Triple Essentiality Concordance Analysis for *Acinetobacter baylyi* ADP1 shows that transposon-based assays and complete-gene knockouts do not measure identical phenotypes. RB-TnSeq (random barcode transposon sequencing) reports the fitness of insertion mutants. A complete knockout instead tests whether removing the entire gene causes lethality under the assay conditions. 1 This distinction places the result within gene-essentiality and condition-space-dimensionality. 1 Evidence from Fitness Effects vs Conservation — Quantitative Analysis supports treating fitness as a graded phenotype. Across approximately 194,000 genes from 43 bacteria, fitness importance and pangenome conservation were positively associated, but the relationship was weak, so fitness importance is not a binary or universal proxy for gene status. 2 The synthesis supports this interpretation across 194,216 protein-coding genes from 43 diverse bacteria. Essential genes were 82% core, whereas always-neutral genes were 66% core, a quantitative but modest conservation gradient rather than a binary essentiality boundary. 3 The linked conservation_fitness report refines that synthesis with a separate 33-organism analysis. Essential genes were 86.1% core versus 81.2% for non-essential genes, with a median odds ratio of 1.56. In 18 of 33 organisms the enrichment was significant after Benjamini-Hochberg false discovery rate correction, a procedure that controls the expected share of false positives among significant results. 4 This modest enrichment supports conservation as a graded correlate of viability requirements, not as an assay-independent definition of essentiality. 4 Evidence from Core Gene Paradox — Why Are Core Genes More Burdensome? further supports this distinction. In that analysis, 25,271 genes, or 17.8% of those examined, were trade-off genes, with fitness below -1 in some conditions and above 1 in others. 5
Core evidence
At an RB-TnSeq essentiality threshold of 0.05 in rich medium, the comparison covered 1,933 genes. Only 18 genes (0.9%) were called essential by both RB-TnSeq and knockout assays, while 1,411 genes (73.0%) were called dispensable by both. The two discordant groups were larger. 211 genes (10.9%) were knockout-essential but RB-TnSeq-dispensable, which the report calls false negatives. 293 genes (15.2%) were knockout-dispensable but RB-TnSeq-essential, which it calls false positives. 1
Agreement was poor at every tested RB-TnSeq threshold. At thresholds of 0.01, 0.025, 0.05, 0.10, and 0.20, Cohen’s kappa was -0.139, -0.122, -0.081, -0.024, and -0.014, respectively. 1 At the 0.05 threshold the counts were True Positives = 18, False Positives = 293, False Negatives = 211, and True Negatives = 1,411. At that threshold, recall was 7.9%, precision was 5.8%, specificity was 82.8%, F1 was 0.067, and Cohen’s kappa was -0.081. 1 The five thresholds were tested without formal multiple-testing correction. The report argues that because Cohen’s kappa was negative at every threshold, correction would not change the conclusion of systematic disagreement with knockout calls. 1
The refined analysis added multiple RB-TnSeq thresholds spanning 0.01–0.20. It compared binary vectors at 5 thresholds (0.01, 0.025, 0.05, 0.10, and 0.20) with continuous essentiality_fraction and continuous fitness. Fitness was inverted for ROC analysis (fitness × -1). 1 The thresholds were chosen from the essentiality_fraction distribution, which had mean=0.026 and median=0.0. None of them produced a positive kappa against knockout data, so the choice of threshold did not resolve the TnSeq–knockout discordance. 1 The rich-medium threshold table covered N=1,933 genes. Its best tabulated F1 was 0.077 at threshold 0.025, with 15.3% recall and 5.1% precision. At the other thresholds, F1 was 0.076 at 0.01, 0.067 at 0.05, 0.039 at 0.10, and 0.008 at 0.20. 1
The knockout reference is a binary viability endpoint. It is measured by complete gene deletion and viable colony formation, and it asks whether the organism can survive without the gene. 1 The knockout ground truth merged essentiality_minimal and essentiality_lb (3,092 genes: 499 essential). The merge prioritized essentiality_minimal and fell back to essentiality_lb, so the minimal-media label mixes in rich-medium calls. Keep this in mind when interpreting the 3,092-gene, 499-essential set. 1 The report’s original analysis had no direct lethality measurements and used TnSeq as a proxy for lethality. 1
The report frames the discordance as a difference in what each method measures. In its framing, FBA (flux balance analysis, a constraint-based metabolic model) captures metabolic necessity, knockouts measure lethality, and TnSeq measures fitness cost. Proteomics indicates expression requirement, and growth measurements capture condition-specific optimization. 1 On this basis it concludes that RB-TnSeq systematically disagrees with knockouts (κ<0 across all thresholds) because TnSeq measures fitness cost rather than lethality. As evidence it cites the 211 knockout-essential genes that were TnSeq-dispensable at the 0.05 threshold against rich-media knockout calls among 1,933 genes. 1 The report treats knockout experiments as the best proxy for lethality. It treats fitness, expression, and growth rates as complementary evidence that reveals additional biology, which is an interpretive framing rather than a separately tested result. 1
The cross-organism pangenome analysis supports the direction of this result while limiting how far it generalizes. It identified 27,693 putative essential genes among 148,826 protein-coding genes across 33 organisms. Its essentiality calls were inferred from RB-TnSeq under the library construction and growth conditions represented in the Fitness Browser. 4 Essentiality calls therefore remain condition- and assay-specific, and their apparent conservation advantage cannot by itself establish complete-knockout lethality. 4
Continuous fitness is more informative than a binary call
Continuous RB-TnSeq fitness identified knockout-essential genes better than the essentiality fraction did. AUC (Area Under the ROC Curve) is a threshold-independent classification measure. Inverted fitness produced an AUC of 0.700 in rich medium (1,933 genes) and 0.725 in minimal medium (2,011 genes). Essentiality fraction produced AUC values of 0.344 and 0.403 in the same media, which the report labels worse than random. Proteomics (log2) in minimal medium had an AUC of 0.743. 1 The report summarizes the contrast as AUC=0.70-0.73 for continuous fitness versus AUC=0.34-0.40 (AUC<0.5) for essentiality_fraction. These are the report’s stated ranges, not recalculated values. 1 It stresses that binary essentiality_fraction classifications systematically disagree with knockout experiments (κ<0), while continuous fitness values show fair prediction (AUC=0.70-0.73). It recommends using continuous fitness scores rather than binary classifications for TnSeq analysis. 1
The discordant groups also differed in their continuous measurements. All four classes come from comparing the 0.05 RB-TnSeq threshold with rich-medium knockout calls across 1,933 genes. 1 The 211 knockout-essential/RB-TnSeq-dispensable genes (10.9%) had mean essentiality fraction = 0.0034 (standard deviation = 0.010) and mean fitness = -0.79 (standard deviation = 0.93). 1 The 293 knockout-dispensable/RB-TnSeq-essential genes (15.2%) had mean essentiality fraction = 0.1023 (standard deviation = 0.118) and mean fitness = -0.34 (standard deviation = 0.66). 1 The 18 genes (0.9%) called essential by both assays had mean essentiality fraction = 0.0969 (standard deviation = 0.041) and mean fitness = -1.18 (standard deviation = 1.49). 1 The 1,411 genes (73.0%) called dispensable by both assays had mean essentiality fraction = 0.0116 (standard deviation = 0.016) and mean fitness = -0.32 (standard deviation = 0.60). 1
These results support treating fitness as a graded phenotype rather than converting it immediately into an essential/dispensable label. 1 They also refine the broader gene-essentiality distinction between predicting lethality and predicting the magnitude of growth impairment. 1 The conservation analysis supports this refinement but also limits how far it can be read. Essential genes were 82% core, compared with 66% for always-neutral genes, while the overall association between fitness breadth and core status was weak (Spearman rho=0.086, p=8.1e-230). 2 Stronger or broader measured fitness effects can therefore align with conservation without making conservation a strong predictor of assay-defined essentiality. 2 The synthesis further supports this graded interpretation. It found that core genes were 1.78x more likely than accessory genes to have strong condition-specific phenotypes, rather than being uniformly essential. 3
The core-gene trade-off analysis further supports keeping continuous, condition-specific measurements rather than treating conserved genes as uniformly essential. It found that trade-off genes were 1.29x more likely to be core (odds ratio 1.29, p=1.2e-44). 5 It also found that core-gene burden depended on function. Core genes were more burdensome in Protein Metabolism, Motility, and RNA Metabolism, while non-core genes were more burdensome among Cell Wall genes. The reported differences were +6.2, +7.8, +12.9, and -14.1 percentage points, respectively. 5 These results refine the conservation comparison by showing that core status can coexist with substantial, reversible laboratory costs rather than implying universal essentiality. 5 The synthesis supports this refinement: 24.4% of core genes were beneficial when deleted, compared with 19.9% of accessory genes, so conservation did not identify an inert housekeeping genome. 3
The new pangenome linkage further supports using conservation as contextual evidence rather than a binary surrogate. In the 33-organism analysis, essential-core genes were 41.9% enzymes and 13.0% hypothetical. Essential-auxiliary genes were 13.4% enzymes and 38.2% hypothetical, and essential-unmapped genes were 44.7% hypothetical. 4 This distribution suggests that some RB-TnSeq-defined essential functions are poorly characterized or strain-specific. It does not resolve whether their insertion phenotype would match complete-gene deletion. 4
Why the assays can disagree
The report reads the pattern as consistent with insertion mutants retaining partial function, whereas complete deletion removes the entire gene. It states that transposon insertion ≠ gene deletion. 1 It proposes several mechanisms by which RB-TnSeq mutants could appear dispensable when complete knockouts are lethal: partial or truncated protein production, read-through transcription, retained functional domains, and aggregation across conditions. 1 Specifically, it proposes that partial function from transposon-disrupted alleles explains the 211 knockout-essential/TnSeq-dispensable genes (mean essentiality_fraction = 0.0034). 1 For the reverse discordance, it proposes that the 293 TnSeq-essential but knockout-dispensable genes (mean essentiality_fraction = 0.10) are fitness-important but not lethal. In this reading, many genes cause fitness defects (slower growth) without being lethal (zero growth). 1 All of these explanations are hypotheses in this analysis, not demonstrated mechanisms. 1
The conservation analysis refines this interpretation. Core genes had heavier tails in both negative and positive fitness effects, and 24.4% of core genes versus 19.9% of auxiliary genes were ever beneficial when deleted. 2 This does not establish a mechanism for any individual assay disagreement. It does show that large effects in either direction occur in conserved genes, so condition-dependent costs should not be attributed to accessory genes alone. 2 Consistently, the core-gene trade-off analysis classified 28,017 genes as Costly + Conserved and 5,526 as Costly + Dispensable. It interprets the former as genes maintained despite laboratory cost, while cautioning that this is an inference from laboratory fitness and conservation rather than a direct measurement of natural selection. 5 The synthesis supports this caution while refining the selection interpretation: costly-and-conserved genes are evidence for purifying selection in natural environments, not direct measurements of natural-environment fitness. 3
The conservation analysis supports the measurement-bias interpretation by noting that its essential-gene definition is an upper bound. Missing transposon fitness data can reflect short genes, low-complexity regions, or scaffold-edge placement rather than essentiality. Essential genes were also slightly shorter on average, which suggests some insertion bias. 4 The same analysis found that pangenome coverage and clade size affect the core-versus-auxiliary classification, including trivially high core fractions in clades containing only 2 genomes. 4 These limitations are further reasons not to read an RB-TnSeq essentiality call, or its association with core status, as a direct complete-knockout phenotype. 4
Relation to FBA and growth assays
The assay disagreement is not equivalent to a general failure of every essentiality predictor. 1 In the refined analysis, FBA and knockout data showed moderate concordance. In rich medium, across 724 genes, recall = 60.8%, precision = 64.0%, specificity = 79.7%, F1 = 0.624, and Cohen’s kappa = 0.486. In minimal medium, across 833 genes, recall = 65.6%, precision = 69.2%, specificity = 78.9%, F1 = 0.673, and Cohen’s kappa = 0.493. 1 The report summarizes this as moderate FBA–knockout concordance (κ≈0.49, F1≈0.62-0.67). It nonetheless notes that FBA does not predict growth defects among TnSeq-dispensable genes (p=0.63), a result from the original study. 1
Among 478 genes that were all TnSeq-dispensable, FBA class did not predict quantitative growth-defect status (chi-squared = 0.93, p = 0.63, 2 df). A Kruskal-Wallis test on mean growth rates was also not significant (H = 1.67, p = 0.43). 1 FBA therefore performs moderately at the lethal-versus-dispensable boundary, but in this subset it could not explain how severe the growth defects were among transposon-dispensable genes. 1
This null result has a built-in sampling limitation. All 478 triple-covered genes (TnSeq + FBA + growth data) are TnSeq-dispensable on minimal media. TnSeq-essential genes have no viable deletion mutants, so they cannot have growth-rate measurements. The analysis can therefore only ask whether FBA predicts which dispensable genes matter for growth. 1 Among the 78 FBA-essential but TnSeq-dispensable genes, 57/78 (73%) showed growth defects, matching the background rate (72% of all 478 genes). The remaining 21/78 (27%) grew normally, which the report reads as FBA over-prediction. 1 Because the growth-defect rate was uniform across all FBA classes, growth measurements could not tell FBA-essential predictions apart from FBA-blocked ones. They therefore did not resolve FBA–TnSeq disagreement in this selected sample. 1 The original study also reported aromatic-degradation enrichment among discordant genes (odds ratio 9.7). 1
Condition dependence
Assay discordance is partly condition-dependent, because mutant growth phenotypes vary across environments. 1 Across eight carbon sources, 333 of 478 genes (70%) showed condition-specific growth defects. 10 genes (2%) showed defects in all eight conditions, and 135 genes (28%) showed no defect in any condition. 1 Mean pairwise defect correlation was 0.38, with a range of -0.03 to 1.0. 1 These findings support condition-specific-fitness. An assay comparison made across mismatched or aggregated conditions can conflate perturbation modality with environmental dependence. 1
The TnSeq data in this comparison came from the kescience-fitnessbrowser (Fitness Browser), which spans diverse conditions across many studies. The essentiality_fraction aggregates across those conditions. 1 The report therefore cautions that essentiality fraction mixes diverse conditions and that condition-matched comparisons may be more informative. 1 It proposes running TnSeq in the same media as the knockout experiments. That would test whether condition matching improves concordance and would separate condition effects from method effects. This remains a proposal, not a completed analysis. 1
The conservation analysis supports keeping this condition dependence. Genes with strong specificphenotype annotations were 77.3% core versus 70.3% without such annotations. 4,450 genes (2.7%) were neutral overall but critical in one condition. 2 These conditional effects were more common among core genes (3.0%) than among auxiliary genes (1.7%) or singleton genes (1.6%), so condition-specific behavior is not restricted to accessory genes. 2 This is consistent with the conclusion that aggregating conditions can hide phenotype differences. It also cautions that laboratory conditions do not cover all ecological niches. 2 The core-gene trade-off analysis reinforces this point. It found that the burden paradox reverses across functional categories, and it interprets conserved, costly genes as conditionally valuable rather than universally essential. 5 The synthesis supports this interpretation. It identified 1,116 co-regulated fitness modules across 32 organisms; 86% of their genes were core versus an 81.5% baseline, and 59% of modules were more than 90% core genes. 3 Because effects are organized into conserved modules, context-dependent fitness should not be reduced to isolated binary essentiality calls. 3
The new analysis supports making condition matching a priority. Its RB-TnSeq essentiality calls came from the growth conditions used for library construction, so genes essential only under stress were not captured. 4 Its proposed extension, testing genes with fitness < -2 under stress conditions, directly asks whether conditionally important genes show different conservation patterns. 4
A separate conditional-essentiality prediction likewise depends on complete deletion as the decisive test. Condition-Specific Respiratory Chain Wiring in ADP1 predicts that deleting ndh 2 makes complex-i essential on glucose. That prediction remains untested. The report states that it can only be tested experimentally and that an NDH-2 knockout in ADP1 would be the definitive test. 6
Tensions
The data support two conclusions that must be kept separate. Continuous transposon fitness contains useful information about knockout lethality, but thresholded RB-TnSeq essentiality agrees poorly with complete-knockout calls. 1 The first conclusion rests on AUC values of 0.700 and 0.725 for inverted fitness. The second rests on Cohen’s kappa of -0.081 at the 0.05 threshold, with negative kappa at every tested threshold. Continuous essentiality fraction also had a rich-medium AUC of 0.344, a threshold-independent result. 1 The report’s resolution is to distinguish graded fitness prediction from binary assay concordance, rather than treating either assay as a universal definition of essentiality. 1
The report is also internally inconsistent about which threshold is optimal. Its concordance breakdown uses cutoff 0.05, whereas its threshold table labels 0.025, with 15.3% recall, as optimal for F1. 1 One passage states that only 7.9% (18/229) of knockout-essential genes are caught by TnSeq at the optimal threshold. However, 7.9% is the recall reported at 0.05, not at the table’s optimal-F1 threshold. The same passage notes that the original study analyzed only TnSeq-dispensable genes (n=478). 1 Because no tested threshold produced a positive kappa, this inconsistency affects which threshold is called optimal but not the conclusion of systematic discordance. 1
The conservation evidence supports the graded-fitness conclusion but adds a qualification rather than a contradiction. The conservation gradient was statistically robust yet weak, and singleton neutrality may reflect poor transposon coverage rather than true neutrality. 2 Conservation therefore cannot by itself resolve whether an apparently neutral insertion mutant retains function or simply lacks a reliable fitness measurement. 2
The core-gene trade-off evidence refines the interpretation of costly conserved genes but does not directly establish natural selection. Its Costly + Conserved category is inferred from laboratory burden and conservation, whereas the assay-discordance evidence shows that laboratory measurements can depend on perturbation modality and condition. 5 Costly conservation should therefore be treated as a selection hypothesis that needs ecological validation, not as direct evidence that a gene is essential in nature. 5 The new synthesis supports this tension. It identifies 28,017 genes as both costly and conserved and 5,526 as costly and dispensable, but explicitly treats these categories as selection signatures rather than direct measurements of natural fitness. 3
The two conservation analyses should also not be conflated. One reports 82% core for essential genes versus 66% for always-neutral genes across 43 bacteria. The new 33-organism linkage reports 86.1% versus 81.2% for essential versus non-essential genes. 2,4 The results point the same way but use different category definitions and organism sets, and neither resolves the poor binary agreement between insertion and complete-knockout assays. 2,4
Open Directions
- Perform condition-matched RB-TnSeq and complete-knockout experiments, then test whether discordance remains after aligning media and growth conditions. 1
- Analyze insertion positions, gene domains, read-through transcription, and truncated-protein potential for the 211 knockout-essential/RB-TnSeq-dispensable genes to test the proposed partial-function mechanisms. 1
- Compare continuous fitness, essentiality fraction, knockout calls, and growth rates with a preregistered threshold-selection and multiple-testing procedure to determine which representation generalizes across conditions. 1
- Fit combined FBA-plus-fitness-plus-proteomics models and evaluate whether multi-omic predictors improve knockout-essentiality classification over any single assay. 1
- Stratify the condition-matched comparison by core, auxiliary, and singleton status, while testing insertion coverage, to determine whether the weak conservation gradient reflects biological buffering or measurement bias. 2
- Test the 28,017 Costly + Conserved genes in ecologically matched environments and with complete knockouts to distinguish natural selection for conditionally valuable functions from assay-specific burden. 5
- Test whether the 1,116 conserved fitness modules remain coordinated in condition-matched complete-knockout experiments, distinguishing module-level biology from insertion-specific effects. 3
- Extend the 33-organism linkage from binary essentiality to fitness < -2 under stress conditions, and test whether conditionally important genes have distinct core, auxiliary, and unmapped conservation patterns. 4
- Reanalyze the 3,683 essential-auxiliary genes with ortholog data and complete knockouts to determine whether they compensate for missing core functions or reflect insertion-specific measurement artifacts. 4
- Construct an NDH-2 deletion mutant in ADP1 and measure Complex I essentiality on glucose by complete knockout, testing the untested conditional-essentiality prediction. 6