property of the input data. ₁ Comparisons should distinguish biological evidence from observation opportunity and should report whether a gene was tested under a relevant condition, tested under unrelated conditions, or not tested sufficiently to support a negative conclusion.

Relation to experimental design

The report’s 10-experiment set-cover sequence prioritizes experiments that cover many high-ranking genes, including MR-1 stress, nitrogen-source, and carbon-source experiments; P. putida N2C3 stress and carbon-source experiments; Sinorhizobium meliloti carbon-source and stress experiments; P. stutzeri RCH2 stress; Marinobacter stress; and P. fluorescens GW456-L13 carbon-source experiments. ₁ A coverage-aware design should retain this efficiency objective while also allocating experiments to organisms and condition classes whose current profiles are sparse, so that campaign efficiency does not reinforce the existing observation imbalance. ₁

This consideration complements fitness-module-detection-sensitivity, because more conditions can increase the evidence available for fitness modules while also making organisms appear more information-rich than sparsely tested organisms. ₁ It also complements condition-space-dimensionality, because the biological space of possible stresses, nutrient sources, and environmental states is larger than the subset represented in any particular assay collection. ₁

The field-versus-lab study shows why coverage-aware design should not assume that field labels solve this problem: its 52 ICA fitness modules had mean core fraction 0.886, but module conservation did not significantly correlate with field-condition activity (Spearman rho=0.071, p=0.62). ₃ This supports balancing condition opportunity and fitness evidence rather than using ecological category as a ranking shortcut. The same study found that ENIGMA CORAL contained field samples and community data but no D. vulgaris gene-level fitness data, so apparent field coverage cannot be treated as equivalent to organism-specific perturbation coverage. ₃

The costly-but-conserved result further supports allocating experiments to underprofiled condition classes rather than treating conservation as evidence that additional testing is unnecessary: 28,017 genes were classified as Costly + Conserved, a group interpreted as potentially maintained despite laboratory-measured cost, although this interpretation does not directly measure selection in nature. ₄

Route-dependent organism rankings

The functional-dark-matter report itself attributes the leading organism’s position to observation depth. shewanella-oneidensis MR-1 contributes 25/100 top candidates, which the report calls a consequence of its deep condition coverage (121 conditions) rather than inherent biology. ₁ Elsewhere the report says the 25/100 share is consistent with MR-1’s position as a model organism with extensive condition coverage and a large hypothetical gene complement. ₁

Route A weights condition-specific evidence depth, and it selected MR-1 (587 genes, tractability 0.8) first. The report gives three reasons: deep condition coverage (121 conditions historically), a large dark gene complement spanning all tiers (172 T5 Dawn, 257 T4 Penumbra), and high genetic tractability. Of its dark genes, 544 have specific condition recommendations (stress, nitrogen, carbon). ₁ A greedy weighted set-cover algorithm optimized priority value × tractability × phylogenetic diversity. It selects 42 organisms (28 genera) covering 95% of total priority, ranks MR-1 first, and reaches 80% coverage with 32 organisms. ₁

Route B supports the coverage-bias reading of MR-1’s rank. It ranks Sinorhizobium meliloti first (1,630 OGs, 195 kingdom-level gaps) and MR-1 third (805 OGs, 105 kingdom gaps, tractability 0.8), because Route B weights conservation breadth over condition-specific evidence depth. ₁ The report also cautions about some of the most universally conserved but functionally uncharacterized ortholog groups. These groups may be invisible to Route A because they may lack condition-specific fitness effects; they may be essential under all conditions and so produce no differential fitness signal. ₁

Essential dark genes have no viable transposon mutants. They represent 55% of the experimentally actionable dark matter but score poorly in the fitness-centric framework. ₁ A separate prioritization ranked all 9,557 essential dark genes using five evidence dimensions that do not require fitness magnitudes, with these weights: gene neighbor context (0.25), cross-organism conservation (0.20), phylogenetic breadth (0.20), domain annotations (0.15), and CRISPRi tractability (0.20). CRISPRi is CRISPR-interference knockdown. ₁ This refines the modality difference described above, because it shows that a large share of actionable dark matter falls outside fitness-centric scoring. ₁

Coverage bias also operates at the taxonomic level. The Fitness Browser collection is heavily skewed toward Gammaproteobacteria (21/48 organisms, 78% of top-500 dark genes), and several major bacterial phyla are absent or severely underrepresented. ₁ A later table in the same report lists 417 for Gammaproteobacteria (21 organisms), a count that is not consistent with 78% of the top 500. The report does not reconcile the two, so the 78% figure should be treated as internally inconsistent until it is rechecked. ₁

Experiment-count sensitivity of fitness thresholds

Whether a gene’s fitness effect counts as significant depends on the chosen |fitness| threshold and on the number of experiments. Organisms with fewer carbon source experiments have less statistical power. ₅ This supports the page’s central claim at the level of individual gene calls, not only at the level of aggregated rankings. ₅

synechococcus-elongatus (alias SynE) is the clearest documented case. It has 12 NaCl dose-response experiments spanning 0.5–250 mM, far more than any other organism (1–6 NaCl experiments). That makes the n_sick >= 1 threshold much easier to satisfy; the threshold requires at least one NaCl experiment with fit < -1. _6,7,8 SynE shows an 88.6% shared-stress rate _6,7,8, comprising 565/638 genes ₆. It flags 32.6% of genes as NaCl-important, 3× higher than the next organism. ₈ Excluding SynE lowers overall overlap from 39.8% to 36.7% (3,739/10,183). _6,7,8 The counter-ion report reads this as confirming that the pooled finding is not driven by the outlier. The next-highest organisms (Korea 58.6%, Phaeo 56.4%, Pedo557 53.0%) have typical NaCl experiment counts. ₆

The pitfalls digest gives two corrections for comparing gene-level importance across conditions with unequal experiment counts. Analysts can require n_sick >= 2 or use mean_fit < threshold instead of n_sick >= 1. Alternatively, they can report results with and without outlier organisms as a sensitivity check. The digest calls the drop from 39.8% to 36.7% modest but worth documenting. ₈ The discoveries digest likewise says that analyses using NaCl fitness data should account for experiment count when setting importance thresholds. ₇

Experiment-count bias also appeared in metal-specificity scoring. An initial analysis showed 0% metal-specificity for essential metals (Mo, W, Se, Mn), which seemed to be because DvH’s 608 non-metal experiments made specificity impossible. In fact, a locusId type mismatch had silently excluded DvH (desulfovibrio-vulgaris-hildenborough). After that fix, essential metals showed these metal-specificity values: Manganese (61%), Molybdenum (61%), Tungsten (57%), Selenium (46%). ₇ The digest concludes that the experiment-count bias is real but less severe than initially thought. Its stated 47-61% range conflicts with the listed 46% selenium value, and the source does not reconcile the discrepancy. ₇

Among 149 novel metal candidate families, 45.6% have a dominant specificity of metal-specific, compared to 58.2% for annotated families (Fisher exact OR=0.60, p=0.003). ₉ The report attributes this partly to the composition of the novel set. Many novel candidates were identified in deeply profiled organisms (DvH, Btheta, psRCH2), where the high experiment count provides more opportunities to detect pleiotropic effects, pushing genes toward general sick. ₉ Read alongside MR-1’s coverage-driven prioritization, this suggests the hypothesis that deep profiling cuts both ways. It can raise detection of specific phenotypes for ranking, yet it can also shift genes from specific toward pleiotropic labels in specificity scoring. _1,9

The Fitness Browser contains 559 metal-related experiments covering 16 metals. The metal-fitness atlas reports these as 8.2% of 6,804 total experiments ₁₀, while the metal-specificity report gives 559 (8.6%) of 6,504 total experiments classified ₉. The two percentages differ because their denominators differ; neither report explains why the classified total is smaller than the atlas total. _10,9 Six metals have cross-species coverage (≥3 organisms): Cobalt (27 orgs), Nickel (26), Copper (23), Aluminum (22), Zinc (17), and Iron (3). DvH is the most metal-profiled organism (149 experiments, 13 metals). ₁₀

Across organisms, metal experiment counts range from 3 to 112, which affects the reliability of per-organism matrices. ₁₁ Imbalance within one organism also matters. When metals are a small fraction of total experiments (e.g., 12/176 for MR1), per-module z-normalization across all experiments produces max |z| < 2.0 for metal experiments. This happens because the non-metal majority dominates the z-score denominator, and it can mask metal-responsive modules. ₈ This supports the coupling to fitness-module-detection-sensitivity: the share of experiments in a condition class, not only the absolute number, conditions what can be detected. ₈

Aggregated summary metrics carry the same problem. The ADP1 triple-essentiality report notes that its TnSeq essentiality fraction mixes diverse conditions and that condition-matched comparisons are more informative. ₁₂

Untested and under-tested condition classes

An absent experiment is different from a measured null result: where a condition was never assayed, no inference about gene importance under it is possible. The Caulobacter RB-TnSeq compendium in the Fitness Browser (orgId=Caulo) has 198 experiments. These cover 95 stress experiments (envelope-, drug-, metal-related stresses), 46 nitrogen-source, 42 carbon-source, 10 PYE control, plus 5 others, and zero pure iron-limitation experiments. ₁₃ A separate classification of the same 198 experiments lists envelope=22, oxidative=4, carbon=48 and iron=0. Its carbon count differs from the 42 carbon-source experiments in the compendium description, and the report does not reconcile the two category schemes. ₁₃

As a result, the Fur-released gene sets could be ranked by phenotypic importance only under envelope stress. The iron-limitation arm of H2 could not be tested, and under the plan’s preflight rule it was descoped to exploratory, leaving an envelope-axis-only test; this is an untested arm, not a negative finding. ₁₃ The 22-experiment envelope-stress subset was large enough to compare gene-set phenotype-bearing rates against the genome background (33.25%). ₁₃ To close the iron axis, the report proposes a small targeted set of RB-TnSeq experiments under bipyridyl chelation, ferric salt supplementation, and hemin, or a cross-walk to Leaden 2018’s published Δfur phenotype. These would test whether the Fur-released subset that scores phenotype-bearing under envelope stress also scores under iron limitation. ₁₃

Antibiotic validation shows the same structure. Only 4 AMR classes (beta-lactam, aminoglycoside, chloramphenicol, tetracycline) have matched antibiotic experiments, and the macrolide, glycopeptide, and polymyxin classes could not be validated. ₁₄ For those classes, the missing matched experiments prevent validation of whether resistance genes become fitness-important under the corresponding antibiotic. This is an absence of tests rather than a negative result, and it does not by itself bear on the cost estimates discussed in antimicrobial-resistance-fitness-cost. ₁₄

Metal coverage is similarly uneven. The atlas’s limitations section says cobalt and nickel were tested in 27 organisms, while uranium, chromium, mercury, cadmium, selenium and manganese were tested in only 1-2 organisms. Cross-species patterns for the rare metals therefore reflect DvH and psRCH2 biology. ₁₀ The 27-organism nickel count conflicts with the atlas’s earlier count of 26, and the report does not reconcile the two. ₁₀ Six metals on the USGS critical minerals list are tested in only 1-2 organisms. The report proposes expanding coverage to 5+ organisms each to enable cross-species conservation analysis. ₁₀

Where a metal was tested in only one or two organisms, results are extrapolation-limited. Manganese coverage is 1 organism (DvH), which shows 100% core enrichment (delta=+0.198), and whether this is universal is untested. Tungsten coverage is also 1 organism (DvH), with strong core enrichment (delta=+0.145); whether this extends beyond sulfate reducers is untested. ₁₀ Uranium differs in kind: it was measured, in 2 organisms (DvH, psRCH2), and gave a non-significant result, delta=+0.035 (p=0.34). The report says more organisms would be needed to tell whether uranium fitness genes are truly neutral or the effect was underpowered. ₁₀ Chromium coverage is 2 organisms (DvH, psRCH2). The Oak Ridge FW300 Pseudomonas strains were isolated from Cr-contaminated groundwater but have never been tested against chromate. ₁₀

Some condition classes and organisms are missing altogether. No rare-earth-element (REE) experiments exist in the Fitness Browser. The report proposes, but has not completed, RB-TnSeq experiments with lanthanum chloride (0.1-1 mM) and cerium chloride (0.1-1 mM) in Putida, Cup4G11, Keio and Marino. Comparing the La and Ce profiles would show whether REE tolerance uses a shared or element-specific gene set. ₁₀ Pseudomonas putida KT2440 has fitness matrices in the Fitness Browser but zero metal experiments, and the report calls it the only Fitness Browser organism with no metal data. ₁₀

In the opposite case, prediction runs ahead of testing. In the atlas’s table, Methanococcus maripaludis S2 and JJ each carry 54 novel metal candidate ortholog groups (the accompanying prose calls them candidate genes), yet both have been tested against only 1 metal (iron). The report proposes Ni, Co, Cu, Zn, W and Mo tests, plus an S2–JJ cross-strain comparison, to validate or refute these predictions. ₁₀ The report also projects what the proposed experiments would achieve. The atlas would expand from 14 to 16+ metals, from 24 to 27+ organisms with fitness data, and from 1-2 to 5+ organisms for critical metals. The 149 novel gene candidates could also be validated directly. These are conditional, prospective impacts, not observed results, and the 14-metal baseline differs from the 16 metals counted elsewhere in the same report. ₁₀

Lab-convenience bias in condition panels

Several projects independently describe the condition panels as skewed toward the laboratory:

• Lab conditions capture only a fraction of the environmental conditions bacteria face in nature, and Fitness Browser condition types are biased toward what is experimentally convenient rather than what is ecologically relevant. ₄
• Lab conditions are biased toward rich media and standard stresses, leaving many ecological niches unrepresented. ₂
• Condition-specific phenotype data are biased toward lab-testable conditions. ₁₅
• Lab media may not reflect the full range of conditions under which pathways are expressed, which may inflate the latent fraction. ₁₆

Each report states these as caveats rather than measuring them, so the size of the ecological under-coverage remains unquantified. _4,2,15,16

Which conditions reveal dark-gene phenotypes

In the functional-dark-matter analysis, stress conditions (metals, oxidative, osmotic) dominate among dark genes with strong fitness effects. Carbon source utilization and nitrogen source utilization follow. ₁

The truly-dark-genes analysis compared truly dark genes with annotation-lag genes among those with strong fitness (|f| ≥ 2). In its Results section, truly dark genes were enriched in mixed community (7.5% vs 0%), iron (0.7% vs 0%), nutrient time-series and rich media, and depleted in stress (43.3% vs 54.7%), carbon source (13.7% vs 21.7%) and motility (1.2% vs 3.2%). ₁₇ The report’s Key Findings section instead gives the stress depletion as 28.7% vs 43.2% (OR = 0.53, p < 0.001); the two pairs of stress percentages are not reconciled in the report. ₁₇ The discoveries digest describes the stress depletion as contrary to expectation, with OR=0.53, p<0.001. ₇

Tensions

The prioritization system has a practical tension between exploiting existing evidence and correcting for unequal observation. ₁ Favoring organisms with many informative profiles improves immediate experimental yield, as illustrated by the 242-of-500 coverage achieved by 10 organism–condition experiments, but may underexplore genes from organisms whose condition panels are shallow. ₁ The report does not establish that the highest-ranked organisms have intrinsically more dark-gene biology, so this distinction should not be inferred from their candidate counts alone. ₁

The conservation analysis introduces a related interpretive tension rather than resolving it: core genes showed stronger fitness effects in both negative and positive directions, while novel singleton genes were near-neutral under tested laboratory conditions. ₂ The former indicates that condition-specific detection is not confined to accessory genes, whereas the latter may reflect poor transposon coverage; neither result establishes how untested ecological conditions would reorder priorities.

The core-gene trade-off analysis refines this tension rather than resolving it: burden was higher for core genes in several functional categories but higher for non-core genes in Cell Wall genes. ₄ This category-specific reversal cautions against applying a single core-versus-accessory correction to fitness-based rankings.

The field-versus-lab result adds a qualification rather than resolving the tension: in one organism, fitness importance predicted conservation more consistently than field-versus-lab classification, but this does not show that unequal condition coverage is harmless. ₃ It remains unresolved whether coverage-adjusted rankings would change substantially across organisms with different condition panels and accessory-genome structures.

The sources also differ on stress dominance versus stress depletion. One report says stress conditions dominate among dark genes with strong fitness effects ₁. Another finds truly dark genes depleted in stress relative to annotation-lag genes (43.3% vs 54.7% in its Results; 28.7% vs 43.2% in its Key Findings) ₁₇, with OR=0.53, p<0.001 ₇. The two need not conflict: one describes the overall composition and the other contrasts two dark-gene subsets. However, neither analysis is normalized by condition coverage. That leaves open the hypothesis that stress dominance partly reflects how numerous stress experiments are, as in the 95 stress experiments of the Caulobacter compendium. _1,17,13

The direction of the deep-profiling effect is also contested across projects. One report attributes MR-1’s 25/100 top candidates to its 121 conditions ₁. Another attributes the lower metal-specific fraction among novel families (45.6% versus 58.2%) to deep profiling raising pleiotropic calls ₉. The corrected DvH result shows that 608 non-metal experiments did not drive essential-metal specificity to zero ₇. Whether deep coverage inflates specific phenotypes, pleiotropic phenotypes, or neither depends on the scoring rule, and no analysis here tests the three cases under a common threshold. _1,9,7

Open Directions

• Use the Fitness Browser condition metadata to construct organism-by-condition coverage matrices, then apply inverse-probability weighting or matched-depth resampling to test whether candidate rankings change after equalizing condition opportunity. ₁
• Recompute the 7,787 strong-fitness-effect count and the top-100 ranking after downsampling MR-1’s 121 historical conditions to the coverage of less-profiled organisms, asking how much of MR-1’s prioritization advantage is attributable to condition depth. ₁
• Compare condition-class-balanced rankings for stress, carbon-source, and nitrogen-source assays, asking whether cross-organism concordance among the 65 ortholog groups changes when only shared condition classes are evaluated. ₁
• Add a coverage penalty or an explicit untested-condition uncertainty term to the six-axis score, then compare top-50 retention against the reported 64% fitness-active retention under conservation-dominant or drop-tractability settings. ₁
• Design targeted RB-TnSeq or CRISPRi experiments for high-priority genes from poorly profiled organisms, asking whether missing condition coverage explains their absence from the current top candidates rather than a lack of measurable phenotype. ₁
• Stratify the coverage-adjusted analysis by core, auxiliary, and singleton status, asking whether the observed 77.3% versus 70.3% condition-specific phenotype contrast persists after matched condition depth and transposon-callability controls. ₂
• Replicate the field-versus-lab comparison across organisms with Fitness Browser data and unequal ecological condition panels, testing whether the weak field/lab AUCs and the stronger gene-length-inclusive model persist after matched condition depth and callability controls. ₃
• Partition the coverage-adjusted analysis by functional category, testing whether the core-gene burden advantage in Protein Metabolism, Motility, and RNA Metabolism and the Cell Wall reversal persist after matched condition depth. ₄
• Apply the pitfalls-recommended n_sick >= 2 or mean_fit < threshold rules to every Fitness Browser condition class, not only NaCl, and report with-and-without-outlier results, asking whether outliers like SynE’s 88.6% shared-stress rate recur in other classes. _8,6
• Downsample MR-1’s conditions and rerun both Route A and Route B, asking whether MR-1’s first-versus-third rank gap closes when condition depth is equalized. ₁
• Recheck the set-cover table’s percentage denominator and the Gammaproteobacteria top-500 share against the underlying candidate lists, so that campaign-yield and taxonomic-skew figures rest on one consistent denominator. ₁
• Run RB-TnSeq under bipyridyl chelation, ferric salt supplementation, and hemin in Caulobacter to test whether the Fur-released subset that is phenotype-bearing under the 22 envelope-stress experiments also scores under iron limitation. ₁₃
• Extend manganese, tungsten, uranium, and chromium experiments to 5+ organisms each, testing whether the DvH-only core-enrichment deltas persist and whether the non-significant uranium delta reflects neutrality or low power. ₁₀
• Normalize the truly-dark condition-class enrichments by per-class experiment counts, asking whether stress depletion (OR=0.53) and iron or mixed-community enrichment survive coverage adjustment. _17,7