Conservation metrics are not interchangeable: the apparent priority of an unknown bacterial gene depends on how ortholog breadth is defined, how species and taxonomic coverage are counted, and whether the metric distinguishes broad conservation from annotation or database-coverage artifacts. 1 The Functional Dark Matter — Experimentally Prioritized Novel Genetic Systems provides a direct calibration case by comparing an eggNOG-based ortholog-group breadth ranking with a species-count variant and a larger GTDB r214 pangenome analysis. 1 The Conservation vs Fitness — Linking FB Genes to Pangenome Clusters adds a complementary calibration against Fitness Browser essentiality: across 33 organisms, 86.1% of essential genes were core versus 81.2% of non-essential genes, with a median odds ratio of 1.56. 2 This supports conservation as a measurable, but modest, discriminator of biological importance rather than a standalone proxy for function. 2 The Fitness Effects vs Conservation — Quantitative Analysis supports and refines this conclusion across approximately 194,000 genes from 43 diverse bacteria: essential genes were 82% core versus 66% for always-neutral genes, while fitness breadth had only a weak association with core status (Spearman rho=0.086, p=8.1e-230). 3 The synthesis in Gene Conservation, Fitness, and the Architecture of Bacterial Genomes supports this calibration across 194,216 protein-coding genes from 43 bacteria, but shows that conservation is organized partly into coordinated modules rather than acting only as an individual-gene score. 4 The new Core Gene Paradox — Why Are Core Genes More Burdensome? refines the calibration by showing that conservation–burden relationships are function-specific: core genes were more burdensome in several categories, whereas the pattern reversed for Cell Wall genes. 5
Why Metric Choice Matters
At an earlier analysis stage (NB02), the report placed phylogenetic breadth in the context of the 27,690-species pangenome and mapped 30,756 dark-gene clusters to taxonomy. 1 The eggNOG ortholog-group breadth classification assigned 30,721 of 30,756 clusters (99.9%) to “universal” breadth, meaning root-level eggNOG ortholog groups present across domains of life. 1 As a result, the NB05 s_pangenome score built on this classification was non-discriminative and produced identical conservation scores, making that category poorly discriminative for ranking unknown genes. 1 At this stage, species counts were the number of pangenome species sharing the same root ortholog group in the Fitness Browser-only analysis. 1 They ranged from 1 to 33 per ortholog group, with median 1 and mean 2.2. 1 This 1-to-33 range belongs to the earlier analysis and is not the later full-pangenome range. 1
Replacing the original binary breadth score with a continuous species-count scoring variant (min(n_species / 20, 1.0) × 0.5) produced Spearman ρ = 0.982 with the original ranking, but top-50 overlap was 62% and top-100 overlap was 58%. 1 This refines the interpretation of rank correlation: a high global correlation did not imply stable membership among the highest-priority candidates. 1
The report treats phylogenetic breadth as a prioritization signal. 1 Dark-gene clusters ranged from clade-restricted patterns (a single phylum) to widespread patterns (3+ phyla). 1 The report designated widespread families that still lack functional annotation as the highest-priority unknowns. 1 This prioritization rests on the premise that broad conservation signals fundamental function; it is not a direct functional measurement. 1 Separately, the report’s NB09 evidence-weighted approach relied on Fitness Browser data, which limited its conservation assessment to the 48 Fitness Browser organisms. 1
The conservation-versus-fitness analysis supports this distinction by showing that a core-versus-auxiliary metric can detect essentiality enrichment while retaining limited discriminative power: essential genes were only 1.56 times more likely to be core, and 18 of 33 organisms showed significant enrichment after Benjamini-Hochberg false discovery rate correction. 2 The result is therefore consistent with conservation helping prioritize candidates, but not uniquely determining their importance. 2 The larger fitness-effects analysis supports the same calibration: conservation increased across fitness-importance categories, but the gradient was only 16 percentage points between essential and always-neutral genes, so statistical robustness did not make fitness importance a strong standalone predictor of conservation. 3 The synthesis supports this modest-gradient interpretation: essential genes were 82% core, whereas always-neutral genes were 66% core, and it identifies 28,017 genes that were both costly in the laboratory and conserved as a joint selection signature rather than as direct evidence of natural-environment fitness. 4
The core-gene trade-off analysis refines the interpretation of conservation as a prioritization metric by identifying 25,271 true trade-off genes, representing 17.8% of the genes examined; these genes had fitness below -1 in some conditions and above 1 in others. 5 Trade-off genes were 1.29x more likely to be core (OR=1.29, p=1.2e-44), suggesting that conserved genes can encode functions whose costs and benefits vary across conditions rather than uniformly important functions. 5 This supports retaining condition-specific fitness alongside conservation and qualifies interpretations of core status as a direct measure of biological importance. 5
Expanded GTDB Calibration
The report states that two earlier inputs failed to meaningfully distinguish conservation patterns among dark genes: the coarse eggNOG breadth classification and the sparse 48-organism Fitness Browser sampling (37/48 Proteobacteria). 1 To address this, NB11b queried the full 27,690-species gtdb r214 pangenome. 1 A Spark query exploded eggNOG_OGs annotations across 93.5M gene clusters and matched all comma-separated ortholog-group entries against 11,774 dark-gene root ortholog groups. 1 Ortholog-group identifier propagation then recovered 5,206 additional dark genes (9.1%). 1 It did so by transferring root_og assignments from genes with known pangenome links to genes in the same 48-organism ortholog group that lacked them. 1 Together these steps expanded conservation coverage from 32,791 (57.5%) to 37,997 (66.6%) dark genes. 1 Species counts now ranged from 1 to 27,482 (median 135, mean 2,128), replacing the previous 1-to-33 range from the Fitness Browser-only analysis. 1 Phylum counts ranged from 1 to 142. 1 The expanded reference space therefore substantially altered the scale of observed conservation. 1 Mobile elements were detected via phylogenetic patchiness, meaning presence in distant phyla but in few species per phylum. 1
Among 11,774 root ortholog groups, 55.9% were kingdom-level, 11.0% class-level, 10.5% family-level, 6.9% genus-level, 6.5% mobile, 4.8% phylum-level, 3.9% order-level, and 0.5% species-level. 1 Each group’s tier was the narrowest rank containing all of its member species. 1 The report interprets the kingdom-level majority as evidence that functional dark matter is a fundamental limitation in understanding broadly conserved biology, not a minor annotation gap. 1 At the gene level, the report covers 57,011 dark genes, including 19,014 FB-only fallback assignments. 1 The gene-level distribution differed: kingdom 51.4%, species 29.8%, genus 6.2%, class 4.0%, family 3.5%, mobile 1.9%, phylum 1.9%, and order 1.3%. 1 Tier frequencies therefore depend on whether the denominator is ortholog groups or genes, particularly for the species tier. 1 These taxonomic tiers provide a more graduated representation of conservation than the initial near-universal eggNOG classification. 1
The expanded result supports using broader and more explicitly taxonomic reference data when the goal is to separate widely conserved unknown genes from narrowly distributed ones, but it does not establish that broad conservation alone predicts gene function. 1 Under a conservation-by-ignorance scheme, the report’s Finding 14 classified 6.0% of dark genes as strong testable hypotheses, 52.5% as weak leads, and 41.5% as true knowledge gaps. 1 The report’s methods section, however, assigns the eight taxonomic-tier percentages (kingdom 55.9%, class 11.0%, family 10.5%, genus 6.9%, mobile 6.5%, phylum 4.8%, order 3.9%, species 0.5%) to ortholog groups. It also phrases the same hypothesis-status percentages (strong 6.0%, weak 52.5%, gap 41.5%) as ortholog-group classifications. 1 The denominator of the hypothesis-status shares is therefore inconsistent within the source, and dark-gene and ortholog-group shares should not be conflated. 1
The report’s conservation-weighted Route B (NB11, NB11b) queries the 27,690-species GTDB r214 pangenome to measure the taxonomic breadth of each dark-gene ortholog group (species range 1–27,482). 1 It classifies each group by taxonomic tier and hypothesis status, then ranks groups by importance = conservation × ignorance. 1 The report recommends this route for discovery experiments that target broadly conserved unknowns. 1 The importance score is (tier-adjusted conservation + log₂-scaled species fraction) × an ignorance multiplier, which places kingdom-level true knowledge gaps at the top. 1 Because this ranking depends directly on the reference-breadth measurement, the reference-database sensitivity described above carries through to discovery priorities. 1
Supporting figures are fig30_species_count_scoring.png (species-count scoring distributions, rank comparison, rank changes and a top-20 table), fig33_conservation_tiers.png (absolute and normalized stacked bars of conservation tier × hypothesis status), fig34_classification_heatmap.png (a heatmap of taxonomic tier × hypothesis status with counts and percentages) and fig38_pangenome_species_distribution.png (old Fitness Browser-only versus new full-pangenome species-count distributions, plus phylum distribution). 1
The Fitness Browser integration refines this calibration: its pangenome categories placed 82.0% of linked genes in core clusters and 18.0% in auxiliary clusters, but clades containing only 2 genomes could have trivially high core fractions because presence in both genomes equals 100% core. 2 Thus, even a conservation metric that is associated with essentiality remains sensitive to clade size, coverage, and the reference construction procedure. 2 The newer cross-species analysis supports retaining continuous fitness magnitude and breadth alongside binary core status, because essential genes were 82.2% core, genes with min_fit < -3 were 77.7% core, and genes with min_fit from -1 to 0 were 66.4% core, despite the weak overall association. 3 The synthesis supports and extends this recommendation by showing that the conservation gradient is modest across 194,216 genes and that 5,526 genes were costly yet dispensable, a category requiring explicit treatment rather than automatic classification as conserved functional priorities. 4
Conservation and Experimental Prioritization
The conservation-weighted covering set selected 42 organisms covering 95.6% of importance-weighted priority across 28,584 high-priority dark genes. 1 A separate darkness-spectrum analysis selected 42 organisms from 28 genera to cover 95% of scored priority, with 32 organisms sufficient for 80% coverage. 1 These results support using conservation metrics not only to rank individual genes but also to design organism panels that maximize coverage of experimentally actionable unknowns. 1
Conservation was one of six axes in the prioritization score, alongside fitness importance, inference quality, pangenome distribution, biogeographic signal, and experimental tractability. 1 Overall rank correlations remained ρ > 0.93 across six alternative configurations, but only 64% of the original fitness-active top 50 remained under conservation-dominant or drop-tractability settings. 1 Essential-gene top-50 retention was 36% when tractability was dropped and 48% when neighbor context was dropped. 1 These sensitivity results qualify conservation-weighted ranking as a useful prioritization input rather than a uniquely correct ordering. 1
The essentiality comparison supports combining conservation with an independent functional axis: essential-core genes were 87% annotated with known function, whereas essential-auxiliary genes were 38.2% hypothetical and essential-unmapped genes were 44.7% hypothetical. 2 This suggests the highest-value unknowns may not be those with the broadest conservation alone, but those whose conservation, fitness importance, and annotation uncertainty jointly identify testable gaps. 2 The fitness-effects analysis refines this prioritization logic: core genes were more likely than auxiliary genes to show beneficial deletion effects (24.4% versus 19.9%), and strong condition-specific phenotypes were 77.3% core versus 70.3% for genes without such annotations. 3 Conservation therefore does not simply identify universally important genes; it can also capture genes with condition-dependent trade-offs that should be tested across environments. 3 The synthesis supports this interpretation: core genes were 1.78x more likely to have strong condition-specific phenotypes and 1.29x more likely to be both important and burdensome depending on condition, while the 28,017 costly-and-conserved genes are evidence for, rather than direct measurements of, purifying selection in nature. 4 The core-gene trade-off analysis supports and sharpens this condition-aware prioritization: its Costly + Conserved group contained 28,017 genes, while 5,526 were Costly + Dispensable, indicating that burden and conservation should be interpreted jointly rather than used as interchangeable priority labels. 5
Coordinated Conservation and Functional Resolution
The synthesis refines gene-level calibration by identifying 1,116 independent-component-analysis fitness modules across 32 organisms; 86% of module genes were core versus an 81.5% baseline (odds ratio 1.46, p=1.6e-87), and 59% of modules were more than 90% core. 4 This supports interpreting conservation as partly organized into coordinated functional units, while preserving the existing caution that broad conservation alone does not establish function. 1,4
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, whereas non-core cell-wall genes were more burdensome. 4 The core-gene trade-off analysis supports these category-dependent differences, reporting core-gene burden excesses of +7.8 percentage points for Motility and +12.9 percentage points for RNA Metabolism, a +6.2 percentage-point excess for Protein Metabolism, and the reverse pattern for Cell Wall genes, where the difference was -14.1 percentage points. 5 These results refine conservation-weighted prioritization by indicating that candidate interpretation should retain functional category and environmental condition, not only taxonomic breadth.
The core-gene analysis further interpreted 28,017 Costly + Conserved genes as genes maintained despite laboratory-measured cost, 5,526 Costly + Dispensable genes as candidates for ongoing gene loss, 86,761 Neutral + Conserved genes as classic housekeeping genes, and 21,886 Neutral + Dispensable genes as niche-specific genes. 5 This interpretation supports using a selection-signature matrix to separate conservation from laboratory burden, but remains an inference from laboratory fitness and conservation patterns rather than direct measurement of natural selection. 5
Strain-Resolution Check of Conservation Classes
An antimicrobial-resistance (AMR) strain analysis supports the validity of atlas-level conservation classes at strain resolution. 6 Within species, 77.3% of atlas-defined Core AMR genes were fixed, while 78.7% of Singletons were rare. 6 The report interprets this as validating the atlas classification and as evidence that the core/accessory distinction captures real biological variation in AMR gene inheritance. 6 This complements the clade-size caution above: that caution concerns how core fractions are computed, whereas this result tests whether core classes track within-species prevalence. 6,2
The same analysis also qualifies downstream statistics built on conservation classes. 6 Core AMR genes are nearly universal (>=95% prevalence by definition), so they have near-zero Jaccard distances with little variance. 6 This inherently suppresses distance-based correlation metrics such as the Mantel r. 6 The weaker core-gene signal is therefore partly a statistical artifact of the conservation definition, independent of biology, so core-versus-non-core contrasts in distance-based tests are partly methodological. 6
Ortholog Breadth Among Truly Dark Genes
A truly-dark-gene analysis refines the calibration by showing that conservation and ortholog-breadth metrics separate classes of unknown genes. 7 Compared with annotation-lag genes, truly dark genes had a median length of 121 versus 194 aa (d = −0.432, p < 1e-100) and a core genome fraction of 43.1% versus 72.7% (OR = 0.284, p < 1e-100). 7 Their essential fraction was 18.0% versus 13.4% (OR = 1.420, p < 1e-10), and their mean GC content was 0.542 versus 0.584 (d = −0.395, p = 2e-115). 7 They had orthologs in 29.3% versus 63.7% of cases (OR = 0.236, p < 1e-100), with a median ortholog breadth of 1 versus 4 organisms (d = −1.072, p < 1e-100). 7 Truly dark genes were thus less often core yet more often essential than annotation-lag genes. 7 This comes from a single analysis, and it suggests the hypothesis that breadth-weighted conservation schemes would systematically under-rank them. 7,1 Coverage limits that hypothesis: BBH orthologs cover only 32 of 48 Fitness Browser organisms, and truly dark genes from the remaining 16 organisms lack concordance and ortholog breadth data. Part of their apparent taxonomic restriction, and any apparent under-ranking, may therefore reflect missing ortholog data rather than biology. 7
Truly dark genes had lower odds of having cross-organism orthologs (OR = 0.236). 7 When they did have orthologs, those orthologs spanned fewer organisms (median 1 vs 4). 7 Only 3 of 65 dark-gene ortholog groups with cross-organism concordance data contained truly dark genes, making these genes nearly invisible to cross-organism analysis. 7 The source describes truly dark genes as “4.2× less likely” to have orthologs, but that phrase describes the odds ratio imprecisely. 7 It is not a 4.2-fold probability comparison; the underlying proportions are 29.3% versus 63.7%. 7
Tensions
The initial eggNOG metric suggested near-universal breadth for 99.9% of clusters, whereas the GTDB r214 analysis resolved root ortholog groups across kingdom, phylum, class, order, family, genus, species, and mobile categories. 1 This tension indicates that apparent conservation can depend strongly on reference-database composition, ortholog-group propagation, and the taxonomic level used for scoring. 1
The species-count variant was highly correlated with the original ranking at the global level, with Spearman ρ = 0.982, yet its top-50 and top-100 overlaps were only 62% and 58%, respectively. 1 Thus, rank correlation and decision stability answer different questions and should be reported together when conservation metrics guide experimental selection. 1
The Fitness Browser comparison introduces a further metric-resolution tension rather than a contradiction: essential genes showed only modest core enrichment (median odds ratio 1.56), while the expanded GTDB analysis produced conservation categories spanning kingdom to species and mobile levels. 2,1 The difference refines the interpretation of “conserved”: core membership within a species clade and broad taxonomic ortholog breadth are related but not interchangeable measurements. 2,1 The broader fitness-effects result adds a related qualification: core genes showed heavier fitness-effect tails in both directions, so core status may identify genes with stronger conditional costs and benefits rather than genes with uniformly greater importance. 3
The synthesis adds a further qualification rather than resolving the metric tension: module-family breadth did not predict conservation (rho=-0.01, p=0.91), and condition-specific fitness did not establish niche-specific fitness. 4 Thus, coordinated architecture and laboratory condition dependence can improve interpretation without making conservation a direct readout of natural ecological importance. 4
The core-gene trade-off analysis reinforces this qualification: its interpretation of costly conserved genes as maintained by natural selection is based on laboratory fitness measurements and conservation patterns, while laboratory conditions capture only a fraction of natural environments and Fitness Browser conditions are biased toward experimentally convenient settings. 5 It also cautions that defining burden as fitness above 1 may capture trade-offs rather than true dispensability, and that a 90% identity threshold for DIAMOND matching may miss rapidly evolving genes. 5
Relation to Other Concepts
This concept refines pangenome-integration by focusing on how pangenome reference breadth and ortholog-group propagation change conservation estimates rather than on pangenome integration generally. 1
It supports comparative-conservation-metric-calibration as a framework for testing whether conservation scores are discriminative, stable at decision thresholds, and appropriate for experimental design. 1
It also connects to gene-essentiality because conservation was combined with fitness importance in candidate prioritization, while the sensitivity analysis showed that changing conservation emphasis altered high-priority membership. 1 The cross-species essentiality analysis supports this connection by finding core enrichment among essential genes, while also showing that the association is modest and condition-dependent because essentiality was inferred from RB-TnSeq under represented growth conditions. 2 The synthesis supports and broadens this connection by showing that fitness breadth and conditional phenotypes add information beyond binary essentiality, while also warning that laboratory condition coverage and transposon coverage limit interpretation. 4 The core-gene trade-off analysis refines it further by showing that core status is enriched among condition-dependent trade-off genes and that burden reverses by functional category. 5
Open Directions
- Recalculate the same dark-gene rankings with matched reference sets and identical ortholog-group definitions, then test whether the 62% top-50 and 58% top-100 overlap values improve when database breadth is held constant. 1
- Compare kingdom-, phylum-, class-, family-, genus-, and species-level conservation scores against independent fitness measurements to test which taxonomic resolution best predicts experimentally measurable phenotypes. 1
- Use bootstrap resampling of species and phyla in the 93.5M gene-cluster reference to quantify confidence intervals for conservation ranks and identify candidates whose priority is database-sensitive. 1
- Re-run the 42-organism covering-set optimization under alternative conservation metrics and compare coverage of the 28,584 high-priority dark genes, asking whether the selected experimental panel is robust to metric choice. 1
- Test whether conservation-weighted candidates outperform fitness-only candidates in CRISPRi or RB-TnSeq follow-up, separating broad conservation from experimentally validated functional importance. 1
- Recompute essential-core enrichment across matched clade sizes and coverage thresholds, then compare binary essentiality and continuous fitness effects to determine whether the 1.56 median odds ratio is stable across reference designs. 2
- Test whether conservation-weighted candidates with strong condition-specific or beneficial deletion effects replicate across underrepresented ecological conditions, separating genuine conditional importance from laboratory and transposon-coverage bias. 3
- Characterize the 5,526 costly-and-dispensable genes and the 48 accessory co-regulated modules, then test whether mobile-element status, recent acquisition, or gene loss explains their departure from conservation-based expectations. 4
- Compare the 1,116 fitness modules with taxonomic conservation tiers and ecological metadata to test whether module-level conservation predicts reproducible phenotypes better than gene-level breadth. 4
- Reanalyze the 25,271 trade-off genes and the 28,017 Costly + Conserved genes across ecologically relevant conditions, testing whether category-specific burden differences replicate outside experimentally convenient Fitness Browser conditions. 5
- Test the 90% identity matching threshold and alternative orthology methods on rapidly evolving genes, asking whether apparent conservation–burden relationships change when homology-search negative evidence is strengthened. 5
- Recompute AMR within-species prevalence associations with variance-aware statistics instead of Jaccard-based Mantel r, testing whether the weaker core-gene signal persists once the variance suppression from the >=95% prevalence definition is removed. 6
- Rescore the conservation × ignorance ranking with truly dark and annotation-lag genes separated, restricted to the 32 Fitness Browser organisms with BBH ortholog coverage, testing whether truly dark genes, with their 18.0% essential fraction, are under-ranked by breadth-based conservation. 7,1