Core claim
Laboratory-measured pairwise co-fitness showed only a weak positive relationship with gene co-occurrence across bacterial pangenomes, whereas coordinated multi-gene modules identified by independent component analysis (ICA) showed a stronger co-inheritance signal. 1 This supports cofitness-network-architecture by indicating that multi-gene organization may be more informative than isolated pairwise links for understanding which genes travel together through a pangenome. 1
The study integrated Fitness Browser co-fitness measurements with pangenome presence/absence data, phylogenetic distances, ICA modules, and SEED functional annotations; its full project summary is Co-fitness and Co-inheritance in Bacterial Pangenomes. 1 The prevalence-ceiling analysis refines this integration: nearly universal genes provide too little presence/absence variation for reliable co-occurrence inference, so weak associations may reflect limited detectability rather than weak biological coupling. 1
The linked Conservation vs Fitness — Linking FB Genes to Pangenome Clusters analysis provides a complementary conservation benchmark: among 33 organisms, essential genes were 86.1% core versus 81.2% for 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. 2 This supports the interpretation that fitness-linked genomic architecture is generally core-conserved, while refining it by showing that essentiality predicts only a modest shift against an already core-rich background. 2
Pairwise signal and the prevalence ceiling
Across 9 organisms, 2,253,491 cofit pairs were compared with 22,534,910 prevalence-matched random pairs. 1 The mean delta phi, defined as cofit-pair co-occurrence minus matched-random co-occurrence, was +0.011 across organisms, with 7 of 9 organisms showing positive effects and 8 of 9 significant at p<0.05 by two-sided Mann-Whitney testing. 1 The aggregate effect was delta=+0.003 with Mann-Whitney p=1.66e-29, but the Wilcoxon signed-rank test across organisms was not significant (W=9, p=0.13), demonstrating high inter-organism variance. 1
Methodologically, cofit pairs were mapped to pangenome cluster pairs through fb_pangenome_link.tsv, deduplicated, and scored with phi coefficients computed from binary genome x cluster presence vectors. 10 prevalence-matched random pairs were generated per cofit pair, matching each cluster’s prevalence independently (tolerance +/-5%), and adjacency was defined from Fitness Browser gene coordinates as within 5 loci on the same scaffold. Overall mean phi was 0.092 for cofit pairs and 0.089 for random pairs, and the organism-level Wilcoxon signed-rank test of delta > 0 gave W=9, p=0.13, a null result. 1
The organism-level results ranged from Ddia6719 at +0.093 to Korea at -0.042, with +0.026 for pseudo3_N2E3, +0.009 for Phaeo, +0.006 for SyringaeB728a, +0.003 for Koxy, +0.002 for Smeli, +0.001 for Btheta, and -0.000 for Putida. 1 Korea’s negative result is difficult to interpret because 95.2% of its cofit pairs had NaN phi values, only approximately 8,000 of 166,601 pairs were computable, and both genes were present in 100% of 72 genomes for the affected pairs, producing zero-variance vectors. 1 The report interprets Korea’s negative delta as statistical noise in this small effective sample rather than a biological signal. 1
In the pairwise organism table, Ddia6719 had the largest positive delta, listed with a count of 16,957, cofit and random phi of 0.182 and 0.089, delta +0.093, and p<1e-300; Putida showed no significant difference (205,323; phi 0.171 versus 0.171; delta -0.000; p=0.41); and Korea was negative (7,994; phi 0.447 versus 0.489; delta -0.042; p<1e-6). 1
A prevalence ceiling occurs when gene clusters occur in nearly every genome: the phi coefficient, measuring association between two binary presence/absence patterns, can then approach zero for both cofit and random pairs. 1 Most Fitness Browser genes mapped to core clusters (>95% prevalence), where phi approaches 0 for both groups. 1 Co-fitness strength was weakly anti-correlated with co-occurrence across 1.04M pairs (Spearman rho=-0.109, p<1e-300), a pattern attributed to strong co-fitness pairs often being core genes with near-universal prevalence. 1
Only 0.7% of cofit pairs were genomically adjacent within 5 genes, and excluding adjacent pairs did not change the result pattern, so operon proximity did not explain the weak pairwise association. 1 The analysis therefore supports interpreting delta phi together with prevalence distributions, core-versus-auxiliary status, and the amount of computable data rather than treating raw phi as directly comparable across species. 1 Restricting primary comparisons to auxiliary pairs in which both clusters are below 95% prevalence would target the variable portion of the pangenome, although it would exclude genuinely coupled nearly universal genes. 1
This limitation interacts with phylogenetic-confounding-of-pangenome-associations: cofit-pair phi was higher among near genomes than medium-distance genomes (mean=0.102 versus mean=0.067), most species lacked genomes in the far stratum (>0.05 branch distance), and phylogenetic stratification was available for 7 of 9 organisms. 1 Thus, insufficient prevalence variation and incomplete phylogenetic separation both constrain interpretation. 1 Near-clonal organisms nevertheless retained detectable signals: Ddia6719 had ANI 99.47% and delta=+0.093, while pseudo3_N2E3 had ANI 99.66% and delta=+0.026. 1
ICA modules show stronger co-inheritance
Across 195 ICA modules in 6 organisms, within-module co-occurrence had mean phi=0.229, compared with a prevalence-matched null mean of 0.177 from 1000 permutations, for delta=+0.053. 1 A total of 51/195 modules (26%) were significant at p<0.05, and 21/195 (11%) remained significant at q<0.05 after Benjamini-Hochberg FDR correction, where FDR means false discovery rate. 1 This module-level result supports the conclusion that coordinated multi-gene structure captures a stronger pangenome inheritance signal than pairwise co-fitness, even when prevalence ceilings weaken individual-pair inference. 1
The module test covered the 6 organisms with ICA module data (Koxy, Btheta, Putida, Korea, Phaeo, pseudo3_N2E3): mean pairwise phi within each module was compared with 1000 prevalence-matched random gene sets, and p-values were corrected with Benjamini-Hochberg FDR. 1 The ICA fitness modules used are the co-regulated gene-module dataset produced by the fitness_modules project. 1
The organism-level module results were Koxy: 44 modules, 14 (32%) significant, mean phi=0.079, null mean=0.040; Btheta: 36, 22 (61%), 0.192, 0.086; Putida: 38, 6 (16%), 0.266, 0.202; Korea: 29, 0 (0%), 0.357, 0.308; Phaeo: 37, 3 (8%), 0.108, 0.095; and pseudo3_N2E3: 40, 4 (10%), 0.444, 0.409. 1 Btheta showed the strongest module-level signal, whereas Korea had no significant modules because all Korea modules were greater than 90% core with prevalence near 1.0. 1 The results support condition-space-dimensionality only indirectly: variation is organism-dependent, but environmental condition space was not tested as a causal explanation. 1
The strongest enrichment occurred in accessory modules, defined as modules with less than 50% core genes: mean delta phi +0.108, with 8/11 (73%) significant at p<0.05 and 4/11 (36%) at FDR q<0.05. 1 Core modules, defined as more than 90% core genes, had mean delta +0.059, with 29/120 (24%) significant at p<0.05 and 13/120 (11%) at q<0.05. 1 Mixed modules, defined as 50–90% core genes, had mean delta +0.031, with 14/64 (22%) significant at p<0.05 and 4/64 (6%) at q<0.05. 1 The accessory-versus-core difference trended toward significance by Mann-Whitney testing (p=0.051). 1 Because that contrast did not meet a p<0.05 threshold, the accessory advantage over core modules is suggestive rather than established. 1
The Metabolic Capability vs Dependency project cites co-inheritance with co-fitness in 73% of accessory modules as validation for pathway-level fitness aggregation. 3 This reuse depends on a small accessory sample: the 73% appears to correspond to the 8/11 accessory modules significant at p<0.05, of which only 4/11 (36%) remained significant at FDR q<0.05, so the validation should be weighted accordingly. 1,3
This accessory result refines rather than replaces the broader module picture. Across 1,116 ICA modules from 32 organisms, the independent Pan-bacterial Fitness Modules via Independent Component Analysis analysis found 94.2% with significantly elevated within-module cofitness, mean within-module |r|=0.34 versus 0.12 in the background, and 22.7x genomic-adjacency enrichment. 4 Its held-out benchmark found Module-ICA had <1% strict precision for exact KEGG KO assignments, whereas ortholog transfer achieved 95.8% precision, 91.2% coverage, and 0.934 F1. 4 Modules are therefore supported as process-level, multi-gene units, but module membership does not establish an individual gene’s molecular function. 4
Cross-organism alignment further supports the conservation aspect: 156 module families spanned 2+ organisms, including 28 spanning 5+, 7 spanning 10+, and 1 spanning 21 organisms; 145 families had consensus functional labels. 4 These are conservation patterns, not proof that every member has an identical molecular function. 4 Module membership is also not generally an accessory-genome phenomenon: module genes were 86.0% core versus 81.5% for all genes (+4.5 percentage points; OR=1.46, p=1.6e-87). 5 Among 974 modules with >=3 mapped genes, 577 (59%) were >90% core, 349 (36%) were 50–90% core, and 48 (5%) were <50% core; the median module was 93.4% core. 5
Independent digests support this core enrichment. The conservation-fitness synthesis reports that ICA decomposition identified 1,116 co-regulated fitness modules across 32 organisms, enriched in core genes (86% core vs 81.5% baseline, OR=1.46, p=1.6e-87), with 59% of modules >90% core. 6 The central discoveries digest reports the enrichment across 29 organisms as 86.0% core vs 81.5% baseline (Fisher OR=1.46, p=1.6e-87), with a per-organism paired Wilcoxon p=1.0e-03 and 22/29 organisms showing enrichment. The organism counts differ between module identification and the enrichment test, and both are reported as given. 7 The module-conservation report describes module genes as co-regulated functional units that skew toward the conserved core genome. 5
The Field vs Lab Gene Importance in *Desulfovibrio vulgaris* Hildenborough analysis is consistent with the predominantly core picture in a smaller, single-organism module set. It is limited to desulfovibrio-vulgaris-hildenborough, and its ‘field’ activity means laboratory fitness experiments classified as field-relevant conditions, not activity measured in natural environments. 8 Across 52 ICA fitness modules the mean core fraction was 0.886 and the median 1.000. No significant correlation was detected between field condition activity and module conservation (Spearman rho=0.071, p=0.62). 8 Its categories were 21 field-active conserved (‘ecological’) modules with mean core fraction 0.980, 17 conserved-quiet modules (low field activity) at 0.983, and 5 field-variable modules (field-active but less conserved) at 0.829. 8 This null result does not rule out a relationship between field-relevant activity and module conservation. It shows only that none was detected, and the single-organism scope and the small field-variable class limit the inference for lab-field-fitness-concordance. 8
The conservation-versus-fitness integration supports this predominantly core interpretation: 27,693 putative essential genes among 148,826 protein-coding genes across 33 organisms were 86.1% core, compared with 81.2% of non-essential genes, while essential-auxiliary genes numbered 3,683 and were 38.2% hypothetical. 2 Essential-core genes were 41.9% enzymes and 13.0% hypothetical, whereas essential-auxiliary genes were 13.4% enzymes and 38.2% hypothetical; this refines the module result by distinguishing a well-annotated conserved functional backbone from poorly characterized strain-specific essential functions. 2 The enrichment was modest overall (median odds ratio 1.56), and 18 of 33 organisms were significant after correction, so essentiality does not imply universal or perfectly conserved module membership. 2
The COG functional-category comparison supports and functionally refines this core-versus-novel distinction: across 32 species spanning 9 phyla and 357,623 genes, core genes were concentrated relative to novel or singleton genes in conserved translation, nucleotide, coenzyme, amino-acid, and energy functions, while novel or singleton genes were enriched in mobile elements (COG L, +10.88%), defense mechanisms (COG V, +2.83%), and unknown function (COG S, +1.64%). 9 The COG L enrichment was consistent across 100% of species, whereas COG V was also consistent across 100% and COG S across 69%. 9 This supports interpreting predominantly core fitness modules as part of a conserved functional architecture, while suggesting that accessory-module co-inheritance may be especially relevant to mobile, defense, and niche-associated functions; it does not establish that any individual module has one of those functions. 9
Thus, the stronger co-inheritance signal is concentrated most clearly in the accessory subset, while the conservation architecture of fitness modules is predominantly core. 1,5 The 48 accessory modules come from the module-conservation classification of 974 modules. Its 38 module families with <50% core genes are a separate family-level summary. That report proposes these families may be horizontally transferred functional units or niche-specific operons, which remain hypotheses rather than established identities. 5 The cofitness report asserts that the 48 accessory modules identified in module_conservation are functionally coherent units that travel together through the pangenome. Its own module test classified only 11 modules as accessory, so the assertion is a cross-project interpretation rather than a result its analyses showed for all 48. 1,5 The conservation-fitness synthesis likewise leaves unresolved which co-regulated functions the 48 accessory modules carry in the flexible genome. 6 The conservation-versus-fitness result further supports this distinction by showing that essential genes are preferentially core but that essential-auxiliary genes remain numerous and poorly characterized. 2 The report reads the module result as suggesting that coordinated regulation across multiple genes, not just pairwise functional similarity, is what most strongly constrains co-inheritance. This is an interpretation, not a separately measured result, and the evidence does not establish that regulation causes co-inheritance. 1
Prevalence-aware interpretation and pangenome integration
The module contrast supports testing coordinated units alongside pairwise relationships: modules can retain detectable variation when individual pairs are weak, but Korea demonstrates that modules also lose detectability when all members are >90% core and prevalence is near 1.0. 1 Ddia6719’s delta=+0.093 and pseudo3_N2E3’s delta=+0.026 despite ANI values of 99.47% and 99.66% show that even near-clonal organisms can retain enough accessory variation to detect co-inheritance, although their high baseline phi values limit interpretation. 1
The module-conservation analysis supports a related caution: family breadth was not associated with conservation (Spearman rho=-0.01, p=0.914), and the high all-gene core baseline of 81.5% limits absolute enrichment to the observed +4.5 percentage points. 5 The conservation-fitness synthesis reports the same null as rho=-0.01, p=0.91, and attributes it to a baseline already so high that there is no room for a gradient. 6 The discoveries digest reports Spearman rho=-0.01, p=0.914 and notes that the baseline core rate (~82%) leaves families nearly all core regardless of breadth. It calls this a ceiling effect rather than evidence against the conservation-function relationship, paralleling pangenome-core-boundary-and-clade-size-bias. 7 The pathway-capability analysis contrasts this null breadth result with its own pathway-level view, in which variability rather than breadth is the informative metric. 3 This suggests that module-family scope is less informative than gene-level conservation for explaining the genomic background against which co-inheritance is measured. 5 The conservation-versus-fitness integration supports the same caution: pangenome core fractions are sensitive to clade size, and clades containing only 2 genomes can have trivially high core fractions because presence in both genomes equals 100% core. 2
The study supports pangenome-integration by showing that linked functional-fitness and pangenome data can test whether laboratory relationships correspond to population-level gene co-occurrence, while prevalence is an explicit quality criterion for that cross-dataset analysis. 1 It also supports cross-tenant-data-bridging because it joined Fitness Browser data, KBase pangenome clusters, phylogenetic distances, ICA modules, and SEED annotations through linked files and Spark extraction. 1 The conservation-versus-fitness integration supports this cross-dataset premise by mapping 44 of 48 Fitness Browser organisms to pangenome species clades and producing 177,863 gene-to-cluster links, but refines it with coverage limits: 33 organisms were used downstream, four organisms were unmatched because their species had too few GTDB genomes, and ten organisms were excluded for <90% DIAMOND coverage. 2
The evidence does not establish that co-fitness causes co-inheritance: shared ancestry, prevalence structure, genome architecture, horizontal transfer, and other functional coupling can produce correlated presence patterns. 1 The narrower supported conclusion is that coordinated module membership was more strongly associated with pangenome co-occurrence than pairwise co-fitness, particularly for accessory modules, within a larger population of predominantly core co-regulated fitness modules. 1,5
The COG comparison refines the interpretation of horizontal transfer in this integration: the strongest cross-species novelty signal was COG L mobile-element enrichment, +10.88%, but the COG analysis was comparative functional annotation rather than a direct transfer test, and approximately 70% of genes had COG annotations. 9 Therefore, mobile-element enrichment is consistent with a mechanism that could help generate accessory co-inheritance, but it should not be treated as proof that the observed module associations were caused by horizontal transfer. 9
The cofitness report’s figures document these analyses. Figure 1 (figures/fig1_cofit_cooccurrence.png) plots phi versus prevalence for cofit and random pairs, per organism and aggregated. Figure 2 (figures/fig2_operon_control.png) shows adjacent versus distant cofit pairs and the signal after excluding operons. Figure 4 (figures/fig4_cofit_strength.png) shows co-fitness score against phi with binned means. Figure 5 (figures/fig5_module_coinheritance.png) shows module phi against null expectations by module type, with size effects. 1
Tensions
The cofitness report’s headline module statistics (195 ICA modules across 6 organisms, 51/195 significant at p<0.05) do not match its per-organism table. The table rows list Koxy: 44 modules, 14 (32%) significant, mean phi=0.079, null mean=0.040; Btheta: 36, 22 (61%), 0.192, 0.086; Putida: 38, 6 (16%), 0.266, 0.202; Korea: 29, 0 (0%), 0.357, 0.308; Phaeo: 37, 3 (8%), 0.108, 0.095; and pseudo3_N2E3: 40, 4 (10%), 0.444, 0.409. These rows do not add up to the headline module or significant-module counts, and the report leaves the discrepancy unresolved. 1 Read the headline delta=+0.053 and the 21/195 FDR-significant count with this unresolved inconsistency in mind; this page does not prefer either side. 1
The report also calls Korea the best pairwise organism when explaining its 0% significant modules. This conflicts with its own pairwise table, where Korea had a negative delta of -0.042 and Ddia6719 had the largest positive delta (+0.093). 1 The module-level explanation itself, that all Korea modules are >90% core with prevalence near 1.0, is not affected by this wording conflict. 1
Open Directions
- Restrict pairwise and module analyses to auxiliary-only pairs in which both clusters are below 95% prevalence; test whether the module advantage and organism-level heterogeneity persist after removing the prevalence ceiling. 1
- Recompute co-fitness from raw genefitness data for Ralstonia UW163 and Ralstonia GMI1000, then repeat prevalence-matched tests to determine whether their exclusion removed informative low-ANI diversity. 1
- Resolve reference-genome mapping and apply stronger phylogenetic controls to test whether module-level delta phi remains higher than pairwise delta phi after ancestry is accounted for. 1
- Combine prevalence-stratified phi estimates with reference-resolved phylogenetic distances to test whether co-fitness predicts co-occurrence after separately controlling for prevalence and shared ancestry. 1
- Build module co-transfer networks and test cross-module prediction after matching module prevalence and auxiliary content, asking whether modules predict one another’s pangenome distributions beyond within-module co-occurrence. 1
- Expand analyses to species with >30% auxiliary genes and existing co-fitness data, testing whether auxiliary-gene fraction and increased accessory variation improve pairwise and module-level detection. 1
- Reanalyze the 974 mapped modules using the same prevalence-matched test, separating the 577 (59%) >90% core, 349 (36%) mixed, and 48 (5%) <50% core modules, to determine whether accessory enrichment persists across the 32-organism conservation set. 5
- Test whether the null family-breadth association (rho=-0.01, p=0.914) remains after matching module families for core fraction and prevalence, distinguishing gene-level conservation from module-scope effects. 5
- Apply prevalence-matched co-inheritance tests to the 156 cross-organism module families and ask whether family-level conservation predicts co-occurrence beyond organism, ancestry, and core-fraction controls. 4
- Link accessory-module membership and COG L/V/S annotations to environmental metadata, testing whether mobile, defense, and unknown-function enrichments predict module co-inheritance after controlling for prevalence and phylogeny. 9
- Extend the conservation-versus-fitness analysis from binary essentiality to condition-specific fitness, especially fitness < -2 under stress conditions, and test whether conditionally important genes show different module and core-fraction patterns. 2
- Use Fitness Browser ortholog data to identify essential gene families conserved across species, correlate mean fitness effects with core-genome fraction, and characterize the 3,683 essential-auxiliary genes to test whether they compensate for missing core functions. 2
- Recount module-level results from the cofitness module output tables to reconcile the per-organism module counts with the headline 195 modules and 51 significant modules, before reusing the +0.053 module delta or the 73% accessory figure downstream. 1,3
- Split the 52 field-scored Desulfovibrio vulgaris Hildenborough modules by accessory status, and test whether the 5 field-variable modules show elevated prevalence-matched co-inheritance relative to conserved modules. Then repeat the field-classification approach in other organisms to test whether the single-organism null generalizes. 8