Functional Dark Matter — Experimentally Prioritized Novel Genetic Systems

Raw report: functional_dark_matter Feeds into: cofitness-network-architecture, comparative-conservation-metric-calibration, computational-pathway-prediction-validation, condition-specific-fitness, confirmatory-exploratory-ecological-association-discordance, core-gene-annotation-paradox, cross-species-fitness-transferability, cultivation-collection-bias-in-ecological-genomics, environment-embedding-geography, evidence-triangulation-for-functional-annotation, experimental-prioritization-of-functional-dark-matter, fitness-condition-coverage-prioritization-bias, fitness-matched-null-models, functional-marker-validation, gene-essentiality, genetic-perturbation-coverage-bias, genomic-dispersal-functional-coupling, genomic-under-representation, lab-field-fitness-concordance, metabolic-model-gapfilling, multi-omics-integration, ontology-and-category-schema-sensitivity, pangenome-core-boundary-and-clade-size-bias, pangenome-integration, pathway-versus-reaction-evidence-resolution, research-attention-inequality, sampling-depth-and-downsampling-effects, selection-on-outcome-leakage, structural-annotation-gap, taxonomic-resolution-dependent-functional-inference, transposon-callability-bias, two-speed-bacterial-genome

Overview

This report integrates Fitness Browser fitness profiles, essentiality, pangenome links, fitness modules, gene domains, environmental metadata, GapMind pathway analysis, cross-organism concordance, synteny, co-fitness, and Bakta reannotation to prioritize experimentally actionable bacterial genes of unknown function. Across 48 organisms, it identifies 57,011 dark genes among 228,709 genes, ranks candidates through evidence-weighted and conservation-weighted routes, and converts gene-level evidence into organism-level experimental campaigns. ₁

Key Findings

Scale and evidence coverage

Of 228,709 genes across 48 Fitness Browser organisms, 57,011 (24.9%) lack functional annotation as hypothetical proteins, DUFs, or uncharacterized genes. Of these, 7,787 show strong fitness effects (|fitness| ≥ 2 in at least one condition) and 9,557 are essential because no viable transposon mutants were recovered; together, 17,344 genes have experimentally measurable phenotypes. Dark genes range from more than 35% to less than 15% of an organism’s genes, a pattern interpreted as annotation-depth variation rather than established differences in functional content. The report notes that its 24.9% dark-gene fraction aligns with published estimates of 25–40% hypothetical genes in typical bacterial genomes (Makarova et al. 2019). ₁ Two figures plot these data: the per-organism annotation breakdown (fig01_annotation_breakdown.png) and the fitness distributions of dark versus annotated genes (fig02_fitness_distributions.png). ₁

Among the 57,011 dark genes, 39,532 (69.3%) have pangenome links, 12,686 are accessory, 511 are both accessory and strongly fitness-active, and 6,142 belong to independent-component-analysis (ICA) fitness modules, which group genes with coordinated fitness profiles. Stress conditions, including metals, oxidative stress, and osmotic stress, dominate strong dark-gene phenotypes, followed by carbon-source and nitrogen-source utilization. ₁ The report treats the 511 dark genes that are both accessory and strongly fitness-active as the prime candidates for biogeographic analysis. Dark-gene evidence coverage across pangenome, module, ortholog and essentiality dimensions is plotted in fig03_dark_gene_coverage.png, and the condition-class distribution for dark genes with strong phenotypes in fig04_condition_classes.png. The census behind these counts integrates four prior observatory projects (fitness_modules, essential_genome, conservation_vs_fitness, module_conservation) with direct Fitness Browser queries into a single table of every dark gene across all 48 organisms, because those projects answered different questions in different formats and no unified catalog existed. The census identifies 57,011 dark genes (24.9% of 228,709 total), of which 39,532 (69.3%) link to the pangenome, 6,142 (10.8%) belong to ICA fitness modules with function predictions, 7,787 (13.7%) show strong fitness effects (|fit| ≥ 2) and 9,557 (16.8%) are essential; the intersection of accessory dark genes with strong fitness is 511 genes. The report concludes that one quarter of bacterial genes remain dark but that the problem is tractable: 17,344 have measurable phenotypes (fitness or essentiality) and 39,532 connect to the pangenome. ₁

Metabolic gaps and pathway hypotheses

GapMind identified 1,256 organism–pathway pairs across 44 Fitness Browser-linked species in which nearly complete pathways had steps_missing_low scores and co-occurred with strongly fitness-active dark genes. The most frequent gaps were carbon-utilization pathways and one amino-acid pathway: fucose utilization in 32 organisms (examples: Marinobacter, P. stutzeri RCH2, D. vulgaris), rhamnose utilization in 31 (Marinobacter, P. putida, Phaeo), sorbitol utilization in 30 (D. desulfuricans, D. vulgaris, Miyama), myoinositol utilization in 28 (P. putida, P. syringae, WCS417), gluconate utilization in 26 (Marinobacter, D. desulfuricans), and asparagine biosynthesis in 24 organisms across diverse phyla. Marinobacter had 49 gapped pathways, while Desulfovibrio desulfuricans ME-23 and Pseudomonas stutzeri had 45 each. The report notes that these organisms also have the largest dark-gene complements and the most specific fitness phenotypes. From this it suggests the hypothesis, not an established assignment, that functionally dark genes may encode their missing steps. GapMind gap-filling candidates are plotted in fig05_gapmind_gaps.png. The underlying query covered 305M GapMind pathway rows, filtered to the 44 FB-linked species. The report frames a dark gene near a gap as a candidate for the missing enzyme, not a confirmed one. ₁

These GapMind results are organism-level co-occurrences, not direct gene-to-enzyme assignments. Across the 1,256 organism–pathway pairs, supplementary domain matching identified 42,239 gene–pathway candidates across 3,186 unique dark genes, including 5,398 high-confidence EC-prefix matches, 4,687 medium-confidence Pfam-family matches, and 32,154 low-confidence keyword matches. The matching used curated pathway–enzyme mappings: EC prefixes, Pfam families and functional keywords. The resulting domain-compatible candidates narrow the search from all dark genes in a gapped organism to those with enzymatically plausible annotations. Compatibility does not validate enzymatic function. The analysis is plotted in fig28_domain_matching.png, which shows matches per pathway and organism, the confidence tiers, and fitness stratified by confidence tier. Confirming individual assignments requires EC matching, structure prediction, enzymology, or other experimental validation. ₁

Conserved phenotypes and phylogenetic breadth

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. The top concordant groups span carbon utilization, stress response and motility, and every listed top group had concordance of 1.00. ₁

Carbon-source groups:

• OG11386: carbon source; concordance 1.00 across 8 organisms; DUF5064; in P. putida, P. syringae and P. stutzeri RCH2.
• OG15006: carbon source; concordance 1.00 across 7 organisms; in Ralstonia spp.
• OG14628: carbon source; concordance 1.00 across 5 organisms; in Ralstonia spp.
• OG12530: carbon source; concordance 1.00 across 4 organisms; DUF2844; in B. thailandensis and Burkholderia sp. 376. ₁

Stress groups:

• OG05812: stress; concordance 1.00 across 8 organisms; Peptidase_M50; in MR-1 and SB2B; module-predicted TIGR01730.
• OG05815: stress; concordance 1.00 across 8 organisms; ParE_toxin; in S. meliloti.
• OG03384: stress; concordance 1.00 across 6 organisms; Metallophos; module-predicted glutathione S-transferase. ₁

Motility groups:

• OG10428: motility; concordance 1.00 across 3 organisms; ThiS; in S. meliloti; max |fit| = 2.2.
• OG10455: motility; concordance 1.00 across 3 organisms; MS_channel mechanosensitive-channel domains; in S. meliloti. ₁

The report gives three interpretations. It reads the carbon-source concordance as suggesting conserved but unannotated components of carbohydrate metabolism. It reads the motility concordance as suggesting unannotated parts of the chemotaxis machinery. It reads OG05812’s Peptidase_M50 domain, from the site-2 protease family, as hinting at a conserved regulatory protease under stress. All three remain functional hypotheses. The cross-organism fitness-concordance heatmap is plotted in fig06_concordance.png. The concordance finding places the strongest concordance in carbon source genes (spanning 3–8 organisms) and motility genes. The report’s hypothesis assessment, which tests the 65 ortholog groups present in 3+ Fitness Browser organisms and calls H1c supported, names only motility-related dark genes as showing the strongest cross-organism concordance, consistent with conserved but incompletely annotated chemotaxis machinery. The two passages differ in emphasis, and the chemotaxis reading is an interpretation. ₁

Phylogenetic analysis mapped 30,756 dark-gene clusters across 27,690 species. The initial eggNOG ortholog-group breadth classification assigned 30,721 of 30,756 clusters (99.9%) to universal breadth, making it poorly discriminative. As used in NB05’s s_pangenome score, it produced identical conservation scores. At this earlier Fitness Browser-only stage, species counts ranged from 1 to 33, with median 1 and mean 2.2. The report stated at that stage that species count (1–33) provides finer resolution than the coarse universal/clade-restricted classification. This 1–33 range is not the later full-pangenome range. A continuous species-count scoring variant (min(n_species / 20, 1.0) × 0.5) had Spearman ρ = 0.982 with the original ranking. Its top-50 overlap was 62% and its top-100 overlap was 58%, so the candidate lists change meaningfully; the variant is provided as a supplementary analysis. Breadth patterns range from clade-restricted (single phylum) to widespread (3+ phyla), and the report treats widespread unannotated families as the highest-priority unknowns. The breadth distribution and the species-count variant are plotted in fig07_phylo_breadth.png and fig30_species_count_scoring.png; the latter shows species-count score distributions, a rank comparison, rank changes and a top-20 table. ₁

The report states that the coarse eggNOG breadth classification (99.9% “universal”) and the sparse 48-organism Fitness Browser sampling (37/48 Proteobacteria) fail to meaningfully distinguish conservation patterns among dark genes. The report adds that the NB09 evidence-weighted approach relies on Fitness Browser data, which limits its conservation assessment to the 48 FB organisms. To address this, NB11 queried the full 27,690-species GTDB r214 pangenome. It exploded eggNOG annotations across 93.5M gene clusters and matched them against 11,774 dark-gene root ortholog groups. Ortholog-group identifier propagation then recovered an additional 5,206 dark genes (9.1%), by transferring assignments from genes with pangenome links to genes in the same 48-organism ortholog group that lacked them. Together these steps raised conservation coverage from 32,791 (57.5%) to 37,997 (66.6%) dark genes. Species counts ranged from 1 to 27,482, with median 135 and mean 2,128, replacing the previous 1-to-33 range from the FB-only analysis; phylum counts ranged from 1 to 142. 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. Each tier reflects the narrowest rank containing all member species. Mobile elements are detected through phylogenetic patchiness, meaning presence in distant phyla but few species per phylum. At the gene level, across 57,011 dark genes including 19,014 FB-only fallback assignments, the tiers were 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%. The report interprets the over-half share of pan-bacterial dark-gene ortholog groups as showing that functional dark matter is a fundamental limitation in understanding broadly conserved biology, not a minor annotation gap. The full-pangenome species-count distribution, compared with the old Fitness Browser-only distribution, and the phyla distribution are plotted in fig38_pangenome_species_distribution.png. ₁

The conservation-by-ignorance classification sorted dark genes into three tiers. 6.0% were strong testable hypotheses: a module prediction with EC number, high cross-organism concordance, a high-confidence GapMind match, or a named domain plus strong fitness. 52.5% were weak leads: a DUF-only domain, a bare module prediction, a medium/low GapMind match, strong fitness without annotation, or a named domain without fitness. 41.5% were true knowledge gaps, with zero functional evidence of any kind. A methods summary instead phrases these 6.0%, 52.5% and 41.5% shares as classifications of dark-gene ortholog groups, placing them alongside the ortholog-group taxonomic-tier percentages. The two denominators, dark genes and ortholog groups, should not be conflated. The highest-ranked knowledge gaps included COG0468 with 27,427 species across 142 phyla, COG0443 with 27,279 species, and COG0491 with 27,393 species. The importance score conservation × ignorance ranks all 11,774 dark-gene ortholog groups; the report defines it as “tier-adjusted conservation score + log₂-scaled species fraction × ignorance multiplier” in one passage. A methods passage gives the grouping explicitly as (tier-adjusted conservation + log₂-scaled species fraction) × ignorance multiplier. Kingdom-level true knowledge gaps score highest and species-specific strong hypotheses score lowest. The report’s description of this score names three top-ranked groups: COG0468 (27,427 species, 142 phyla, true knowledge gap), COG0443 (27,279 species, true knowledge gap) and COG0172 (27,431 species, weak lead). A conservation-weighted covering set selected 42 organisms covering 95.6% of importance-weighted priority across 28,584 high-priority dark genes. It selected organisms by sum(importance) × tractability × phylo_bonus, where phylo_bonus penalizes genus-redundant selections. Sinorhizobium meliloti ranks first (1,630 genes, 195 kingdom-level gaps), followed by P. putida (1,043 genes, 172 kingdom gaps) and S. oneidensis MR-1 (805 genes, 105 kingdom gaps). The report lists Bacteroides thetaiotaomicron with 1,382 genes and 267 kingdom gaps, labels this the “highest raw count”, and says it was deprioritized by 0.3 tractability. That label is inconsistent with its gene count, which is lower than the 1,630 genes listed for S. meliloti, though it would fit the kingdom-gap count; the report does not say which count it means. The first 10 organisms cover 31% of priority. Per-organism plans recommend three experiment types: CRISPRi (CRISPR interference) knockdown for essential genes in tractable organisms, condition-specific TnSeq for fitness-active genes, and broad phenotypic screens for true knowledge gaps. Top conditions come from fitness data or keyword inference. Six figures plot these outputs: the Fitness Browser organism phylum distribution, showing sparse taxonomic coverage (fig32_organism_taxonomy.png); conservation tier × hypothesis status stacked bars, absolute and normalized (fig33_conservation_tiers.png); a 2D heatmap of taxonomic tier × hypothesis status with counts and percentages (fig34_classification_heatmap.png); the top 25 true-knowledge-gap ortholog groups by pangenome species count (fig35_top_knowledge_gaps.png); conservation-weighted covering-set optimization, with cumulative coverage and marginal genes (fig36_covering_set_curve.png); and per-organism dark-gene coverage by conservation tier for the top 20 organisms (fig37_experiment_plan_heatmap.png). ₁

Environmental and lab–field concordance

Within-species carrier-versus-non-carrier analyses tested 151 accessory dark-gene clusters across 31 species; 10 of 137 clusters with environmental-category tests showed significant enrichment at false discovery rate (FDR) < 0.05, and 1 of 67 showed significant AlphaEarth embedding divergence. A pre-registered lab–field mapping found 29 of 47 testable clusters concordant (61.7%); pH-associated genes had 100% concordance across 4 tests and nitrogen-source genes had 78% concordance across 9 tests. ₁ Six clusters reached FDR < 0.2, including:

• K. oxytoca BWI76_RS15525 (carbon |fit| 2.3): 14.3% of carriers versus 2.5% of non-carriers were in the expected soil, freshwater or plant environments (odds ratio (OR) 6.6, FDR 0.069).
• K. oxytoca BWI76_RS15535 (carbon |fit| 2.2): the same 14.3% versus 2.5% split (OR 6.6, FDR 0.069).
• P. putida N2C3 AO356_12450 (carbon |fit| 2.1): 62.5% versus 0% (OR inf, FDR 0.093).
• P. putida N2C3 AO356_11255 (nitrogen |fit| 3.4): 80.0% of carriers versus 8.3% of non-carriers were in the expected soil, freshwater or wastewater environments (OR 44.0, FDR 0.093).
• P. putida N2C3 AO356_25185 (anaerobic |fit| 2.7): 55.6% versus 0% in the expected soil, freshwater or animal environments (OR inf, FDR 0.178).
• P. putida N2C3 AO356_24150 (nitrogen |fit| 3.0): 55.6% versus 0% in the expected soil, freshwater or wastewater environments (OR inf, FDR 0.178). Three figures plot this analysis: carrier-species environmental distributions (fig08_env_distribution.png), carrier-versus-non-carrier test results (fig09_carrier_tests.png) and an AlphaEarth embedding UMAP (fig10_embedding_umap.png). The lab–field concordance matrix is plotted in fig11_concordance_matrix.png. The report’s summary of this analysis adds that NMDC independent validation confirmed all 4 pre-registered predictions and that 76/105 abiotic correlations reached significance. It explicitly cautions that compositional coupling inflates the NMDC signal. It nonetheless argues that directional concordance across independent datasets supports the inference that lab fitness phenotypes reflect real ecological function. ₁

The 10 significant environmental-category clusters were:

• Pseudomonas putida: PP_0025 (stress |fit| 4.8; human_clinical; odds ratio 27.5; FDR 7e-6), PP_3434 (nitrogen |fit| 3.1; human_clinical; odds ratio 28.6; FDR 7e-6), PP_0642 (nitrogen |fit| 2.8; human_clinical; odds ratio 11.6; FDR 0.001) and PP_3105 (stress |fit| 3.7; human_assoc; odds ratio inf; FDR 0.004).
• P. syringae B728a: Psyr_0167 (in planta |fit| 4.6; plant_assoc; odds ratio 11.9; FDR 0.005) and Psyr_2830 (stress |fit| 3.4; plant_assoc; odds ratio 10.9; FDR 0.031).
• P. putida N2C3: AO356_11255 (nitrogen |fit| 3.4; freshwater; odds ratio inf; FDR 0.031; module-predicted D-Ala-D-Ala carboxypeptidase).
• Lower rather than higher carrier association with the category: B. thetaiotaomicron 354052 (stress |fit| 2.4; human_assoc; odds ratio 0.17; FDR 0.005), B. thetaiotaomicron 350920 (stress |fit| 2.2; human_assoc; odds ratio 0.37; FDR 0.031) and K. oxytoca BWI76_RS15640 (carbon |fit| 2.1; human_assoc; odds ratio 0.14; FDR 0.031; module-predicted phage-tail tape-measure protein). ₁

The report draws two patterns from these hits, both interpretations rather than established functions. P. putida stress and nitrogen dark genes are enriched in clinical or human-associated isolates, which suggests host-associated niche adaptation. P. syringae in-planta and stress dark genes are enriched in plant-associated genomes, consistent with their lab phenotypes. Elsewhere, in the lab–field mapping, the report singles out AO356_11255, the project’s top candidate, as the clearest signal: 80% of carrier genomes versus only 8.3% of non-carriers come from soil/freshwater/wastewater environments (OR = 44, FDR = 0.093), matching its lab phenotype of strong nitrogen-utilization fitness. It also describes AO356_11255’s carriers in two ways: as found exclusively in freshwater/soil environments, and as enriched in soil, freshwater, and wastewater environments relative to non-carriers. Under either description the carriers match the gene’s nitrogen-utilization fitness phenotype. ₁

The one-sided binomial test for 29/47 concordant clusters gave p = 0.072, with Wilson score 95% CI [0.474, 0.742], which includes 0.50, so the result is marginal, while Fisher’s combined probability across 47 individual tests gave p = 0.031; these statistical tests are summarized in fig31_statistical_tests.png, which presents a binomial null, a QQ plot, sign tests and a summary table. NMDC, the National Microbiome Data Collaborative dataset, independently mapped 5 of 6 carrier genera to 47 taxon columns across 6,365 metagenomic samples and confirmed all 4 testable pre-registered abiotic predictions: nitrogen-source carriers correlated with total nitrogen (ρ = +0.109, n = 1,231, FDR = 2.3e-4) and ammonium nitrogen (ρ = +0.231, n = 1,230, FDR = 8.0e-16), pH carriers correlated with pH (ρ = +0.157, n = 4,366, FDR = 7.4e-25), and anaerobic carriers correlated negatively with dissolved oxygen (ρ = -0.298, n = 272, FDR = 1.5e-6). The NMDC metagenomic samples are entirely separate from the pangenome-based carrier analysis, so the report treats these nitrogen, pH and oxygen associations as independent confirmations. ₁

Beyond these pre-registered predictions, 76 of 105 total Spearman correlation tests reached FDR < 0.05 (72.4%). The report attributes this high rate largely to confounding: the carrier genera (Pseudomonas, Klebsiella, Bacteroides) are among the most abundant and ubiquitous taxa in NMDC samples, so their abundance correlates broadly with many abiotic variables regardless of condition class. It therefore treats the 4/4 pre-registered prediction rate as the more meaningful metric, because it tests specific directional hypotheses. NMDC abiotic correlation results are plotted as a volcano plot in fig12_nmdc_correlations.png. ₁

The condition–environment mapping departed from the plan. The research plan specified 7 condition–environment mappings, including osmotic, temperature and oxidative stress, but the implementation used 6: stress (consolidating metal, osmotic and oxidative stress), carbon source, nitrogen source, pH, motility and anaerobic. The consolidation was needed because the Fitness Browser expGroup field uses a broad “stress” category rather than sub-categorizing stress types; motility and anaerobic were added because they emerged as prominent condition classes among dark genes with strong fitness effects. ₁

NMDC trait-feature analysis, using the same genus bridge against 76 functional-group columns across 6,365 samples, confirmed all 7 pre-registered trait-condition predictions at FDR < 10⁻²¹, including nitrogen-source carriers with nitrogen fixation (ρ = 0.60) and nitrate denitrification (ρ = 0.52), and carbon-source carriers with aerobic chemoheterotrophy (ρ = 0.73) and fermentation (ρ = 0.59), and anaerobic carriers with fermentation (ρ = 0.59) and iron respiration (ρ = 0.45). However, 441 of 449 exploratory tests also reached FDR < 0.05, and the report attributes this high rate largely to compositional coupling because abundant genera contribute to both carrier abundance and community trait scores, rather than to independent gene–phenotype validation. Separately, a binomial sign test on the 7/7 correct-direction pre-registered trait predictions gave p = 0.0078, which the report takes as showing that the directional hypotheses are significantly non-random. An NMDC trait–condition correlation heatmap covering the 7 pre-registered and the exploratory tests is plotted in fig21_trait_correlations.png. ₁

Candidate prioritization and experimental design

A six-axis composite score combined fitness importance, cross-organism conservation, inference quality, pangenome distribution, biogeographic signal, and experimental tractability, with each dimension scaled 0–1. It ranked all 17,344 phenotype-bearing dark genes, translating diverse evidence types into a single prioritization. Its top-ranked genes are those where the most evidence converges and the experimental path is clearest. The report describes the six scored evidence axes as producing 100 ranked candidates with specific functional hypotheses and suggested experiments. These 17,344 are the strong-fitness and essential dark genes defined above taken together, not two separate populations. The top 100 candidates span 22 organisms, with Shewanella MR-1 contributing 25, P. putida N2C3 18, and Marinobacter 9; 82/100 have high-confidence functional hypotheses supported by at least 3 evidence types, and 85/100 have module-based predictions. Scores range from 0.624 to 0.715, and 97/100 have domain annotations. The report attributes MR-1’s 25/100 top candidates to its deep condition coverage (121 conditions) rather than to inherent biology. It also describes MR-1’s dominance of the top candidates (25/100) as consistent with MR-1’s position as a model organism with extensive condition coverage and a large hypothetical-gene complement. ₁

The highest-ranked fitness-active candidate was P. putida N2C3 AO356_11255, with |fit| = 3.4 under nitrogen conditions, a D-alanyl-D-alanine carboxypeptidase prediction, an EamA domain, and a lab–field odds ratio of 44. The next candidates included MR-1 202463 with |fit| = 6.4 under stress and a YGGT domain, and the MR-1 K03306 paralog trio 199738, 203545, and 202450, with nitrogen-associated fitness effects of 5.5, 4.0, and 3.9 respectively. ₁

The report’s top-20 table follows; module assignments are predictions, not validated functions. The report reads three patterns in it. First, MR-1 genes 199738, 203545 and 202450 all carry K03306 module predictions with different domain architectures, suggesting paralogous members of a conserved nitrogen-responsive system. Second, MR-1 genes 202463 and 203247 both predict PF01145 under stress, pointing to a stress-responsive membrane protein family (YGGT/GBBH-like). Third, AO356_11255 is the only accessory gene in the top 10 and has the strongest biogeographic signal (lab-field OR = 44, NMDC nitrogen correlation). The two functional readings are hypotheses. Scoring-dimension distributions for all dark genes are plotted in fig13_score_components.png, a stacked-bar score breakdown of the top 20 candidates in fig14_top20_dossiers.png, and the top 100 candidates by organism and condition class in fig15_organism_distribution.png. Entries give rank, organism, locus, |fit|, condition, module prediction, domains, core or accessory status, and score: ₁

• 1: P. putida N2C3 AO356_11255; 3.4; nitrogen; D-Ala-D-Ala carboxypeptidase; EamA; accessory; 0.715.
• 2: Shewanella MR-1 202463; 6.4; stress; PF01145; YGGT; core; 0.698.
• 3: Shewanella MR-1 199738; 5.5; nitrogen; K03306; Gcw_chp, TIGR02001; core; 0.698.
• 4: Shewanella MR-1 203545; 4.0; nitrogen; K03306; DUF4124; core; 0.694.
• 5: Shewanella MR-1 202450; 3.9; nitrogen; K03306; Gly_transporter; core; 0.693.
• 6: P. putida N2C3 AO356_18320; 3.8; motility; PF00460; MotY_N, OmpA; core; 0.689.
• 7: P. fluorescens N1B4 Pf1N1B4_3696; 3.7; pH; PF00361; DUF3108; core; 0.687.
• 8: Shewanella MR-1 201124; 5.0; nitrogen; PF01144; HgmA_N, HgmA_C; core; 0.685.
• 9: P. putida N2C3 AO356_15270; 5.6; carbon; PF02589; LrgB; core; 0.685.
• 10: Shewanella MR-1 203247; 4.6; stress; PF01145; GBBH-like_N; core; 0.685.
• 11: Shewanella MR-1 203720; 4.5; nitrogen; no module prediction; Ser_hydrolase; core; 0.680.
• 12: Marinobacter GFF2506; 3.4; stress; ArnT/PqaB; AsmA_2, DUF3971; core; 0.678.
• 13: Shewanella MR-1 203026; 3.5; carbon; no module prediction; AFG1_ATPase; core; 0.677.
• 14: Marinobacter GFF1827; 3.7; stress; PF01145; Bax1-I; core; 0.676.
• 15: Shewanella MR-1 201809; 4.4; stress; no module prediction; PG_binding_3, Glyco_hydro_108; core; 0.674.
• 16: Shewanella MR-1 201731; 4.0; motility; TIGR00254; ZapC; core; 0.674.
• 17: Marinobacter GFF1367; 5.7; stress; PF00270; IMS; core; 0.672.
• 18: Shewanella MR-1 202474; 7.1; carbon; PF00460; YggL_50S_bp; core; 0.672.
• 19: P. putida N2C3 AO356_17245; 3.6; stress; K00763; Biotin_lipoyl, HlyD; core; 0.671.
• 20: P. putida N2C3 AO356_08210; 4.0; stress; K03808; DUF3426, zinc_ribbon; core; 0.670.

The report asked which organism × condition experiments would give the highest return for characterizing dark genes, given that RB-TnSeq (randomly barcoded transposon sequencing) libraries already exist for the Fitness Browser organisms. It ranked all organism–condition pairs by the number of top-500 dark-gene candidates they would address and then applied greedy set-cover optimization. This analysis found that 10 organism–condition experiments would address 242 of the top 500 dark genes, a share the report gives as 45.3%. The first three MR-1 experiments (stress, nitrogen source and carbon source) would address 111 top-500 candidates, which the report gives as 20.8%. These reported percentages do not match the stated top-500 denominator, and the report does not explain the difference, so the counts and percentages are reproduced here as reported. The table’s intermediate cumulative percentages, such as 10.7% for the first 57 genes, likewise do not match a denominator of 500. The report lists the full 10-experiment sequence as MR-1 stress, MR-1 nitrogen source, MR-1 carbon source, P. putida N2C3 stress, Sinorhizobium meliloti carbon source, S. meliloti stress, P. stutzeri RCH2 stress, Marinobacter stress, P. fluorescens GW456-L13 carbon source, and P. putida carbon source. The report calls MR-1 the single highest-value organism for dark-gene characterization, attributing this to deep condition coverage (121 conditions historically), a large dark-gene complement (587 scored) and high fitness-effect magnitudes (142 genes with |fit| ≥ 4). All coverage figures are projections for experiments not yet run. ₁

The proposed MR-1 screens are led by specific loci; the report’s domain-level readings are hypotheses: ₁

• Stress (57 genes): 202463 (YGGT domain, |fit|=6.4), 203247 (GBBH-like_N, |fit|=4.6) and 203631 (|fit|=4.5), all carrying PF01145 module predictions that suggest, but do not establish, a stress-responsive membrane-protein family.
• Nitrogen (31 genes): the K03306 paralog trio 199738 (Gcw_chp/TIGR02001, |fit|=5.5), 203545 (DUF4124, |fit|=4.0) and 202450 (Gly_transporter, |fit|=3.9); comparing single versus double mutants would test functional redundancy.
• Carbon (23 genes): 202474 (YggL_50S_bp, |fit|=7.1) and 202608 (BcrAD_BadFG, |fit|=7.3), the highest fitness magnitudes in the entire top 100.

The second-highest-value organism is P. putida N2C3, with 48 genes in the top 500, where stress conditions alone would resolve 23 candidates. Its #1-ranked gene AO356_11255 (D-alanyl-D-alanine carboxypeptidase prediction, EamA domain, |fit|=3.4 under nitrogen) also has the strongest biogeographic signal (lab–field OR = 44 and an NMDC nitrogen correlation). ₁

Cross-organism ortholog coverage amplifies the projected return: 101 ortholog groups in the top 500 span 2+ organisms, including: ₁

• OG03827: carbon source; 6 organisms (MR-1, Marinobacter, ANA3, PV4, P. fluorescens N1B4, P. simiae N2E2).
• OG02907: nitrogen source; 6 organisms (MR-1, E. coli Keio, K. oxytoca, P. putida, P. fluorescens N1B4, P. simiae N2E3).
• OG01383: pH; 5 organisms (MR-1, S. meliloti, C. metallidurans, P. fluorescens GW456-L13, P. putida N2C3).
• OG03534: stress; 5 organisms (MR-1, ANA3, Dinoroseobacter, PV4, Synechococcus).
• OG01997: stress; 4 organisms (MR-1, ANA3, S. meliloti, P. putida N2C3).

The report proposes that stress experiments in MR-1 and P. putida N2C3 alone would provide cross-organism concordance data for OG01997 and, combined with nitrogen experiments, would test OG02907 and OG01383 across multiple genetic backgrounds, calling this the most efficient path to identifying conserved dark-gene functions. This is an experimental projection, not a completed test, and the report does not explain how stress or nitrogen screens would test OG01383, which its own table assigns to pH. ₁

Essential dark genes require a separate strategy because standard RB-TnSeq, randomly barcoded transposon sequencing, yields no viable knockout fitness profiles for them. The report states that, of the 17,344 scored dark genes, 9,557 (55%) are essential, so essential dark genes represent 55% of the experimentally actionable dark matter but are systematically missed by fitness-centric scoring, since they have zero rows in genefitness; the report notes that no essential genes appeared in the fitness-centric (NB05) top 100. A separate prioritization therefore ranked all 9,557 essential dark genes on 5 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). Among 57,011 dark genes, 30,190 (52.9%) shared a predicted operon with an annotated gene under a same-strand, gap < 300 bp heuristic, which enables guilt-by-association functional hypotheses. A further 97.2% had at least one annotated neighbor within a five-gene window. The report cautions that this rate is expected for any gene given the 75% genome-wide annotation rate (P(≥1 annotated in 10 neighbors) > 99% by chance). More informatively, the mean annotated-neighbor fraction (63.6%) is below the 75% genome-wide baseline. The report reads this as dark genes co-localizing with other dark genes more than expected by chance, consistent with unannotated operonic clusters. It treats proximity-based hypotheses as leads for experimental testing, not validated assignments, noting that published operon predictors such as DOOR and ProOpDB incorporate signals beyond strand and gap distance. The report describes these NB07 operon predictions as a single-genome positional heuristic. ₁

The highest-ranked essential candidates were Escherichia coli Keio 14796 with score 0.875 and a YbeY domain, MR-1 200382 with score 0.874 and RimP_N/DUF150_C domains, and Klebsiella oxytoca BWI76_RS08540 with score 0.865 and OmpA/TIGR02802 domains. The recommended approach is CRISPRi, CRISPR interference for transcriptional knockdown, including Mobile-CRISPRi for less-established organisms, followed by growth measurements under standard and stress conditions. The report’s essential-candidate table annotates the top three as follows; each function is predicted, not established:

• E. coli K-12 14796 (0.875, YbeY): conserved in 5+ phyla, synteny-confirmed operon, predicted ion transport.
• MR-1 200382 (0.874): predicted ribosome assembly factor with a conserved neighborhood.
• K. oxytoca BWI76_RS08540 (0.865): predicted cell division, conserved across Enterobacteriaceae. The report says these candidates score highest because several evidence lines converge: detectable domains, membership in widely conserved ortholog groups, conserved gene neighborhoods with annotated partners, and organisms amenable to CRISPRi. These are functional clues, not experimental confirmation. The top 50 essential candidates score 0.740–0.875, above the fitness-active top 100 (0.624–0.715), which the report attributes to stronger conservation and neighborhood signals. The two sets come from different scoring approaches. ₁

The top 50 essential candidates (scores 0.740–0.875) span 15 organisms. These 50 ranked candidates, amenable to CRISPRi knockdown experiments, come from the separate NB07 prioritization using gene neighbor analysis, phylogenetic breadth, cross-organism conservation, domain annotations and CRISPRi tractability. All of them have high-confidence functional hypotheses derived from neighbor (operon) context and domain annotations. Each includes a specific CRISPRi design: target organism, sgRNA target, growth condition, expected phenotype and validation strategy. These are proposed tests, not validated functional assignments. For example, the #1 candidate Keio:14796 (YbeY domain) is predicted to function in ion transport because it shares an operon with an annotated ion-transport gene. It could be tested by CRISPRi knockdown in E. coli Keio grown on varied carbon/nitrogen sources, monitoring OD600 for growth defects. Three figures plot this analysis: annotated-neighbor and operon-partner rates from the gene-neighbor analysis (fig18_neighbor_analysis.png), essential-dark-gene score distributions across the 5 dimensions (fig19_essential_scores.png) and the top 20 essential dark-gene candidates with their score breakdown (fig20_essential_top20.png). The ten highest-ranked essential candidates, each labelled high hypothesis confidence, are listed with score, domains and operon context: ₁

• 1: E. coli Keio 14796; 0.875; YbeY, TIGR00043; ion transport + NTP hydrolase.
• 2: Shewanella MR-1 200382; 0.874; RimP_N, DUF150_C; tRNA-Met + NusA.
• 3: K. oxytoca BWI76_RS08540; 0.865; OmpA, TIGR02802; TolB + CpoB (cell division).
• 4: P. putida N2C3 AO356_29395; 0.838; Peptidase_M20; ABC transporter + peptidase.
• 5: Shewanella MR-1 201473; 0.835; EarP, TIGR03837; EF-P + flavodoxin.
• 6: E. coli Keio 14768; 0.833; DUF493; lipoate biosynthesis + PBP5.
• 7: Shewanella MR-1 200359; 0.833; YbeY, TIGR00043; CorC + PhoH.
• 8: P. putida PP_1910; 0.828; YceD; 50S ribosomal L32 + PlsX.
• 9: P. putida PP_5002; 0.828; GBBH-like_N; HslVU protease + PhaC1.
• 10: E. coli Keio 11474; 0.825; DUF4109; tRNA-Met + translation.

Cross-species synteny tested 21,011 dark-gene–operon-partner pairs for neighborhood conservation across up to 47 Fitness Browser organisms; these were the pairs with ortholog groups assigned to both genes. It found 17,058 pairs conserved in at least one other organism and 10,150 conserved in at least 3 organisms. The report treats this as STRING-like evidence that the pairs are functionally linked, which supports but does not prove linkage. The median conservation score was 0.95, meaning that when orthologs of a dark gene and its operon partner co-occur in another genome, they almost always remain neighbors. Independent co-fitness analysis tested 32,075 non-essential operon pairs against the Fitness Browser cofit table (top-20 partners, 13.6M rows). It used tiered scores: mutual top-5 cofit (1.0), one-directional top-5 (0.75), mutual top-20 with cofit > 0.5 (0.50) and one-directional top-20 (0.25). It found 2,899 pairs with co-fitness evidence (rank ≤ 20), including 1,129 mutual top-5 pairs. Of these, 998 pairs were double-validated by both conserved synteny (>30%) and strong co-fitness (score ≥ 0.5); the report calls them its highest-confidence functional predictions, though they remain association-based. Incorporating this evidence improved scores for 10,757 fitness-active and 4,028 essential dark genes. Of the top 100 candidates, 85/100 had existing prior module-based predictions, and all of these agreed with the project’s independent inference; the report presents this as orthogonal validation. A methods passage instead describes the synteny test as covering 21,011 pairs across 48 FB genomes. It argues that two genes sitting together in many species are unlikely to be associated by accident. ₁

The evidence-weighted experimental roadmap ranks MR-1 as the top organism with 61 candidates (47 fitness-active + 14 essential, 83 synteny-validated pairs, 31 cofit-validated pairs), followed by P. putida N2C3 (35 candidates) and P. putida (24 candidates). The report states that just 4 experiments now achieve 45% coverage of the top 300 candidates. This uses a different denominator from the earlier set-cover projection, in which 10 organism–condition experiments would address 242 of the top 500 dark genes (reported as 45.3%). The report does not reconcile the two figures, and both are projections for experiments not yet run. The evidence-weighted roadmap’s coverage curve, organism heatmap and priority breakdown are plotted in fig24_improved_experimental_roadmap.png. A methods passage says that re-scoring all 17,344 fitness-active and 9,557 essential dark genes with synteny and co-fitness evidence produced 300 improved candidates (200 fitness-active + 100 essential). That phrasing treats all 17,344 as fitness-active and adds the 9,557 essentials separately, which conflicts with the census definition of the 17,344 as including the 9,557 essential genes. It also implies co-fitness evidence for essentials, which the report elsewhere says is unavailable. ₁

The report cautions that its synteny analysis uses a 5-gene window within Fitness Browser organisms only (48 genomes). Tools such as STRING v12 and EFI-GNT analyze thousands of genomes and use more sophisticated scoring (gene fusion, shared phylogenetic profiles). Its conservation scores should therefore be treated as lower bounds, and true conservation rates are likely higher under broader taxonomic sampling. Co-fitness evidence is unavailable for essential genes: without viable mutants there are no fitness profiles and hence no co-fitness data. Cross-species synteny conservation (score distribution, tested versus conserved pairs and the top 15 pairs) is plotted in fig22_conserved_neighborhoods.png, and co-fitness validation (tier distribution, a synteny-versus-co-fitness scatter and the top 20 combined) in fig23_cofit_validation.png. A methods passage nonetheless says that, among the 32,075 tested pairs, co-fitness-confirmed operons provide the strongest functional inference for essential genes that lack direct fitness data. This conflicts with the stated non-essential scope of those pairs and with the unavailability of co-fitness for essential genes, and the report does not reconcile the two. ₁

The darkness spectrum assigned all 57,011 dark genes to tiers using 6 binary evidence flags: domain annotation, ortholog group, function prediction, co-fitness partner, fitness/essentiality phenotype and pangenome context. The tiers were T1 Void (4,273, 7.5%; no evidence of any kind), T2 Twilight (12,282, 21.5%; a single clue, domain or ortholog only), T3 Dusk (16,103, 28.2%; two converging hints), T4 Penumbra (22,500, 39.5%; substantial evidence and testable hypotheses, described elsewhere as 3–4 evidence lines) and T5 Dawn (1,853, 3.3%; 5–6 evidence lines, nearly characterized). The report proposes tier-specific follow-up: T5 Dawn genes need confirmation, T4 Penumbra genes need targeted experiments, and T1 Void genes need broad screens. Greedy weighted set-cover optimization, maximizing priority value × tractability × phylogenetic diversity over the 16,488 scored dark genes, selected 42 organisms from 28 genera to cover 95% of total priority value. MR-1 ranked first, contributing 587 genes, and 32 organisms sufficed for 80% coverage. Per-organism action plans classified 14,450 genes as hypothesis-bearing, with specific condition recommendations from fitness data, neighbor context or module predictions, and 2,038 as darkest and requiring broad phenotypic screens. They also flagged 8,900 essential genes for CRISPRi-based approaches. This covering set ranks MR-1 first, whereas the conservation-weighted covering set described above ranks S. meliloti first. The two optimize different gene sets: 16,488 scored dark genes versus 28,584 high-priority dark genes. The darkness tier distribution, evidence-flag combinations and composite score by tier are plotted in fig25_darkness_spectrum.png; set-cover diminishing returns, the per-organism tier breakdown and genus-diversity progression in fig26_covering_set.png; and the organism action-plan heatmap, condition recommendations and tractability versus gene count in fig27_action_plan.png. ₁

Reannotation and experimental coverage

Bakta v1.12.0 with DB v6.0 annotations for the 132.5M pangenome cluster representatives were cross-referenced against the 57,011-gene dark catalog. Bakta reclassified 33,105 of 39,532 pangenome-linked dark genes (83.7%) as not hypothetical. Its PSC/PSCC pipeline assigned them a product description where the Fitness Browser had only “hypothetical protein”. These genes represent 58.1% of all dark genes. All 100 top candidates received Bakta product descriptions, including homogentisate 1,2-dioxygenase HmgA (MR-1 rank 8), cell division proteins ZapE/ZapC (MR-1 ranks 13/16), N-acetylglucosamine kinase (MR-1 rank 26) and BolA family iron metabolism protein IbaG (Keio rank 31). The report presents these annotations as independent validation of the experimental prioritization. It says the dark genes with the strongest fitness effects tend to be genes annotated in UniProt but not in the Fitness Browser’s annotation vintage. 5 dark genes received AMRFinderPlus annotations: mercury-resistance transport (MerF in Marinobacter), a yersiniabactin transporter (YbtP in K. oxytoca), an acid-resistance protein (Asr in K. oxytoca), a heat-resistance membrane protein (HdeD-GI in P. stutzeri) and an S8-family peptidase (PV4). A further 6,427 genes remained hypothetical in both Fitness Browser and Bakta but gained UniRef50 links for 79.4%, UniParc/UniRef100 links for 69.1%, and RefSeq links for 62.4%. The report says these links provide cross-reference paths for literature mining even without a product description. ₁

Adding has_bakta_annotation as a 7th evidence flag to the darkness spectrum changed the darkness tier of 18,019 genes. The largest shifts were from T4 Penumbra to T5 Dawn, as genes gained their final missing evidence line. The six-flag tier counts given above do not include this flag. Bakta, the annotation tool used for this reclassification, is the basis of two figures: fig33_bakta_reclassification.png shows a pie chart of annotated, dark and unlinked genes and a tier comparison of the original versus the with-Bakta spectrum, and fig34_bakta_coverage_heatmap.png shows Bakta annotation coverage (%) by Fitness Browser dark-gene class (hypothetical, DUF, uncharacterized, no annotation). ₁

The Fitness Browser collection is taxonomically uneven: 37/48 organisms are Pseudomonadota, and the top 500 candidates include 417 from Gammaproteobacteria, 71 from Alphaproteobacteria, 17 from Deltaproteobacteria, 12 from Bacteroidetes, 10 from Betaproteobacteria, 5 from Firmicutes, 2 from Cyanobacteria, and 0 from Archaea; Actinobacteria and Epsilonproteobacteria have no organisms in the collection, so no gene count is reported for them. The report also describes the collection as heavily skewed toward Gammaproteobacteria, with 21/48 organisms and 78% of top-500 dark genes; that percentage does not agree with the 417 Gammaproteobacteria genes it reports for the top 500, and the report does not reconcile the two. An extended pool of 73 organisms, including 25 non-Fitness-Browser organisms, produced a 50-organism set covering 98.7% of ortholog groups across 6 phyla; the selected non-Fitness-Browser organisms included Pseudomonas aeruginosa PAO1, Vibrio cholerae N16961, Burkholderia cenocepacia K56-2, Mycobacterium tuberculosis H37Rv, and Campylobacter jejuni. A Spark query mapped genus-level ortholog-group membership for the 25 non-Fitness-Browser organisms across the full pangenome (24 genera, 53,970 genus-OG pairs in 2.9 min). The Fitness-Browser-only alternative was 41 organisms covering 100% of ortholog groups across 4 phyla. 16 non-Fitness-Browser organisms were selected, led by P. aeruginosa PAO1 (3,713 OGs, tractability 0.8), V. cholerae N16961 (1,049 OGs) and B. cenocepacia K56-2 (956 OGs). M. tuberculosis H37Rv ranked #6 (131 new OGs from Actinomycetota) and C. jejuni #48 (Campylobacterota). Coverage at N=5 organisms reached 57.7%, versus 38.1% for Fitness-Browser-only candidates (+19.6%). Bacillota organisms (B. subtilis, S. aureus) were not selected because their ortholog groups are subsets of coverage already provided by Pseudomonadota. The report adds that they remain valuable for studying genes in native Gram-positive genomic context. The extended covering-set comparison, covering Fitness Browser-only versus the 73-organism pool, marginal coverage, phylum diversity and kingdom-level gap coverage, is plotted in fig39_extended_covering_set.png. ₁

The report’s taxon-coverage table gives, per taxon, the number of Fitness Browser organisms, top-500 dark genes and gap severity: ₁

• Gammaproteobacteria: 21 organisms; 417 genes; Saturated.
• Betaproteobacteria: 8 organisms; 10 genes; Moderate.
• Alphaproteobacteria: 5 organisms; 71 genes; Low.
• Deltaproteobacteria: 5 organisms; 17 genes; Low.
• Bacteroidetes: 5 organisms; 12 genes; Low.
• Firmicutes: 1 organism; 5 genes; Critical.
• Cyanobacteria: 1 organism; 2 genes; Moderate.
• Archaea: 2 organisms; 0 genes; Severe.
• Actinobacteria: 0 organisms; no gene count; Critical.
• Epsilonproteobacteria: 0 organisms; no gene count; High.

Among the most widespread top-500 dark-gene clusters (present in 15+ pangenome species), the most common domain families include TauE (sulfonate export, 7 clusters), DUF444 (5 clusters), Cu-oxidase_4 (5 clusters), EamA (4 clusters, including the #1 candidate) and DUF484/DUF971/DUF934 (3+ clusters each). ₁

To fill the phylogenetic gaps, the report proposes new RB-TnSeq (randomly barcoded transposon sequencing) libraries. Its first choice is Bacillus subtilis 168 (Firmicutes), the best-studied Gram-positive model with established RB-TnSeq protocols (Koo et al. 2017); it would fill the largest phylogenetic gap and enable Gram-positive versus Gram-negative comparisons for universal dark-gene families (DUF484, DUF971, EamA), since the existing Fitness Browser strain BFirm produced only 5 top-500 genes. Second is Streptomyces coelicolor A3(2), which would add the entirely missing Actinobacteria; its large genome (8.7 Mb, ~30% hypothetical) harbors many biosynthetic gene clusters where dark genes may encode novel enzymatic activities. Third is Clostridium difficile 630, an anaerobic Firmicute with established genetic tools and TnSeq (Dembek et al. 2015), which would allow anaerobic-phenotype dark genes, which showed strong NMDC dissolved-oxygen correlation, to be tested in a strict anaerobe. Fourth is Mycobacterium smegmatis mc²155, a fast-growing non-pathogenic mycobacterium widely used as a model for M. tuberculosis, with well-established Tn-seq. The report calls it the highest-impact Actinobacteria representative, with direct translational relevance. Fifth is Campylobacter jejuni NCTC 11168, which would add the missing Epsilonproteobacteria. TnSeq has been applied to it successfully (de Vries et al. 2017), and it is an important foodborne pathogen with unusual metabolic requirements. Sixth is Lactobacillus plantarum WCFS1, a plant-associated and gut-associated Firmicute with well-developed genetic tools. It would complement B. subtilis with a different ecological niche and allow carbon-source dark genes to be tested in a fermentative bacterium. Seventh is Rhodobacter sphaeroides 2.4.1, a photosynthetic alphaproteobacterium with diverse metabolism (phototrophy, aerobic/anaerobic respiration, nitrogen fixation). It would strengthen cross-organism concordance testing for the 71 top-500 Alphaproteobacteria genes, using an organism with a different metabolic strategy from S. meliloti. The report projects that the first two additions, B. subtilis and S. coelicolor, would fill the two largest phylogenetic gaps: Firmicutes depth and Actinobacteria absence. Together they would enable cross-phylum testing of the ~100 “universal” dark-gene families that currently can only be studied in Proteobacteria. These are proposals, not completed screens. ₁

Experimental campaign: two prioritization routes

The report distills the analysis into an experimental campaign with two complementary prioritization routes. One is optimized for genes with the strongest existing evidence, the other for maximizing discovery of completely unknown functions, and the report says both converge on a tractable set of organisms. Route A, evidence-weighted prioritization (NB05–NB09), scores each dark gene across 6 dimensions using data from the 48 Fitness Browser organisms: fitness importance 0.25, conservation 0.20, inference 0.20, pangenome 0.15, biogeographic 0.10 and tractability 0.10. It identifies genes where multiple evidence lines converge on a testable hypothesis, and is intended for targeted experiments. Route B, conservation-weighted prioritization (NB11, NB11b), queries the full 27,690-species GTDB r214 pangenome to measure the taxonomic breadth of each dark-gene ortholog group (species range 1–27,482). It classifies each group by taxonomic tier (kingdom through species) and hypothesis status (strong hypothesis, weak lead, true knowledge gap), then ranks by importance = conservation × ignorance. It is intended for discovery experiments on broadly conserved genes where the least is known. ₁

In the report’s route comparison, Route A’s top organism is MR-1 (deep condition coverage, many converging evidence lines). Route B’s is S. meliloti (1,630 OGs, 195 kingdom-level gaps). Each route selects 42 organisms (28 genera); Route A covers 95% of composite priority and Route B 95.6% of importance-weighted priority. The routes share 39 organisms but produce different orderings reflecting different experimental strategies. An extended tractable-organism list of 73 organisms (48 Fitness Browser plus 25 literature-curated from 3 additional phyla) provides a path to expand experimental coverage beyond the Proteobacteria-dominated Fitness Browser. Route A uniquely selects 3 (Dda3937, Ddia6719, Miya) and Route B uniquely selects 3 (Kang, Methanococcus_JJ, Methanococcus_S2). The comparison counts S. meliloti’s 1,630 as ortholog groups (OGs). A methods summary likewise labels P. putida’s 1,043 (172 kingdom gaps) and MR-1’s 805 (105 kingdom gaps) as OGs. The conservation-weighted covering-set description above gives these counts as genes, and the report does not reconcile the unit. ₁

Route A’s top fitness-active candidates are genes where fitness data, module co-regulation, domain annotations and biogeographic signals converge on testable functional hypotheses. The wider pool is the 17,344 of 57,011 dark genes across 48 bacteria with experimentally measurable phenotypes: strong fitness defects (|fit| ≥ 2 in at least one condition) or essentiality (no viable transposon mutants). The report says these genes are not computationally predicted to matter but demonstrably matter in the lab. The functional readings below are module- and domain-based hypotheses, not validated assignments. Entries give rank, locus, organism, score, predicted function, condition and rationale: ₁

• 1: AO356_11255; P. putida N2C3; 0.715; D-alanyl-D-alanine carboxypeptidase (EC 3.4.16.4); nitrogen (fit=3.4); module prediction + EamA domain + strongest biogeographic signal (lab-field OR=44) + NMDC nitrogen correlation.
• 2: 202463; S. oneidensis MR-1; 0.698; stress-responsive membrane protein (PF01145); stress (fit=6.4); YGGT domain + co-regulated module. The report calls this the highest fitness magnitude among top candidates, although its top-20 table lists MR-1 202474 at |fit| 7.1; it does not say which candidate set the phrase refers to.
• 3: 199738; S. oneidensis MR-1; 0.698; K03306 family (nitrogen metabolism); nitrogen (fit=5.5); part of a three-gene paralog family with 203545 and 202450, so comparing mutants would test functional redundancy.
• 4: 203545; S. oneidensis MR-1; 0.694; K03306 family (DUF4124 domain); nitrogen (fit=4.0); a paralog of 199738 in the same module but with a different domain, which the report reads as suggesting subfunctionalization.
• 5: 202450; S. oneidensis MR-1; 0.693; K03306 family (Gly_transporter domain); nitrogen (fit=3.9); the third K03306 paralog, whose glycine-transporter domain the report reads as suggesting a nitrogen/amino-acid link.

Route B’s top knowledge gaps are the most broadly conserved genes with zero functional evidence, where the report expects experimental characterization to produce the most novel biological insight. All three listed groups are kingdom-level true knowledge gaps spanning 142 phyla with importance 23.9. This table places COG0491 third, while the report’s description of the importance score names COG0172 (weak lead) among three top-ranked groups. Entries give rank, group, species and the report’s note: ₁

• 1: COG0468; 27,427 species; pan-bacterial, present in virtually all sequenced genomes.
• 2: COG0443; 27,279 species; pan-bacterial, functionally uncharacterized.
• 3: COG0491; 27,393 species; pan-bacterial, zero evidence across all inference layers.

The report notes that these ortholog groups are among the most universally conserved genes in bacteria yet remain functionally uncharacterized. They are invisible to Route A because they may lack condition-specific fitness effects. The report suggests they may be essential under all conditions, producing no differential fitness signal; this is offered as a possible explanation, not a tested one. ₁

The report’s Route A organism ranking places Shewanella oneidensis MR-1 first (587 genes, tractability 0.8). MR-1 was selected because it combines deep condition coverage (121 conditions historically), a large dark-gene complement spanning all tiers (172 T5 Dawn, 257 T4 Penumbra) and high genetic tractability. 544 of its dark genes have specific condition recommendations (stress, nitrogen, carbon), and 25/100 top fitness-active candidates reside in MR-1. This ranking puts P. fluorescens N1B4 second, whereas the evidence-weighted roadmap above puts P. putida N2C3 second by candidate count; the report does not reconcile the two orderings. The next four organisms are: ₁

• 2: Pseudomonas fluorescens N1B4 (624 genes, tractability 0.7), which contributes the most additional uncovered genes after MR-1, particularly in carbon/nitrogen source panels (215 genes) and membrane stress (77 genes).
• 3: Sinorhizobium meliloti (570 genes, tractability 0.6), covering a distinct phylogenetic lineage (Alphaproteobacteria) with 316 T4 Penumbra and 90 T5 Dawn genes.
• 4: Escherichia coli K-12 (Keio) (368 genes, tractability 0.9), the highest tractability, with 259 hypothesis-bearing genes immediately testable and 205 essential genes proposed for CRISPRi.
• 5: Klebsiella oxytoca (396 genes, tractability 0.65), covering Enterobacteriaceae genes not in E. coli, including top essential candidate BWI76_RS08540 (score 0.865).

The NB11 Route B selection uses a conservation-weighted set-cover algorithm that optimizes Σ(importance) × tractability × phylo_bonus. It selected 42 organisms covering 95.6% of total importance-weighted priority. The first 10 Route B organisms cover 31% of importance-weighted priority (9,014 genes across 10 genera). Its first five organisms, with counts given as ortholog groups (OGs), are: ₁

• 1: S. meliloti (1,630 OGs, 195 kingdom gaps, tractability 0.6), selected first because its dark genes span the most kingdom-level true knowledge gaps.
• 2: P. putida (1,043 OGs, 172 kingdom gaps, tractability 0.8), combining high tractability with deep coverage of conserved unknowns.
• 3: S. oneidensis MR-1 (805 OGs, 105 kingdom gaps, tractability 0.8), third here versus first in Route A because Route B weights conservation breadth over condition-specific evidence depth.
• 4: B. thetaiotaomicron (1,382 OGs, 267 kingdom gaps, tractability 0.3), described as having the highest raw OG count of any organism but deprioritized by low tractability (0.3). The report calls it critical for non-Proteobacteria (Bacteroidota) coverage. The ‘highest raw OG count’ label conflicts with the 1,630 OGs listed for S. meliloti.
• 5: Klebsiella michiganensis (568 OGs, 65 kingdom gaps, tractability 0.65), adding Enterobacteriaceae depth beyond E. coli.

The report cautions that both covering sets draw only from the 48 Fitness Browser organisms, which are heavily biased. 37/48 are Pseudomonadota (Proteobacteria), with Bacteroidota (4), Cyanobacteriota (1) and Halobacteriota (2) the only other phyla represented. Bacillota (Firmicutes), Actinomycetota (Actinobacteria) and Campylobacterota have zero Fitness Browser representation. As a result, kingdom-level ortholog groups (55.9% of dark-gene ortholog groups, present across multiple phyla) cannot be experimentally addressed in their non-Proteobacterial hosts using Fitness Browser organisms alone. These phylum counts conflict with the report’s taxon-coverage table, which lists 1 Firmicutes organism (the Fitness Browser strain BFirm) and 5 Bacteroidetes organisms; the report does not reconcile them. ₁

To address this gap, the report curated an extended tractable-organism list of 73 organisms: 48 Fitness Browser organisms plus 25 literature-curated organisms with published TnSeq or CRISPRi resources (data/extended_tractable_organisms.tsv). The 25 non-Fitness-Browser organisms fill specific phylogenetic gaps. The C. difficile strain listed here is R20291, whereas the proposed new RB-TnSeq library above is C. difficile 630; both are cited to Dembek et al. 2015. The additions are: ₁

• Bacillota (6): B. subtilis 168, S. aureus USA300, S. pneumoniae TIGR4, C. difficile R20291, E. faecium and L. monocytogenes EGD-e; TnSeq + CRISPRi; Koo et al. 2017, Bae et al. 2004, van Opijnen et al. 2009, Dembek et al. 2015.
• Actinomycetota (2): M. tuberculosis H37Rv and C. glutamicum ATCC 13032; TnSeq + CRISPRi; DeJesus et al. 2017, Cleto et al. 2016.
• Campylobacterota (2): C. jejuni NCTC 11168 and H. pylori 26695; TnSeq; Gao et al. 2014, Salama et al. 2004.
• Pseudomonadota (15): P. aeruginosa PAO1, Salmonella 14028s, V. cholerae N16961, A. baumannii 17978 and 11 others; TnSeq/CRISPRi; Turner et al. 2015, Langridge et al. 2009, Chao et al. 2013.

An earlier, NB11b-stage passage presents incorporating these organisms into the Route B covering set as prospective. Doing so would enable experimental characterization of kingdom-level dark-gene ortholog groups in their native Gram-positive, Actinobacterial and Campylobacterial genomic contexts. It would require a Spark query mapping each non-Fitness-Browser organism’s genome to dark-gene root ortholog groups via the pangenome, which the report calls a natural extension of the NB11b pipeline. The report elsewhere describes a Spark query of genus-level ortholog-group membership for the 25 non-Fitness-Browser organisms and an extended covering set built from it (see above). This passage does not by itself establish the final covering-set status. ₁

The action plan sorts the covering-set genes into three experiment classes, all proposals rather than completed tests. 14,450 hypothesis-bearing genes across all 42 organisms have at least one condition recommendation from fitness data, module co-regulation or gene-neighborhood context. For these the experiment is targeted RB-TnSeq or CRISPRi under the predicted condition. Examples are MR-1:202463 (YGGT domain, stress-responsive) under oxidative, osmotic and heat stress, and the K03306 paralog trio (199738/203545/202450) under nitrogen limitation with single and double mutant comparisons. For 2,038 darkest genes across all 42 organisms, the report finds no converging evidence to suggest a specific condition. These are T1 Void or T2 Twilight genes with detectable fitness phenotypes, and they call for a broad phenotypic screen: RB-TnSeq across a diverse condition panel (the standard Fitness Browser protocol of ~50-100 conditions). For E. coli K-12 alone, 109 genes fall in this category. The 8,900 essential genes in the covering set cannot be studied by transposon knockout. The recommended approach is CRISPRi knockdown (Mobile-CRISPRi for non-model organisms, Peters et al. 2019) with growth curves under standard and stress conditions. The top 50 essential candidates have specific CRISPRi experiment designs in NB07. ₁

The report’s targeted campaign (Route A) proposes three first experiments, none yet run:

• An MR-1 stress screen: RB-TnSeq under 5 stress conditions (oxidative, osmotic, metal, heat, pH), addressing 161 hypothesis-bearing dark genes, including the #2 overall candidate 202463 (fit=6.4 under stress).
• An MR-1 nitrogen screen: RB-TnSeq under nitrogen limitation and amino acid supplements, addressing the K03306 paralog family (3 genes, fits 3.9–5.5) and 74 total dark genes.
• An E. coli K-12 CRISPRi screen: knockdown of the top 20 essential dark genes with growth curves, using the highest-tractability organism (0.9) and mature CRISPRi tools. The top target is Keio:14796 (YbeY domain, score 0.875). The 161 stress and 74 nitrogen counts differ from the 57 stress and 31 nitrogen top-500 genes given for the MR-1 set-cover screens above. They appear to count different gene sets, and the report does not reconcile them. ₁

The report’s discovery campaign (Route B) is meant for labs focused on novel function discovery rather than hypothesis testing. It starts with S. meliloti (195 kingdom-level gaps), P. putida (172 kingdom gaps) and MR-1 (105 kingdom gaps). It recommends broad phenotypic screens under diverse condition panels, because the target genes are true knowledge gaps without condition predictions. The report says this ordering maximizes the probability of uncovering completely new biology; that benefit is prospective, not demonstrated. ₁

Hypothesis assessment

The report assesses its central hypothesis H1 as partially supported and H0 as partially rejected. Dark genes with strong fitness effects are not randomly distributed but show non-random patterns across multiple evidence dimensions. Among these patterns, 6,142 dark genes co-regulate with annotated genes in independent-component-analysis (ICA) modules, and 85/100 top candidates have module-based function predictions. ₁

H1b, the conservation signal, was not supported. It predicted that stress-condition dark genes would be more accessory than carbon/nitrogen dark genes. A formal Fisher’s exact test on n=7,491 dark genes (NB06) found a significant association in the opposite direction: stress-condition dark genes were 23.0% accessory versus 25.5% for carbon/nitrogen dark genes (OR=1.15, p=0.013), so the predicted direction was rejected. The report reads this as suggesting that the relationship between condition specificity and pangenome conservation is more complex than a simple stress=accessory, metabolism=core dichotomy. The formal test of stress versus carbon/nitrogen accessory rates is plotted in fig16_h1b_test.png. ₁

A matched comparison of cross-organism concordance for 65 dark versus 490 annotated ortholog groups (NB06) found no significant difference (Mann-Whitney p=0.17, KS p=1.0). Dark genes with orthologs in 3+ organisms reached concordance levels indistinguishable from annotated genes (dark median=1.0, annotated median=1.0; dark mean=0.976, annotated mean=0.985). The report says this null control, in which dark genes achieve concordance indistinguishable from annotated genes, supports H1. Dark-versus-annotated concordance distributions are plotted in fig17_concordance_comparison.png. ₁

For the environmental hypothesis, the report cites 10/137 clusters with significant environmental enrichment and an overall concordance rate of 61.7%, which it says exceeds the 50% chance level. The binomial test, however, is not significant at 0.05 (p = 0.072), although Fisher’s combined p = 0.031 is. NMDC independent validation confirmed all 4 testable pre-registered predictions (nitrogennitrogen, pHpH, anaerobic~dissolved oxygen). The strongest within-species signals are in Pseudomonas and P. syringae, and the NMDC correlations provide community-level corroboration across 6,365 metagenomic samples. ₁

Literature context

The report places its approach in a line of fitness-based function prediction. Deutschbauer et al. (2011) pioneered genome-wide fitness profiling for function prediction in Shewanella MR-1, using 121 conditions to propose functions for 40 previously hypothetical genes; this project extends that approach to 48 organisms and 7,552 conditions using the Fitness Browser resource (Price et al. 2018). The report also cites Vaccaro et al. (2016), who showed that fitness profiling in Pseudomonas stutzeri RCH2 could identify novel metal-resistance genes among hypotheticals, and says its own finding of stress-responsive dark genes in Pseudomonas species corroborates that pattern. ₁

The report also positions its work against recent large-scale dark-proteome studies. It cites Pavlopoulos et al. (2023, Nature), who identified 106,198 novel protein families from 26,931 metagenomes, doubling the number of known protein families, and Zhang et al. (2025, Nature Biotechnology), whose FUGAsseM method predicts high-confidence functions for >443,000 protein families from microbial communities using metatranscriptomic coexpression. By contrast, the report asserts that every gene in its catalog has measured fitness phenotypes from controlled lab experiments and that the top candidates come with specific experimental protocols. That assertion is in tension with the report’s own statement that essential dark genes have zero genefitness rows and no fitness magnitudes beyond the essentiality call; the report does not reconcile the two. Its first listed contribution is a unified dark-gene catalog (57,011 genes across 48 bacteria) integrating fitness, conservation, module, ortholog and domain data from 4 prior observatory projects, which had previously been fragmented across separate analyses. It also describes its lab–field concordance approach, testing whether lab fitness conditions predict the field environments of gene carriers, as, to the authors’ knowledge, novel in its systematic application across multiple organisms and condition classes. ₁

Data sources and output files

The project drew on three source collections: ₁

• kescience_fitnessbrowser (tables gene, genefitness, specificphenotype, experiment, cofit, ortholog, genedomain, seedannotation, organism, specog): fitness phenotypes, gene descriptions, co-fitness, orthologs and domains.
• kbase_ke_pangenome (tables gene_cluster, gene, gene_genecluster_junction, genome, eggnog_mapper_annotations, gtdb_species_clade, gtdb_taxonomy_r214v1, gtdb_metadata, ncbi_env, alphaearth_embeddings_all_years, gapmind_pathways): pangenome conservation, phylogenetic breadth, environmental metadata and pathway analysis.
• nmdc_arkin (tables taxonomy_features, abiotic_features, taxonomy_dim, trait_features): independent environmental validation via a genus-level taxonomy bridge, and trait-condition concordance analysis.

Key output files and their row counts: ₁

• data/dark_genes_integrated.tsv: 228,709 rows; full gene table with all cross-references (43 columns).
• data/dark_genes_only.tsv: 57,011 rows; dark-gene subset with pangenome, module and ortholog links.
• data/gapmind_gap_candidates.tsv: 1,256 rows; organism–pathway pairs with dark genes near metabolic gaps.
• data/gapmind_pathway_summary.tsv: 80 rows; per-pathway completeness summary across 44 species.
• data/concordance_scores.tsv: 65 rows; cross-organism fitness concordance per ortholog group.
• data/phylogenetic_breadth.tsv: 30,756 rows; taxonomic breadth per gene cluster.
• data/biogeographic_profiles.tsv: 31 rows; species-level environmental profiles.
• data/carrier_genome_map.tsv: 8,139 rows; gene cluster to carrier genome mapping.
• data/carrier_noncarrier_tests.tsv: 151 rows; within-species carrier vs non-carrier test results.
• data/lab_field_concordance.tsv: 47 rows; pre-registered lab–field concordance test results.
• data/nmdc_validation.tsv: 105 rows; NMDC abiotic correlation tests (7 score types × 15 abiotic variables).
• data/nmdc_trait_validation.tsv: 456 rows; NMDC trait–condition correlation results (7 pre-registered + 449 exploratory).
• data/scoring_all_dark.tsv: 17,344 rows; full scoring for all strong/essential dark genes.
• data/prioritized_candidates.tsv: 100 rows; top 100 ranked candidates with hypotheses and experiments.
• data/h1b_test_results.tsv: 1 row; H1b formal test results (Fisher’s exact, chi-squared).
• data/annotated_control_concordance.tsv: 490 rows; annotated ortholog-group concordance scores for the null comparison.
• data/gene_neighbor_context.tsv: 57,011 rows; gene neighbor profiles for all dark genes (operon predictions, functional keywords).
• data/essential_dark_scored.tsv: 9,557 rows; essential dark genes scored across 5 dimensions.
• data/essential_prioritized_candidates.tsv: 50 rows; top 50 essential dark-gene candidates with CRISPRi experiments.
• data/scoring_sensitivity_nb05.tsv: 6 rows; NB05 scoring sensitivity analysis (6 weight configurations).
• data/scoring_sensitivity_nb07.tsv: 6 rows; NB07 scoring sensitivity analysis (6 weight configurations).
• data/conserved_neighborhoods.tsv: 21,011 rows; per-pair cross-species synteny conservation scores.
• data/cofit_validated_operons.tsv: 32,075 rows; per-pair co-fitness validation of operon predictions.
• data/improved_candidates.tsv: 300 rows; re-scored top candidates (200 fitness-active + 100 essential) with synteny and co-fitness evidence.
• data/experimental_roadmap.tsv: 30 rows; evidence-weighted organism experiment priorities.
• data/dark_gene_census_full.tsv: 57,011 rows; full darkness-spectrum census with tier, evidence flags and composite score.
• data/minimum_covering_set.tsv: 16,488 rows; gene-to-organism assignments from greedy weighted set cover.
• data/experimental_action_plan.tsv: 42 rows; per-organism experiment recommendations (targeted vs broad screen).
• data/og_id_root_propagation.tsv: 7,898 rows; OG_id → root_og propagation mapping (recovers 5,206 additional dark genes).
• data/dark_gene_classes.tsv: 57,011 rows; per dark gene taxonomic_tier, hypothesis_status, importance_score and evidence_summary.
• data/og_importance_ranked.tsv: 11,774 rows; all ortholog groups ranked by importance (conservation × ignorance).
• data/conservation_covering_set.tsv: 42 rows; ordered organism list with cumulative coverage for high-importance ortholog groups.
• data/conservation_experiment_plans.tsv: 42 rows; per-organism experimental plans with tier × hypothesis breakdowns.
• data/extended_tractable_organisms.tsv: 73 rows; 48 Fitness Browser + 25 literature-curated organisms with TnSeq/CRISPRi tractability scores.
• data/non_fb_genus_og_coverage.tsv: 53,970 rows; genus-level dark-gene ortholog-group coverage for 24 non-Fitness-Browser genera from the full pangenome query.
• data/extended_covering_set.tsv: 50 rows; extended covering set from the 73-organism pool, with organism ordering, coverage and phylum breakdown.
• data/gapmind_domain_matched.tsv: 42,239 rows; domain-compatible dark-gene candidates matched to gapped pathways by EC, Pfam or keyword (gene–pathway matches, not 42,239 distinct dark genes).
• data/robust_ranks_fitness.tsv: 17,344 rows; per-gene rank ranges across 6 weight configurations (fitness-active).
• data/robust_ranks_essential.tsv: 9,557 rows; per-gene rank ranges across 6 weight configurations (essential).
• data/scoring_species_count_variant.tsv: 17,344 rows; species-count scoring variant with original vs adjusted ranks.
• data/og_pangenome_distribution.tsv: 11,774 rows; per root_og species/phyla/class/order/family/genus counts from the full 27,690-species pangenome (species range 1–27,482).
• data/og_pangenome_distribution_fb_only.tsv: 11,774 rows; backup of the original Fitness Browser-only distribution (species range 1–33).
• Notebook 11b_extended_conservation.ipynb: full-pangenome Spark query exploding eggNOG_OGs across 93.5M annotations, plus OG_id propagation for extended dark-gene coverage.
• Notebook 11c_extended_covering_set.ipynb: genus-level pangenome query for 25 non-Fitness-Browser organisms and an extended covering set with 73 candidates (50 selected, 6 phyla, 98.7% coverage); this 98.7% is ortholog-group coverage, not importance-weighted priority.

Caveats and Limitations

Environmental metadata are sparse: AlphaEarth embeddings cover only 28% of genomes (83K/293K), and NCBI isolation-source metadata are inconsistent, limiting the power of biogeographic tests. NMDC validation is genus-level (mapping NMDC taxon columns to pangenome genera via ncbi_taxid) and may miss species-specific dark-gene signals, only 5 of 6 carrier genera were matched, and common genera such as Pseudomonas, Klebsiella, and Bacteroides may generate broad correlations with abiotic variables independent of specific dark-gene functions; the report says the high significance rate (76/105) likely reflects the dominance of such common genera in both datasets. ₁

The 57,011 dark-gene count likely overestimates true functional darkness because annotations may exist in databases or releases not checked, such as UniProt, InterPro or recent NCBI updates. Module predictions, which cover 6,142 dark genes in the integrated census and are reported as 6,691 in a later project-specific accounting, are guilt-by-association inferences rather than direct experimental validation; the “high confidence” label in prioritization reflects evidence convergence, not experimental proof. ₁

Fitness Browser condition coverage is uneven, and MR-1 has 121 historical conditions; organisms with deeper condition coverage can therefore produce more specific phenotypes and receive higher prioritization scores. The GapMind analysis is limited to amino-acid biosynthesis and carbon-utilization pathways, identifies organism-level co-occurrence rather than direct gene-to-step assignments, and does not address signaling, regulation, or structural functions comprehensively. Curated pathway–enzyme domain matching (EC prefix, Pfam family and keyword compatibility) partially addresses the gene-to-step gap, identifying 42,239 domain-compatible gene–pathway candidates including 5,398 high-confidence EC matches, but full gene-to-step validation would require AlphaFold structure prediction or experimental enzymology. ₁

Essential genes are penalized in fitness-centric scoring because they lack viable transposon mutants, no genefitness rows, and no differential fitness magnitudes; the separate essential-gene prioritization partly corrects this bias but relies on neighborhood and domain inference. None appear in the NB05 top 100 because dimension 1 (fitness importance) scores them at the essentiality bonus floor and dimension 6 (experimental tractability) penalizes them for being non-knockable. The neighborhood heuristic uses a five-gene window, same-strand orientation, and gaps of no more than 300 bp, whereas tools such as DOOR, STRING, and EFI-GNT use broader taxonomic and multimodal evidence. Cross-species synteny validation across 48 Fitness Browser genomes and co-fitness confirmation add to the positional heuristic, but the analysis still lacks gene-fusion detection, regulatory-element analysis and broad taxonomic sampling. ₁

The NMDC trait correlations are vulnerable to compositional coupling: 441/449 exploratory tests were significant despite only 7/7 pre-registered trait directions being the meaningful directional metric. A full sample-label permutation test remains future work requiring Spark access to raw per-sample matrices. Partial null approximations instead give a sign test of p = 0.0078 for the 7/7 pre-registered trait directions and p = 0.0625 for the 4/4 pre-registered abiotic directions, which the report says confirms non-random directionality; Fisher’s combined probability across all 11 pre-registered tests of p ≈ 0; and an inflation factor of ~20× for exploratory tests (441/449 significant versus 22.5 expected), which quantifies the compositional coupling. The report adds that these NMDC correlations demonstrate ecological co-occurrence rather than causal gene–phenotype links. Likewise, the 29/47 lab–field concordance rate has a one-sided binomial p = 0.072, and equivalent annotated-accessory-gene controls were not run through the complete biogeographic pipeline. ₁

The prioritization weights are expert-assigned and exact ranks are sensitive to them. Overall rank correlations remained ρ > 0.93 across six alternative configurations, but only 32/50 (64%) of the original fitness-active top 50 remained under conservation-dominant or drop-tractability settings, while essential-gene top-50 retention was 18/50 (36%) when tractability was dropped and 24/50 (48%) when neighbor context was dropped. Across all 6 fitness-active weight configurations, 18 genes remain in the top 50 and 35 in the top 100 regardless of weight choice; for essential genes, 6 remain always-top-50 and 19 always-top-100. The report calls these always-top candidates the most defensible targets for experimental follow-up. Robust rank indicators, namely rank-range distributions, a min/max rank scatter and always-top-50 candidates by organism, are plotted in fig29_robust_ranks.png. ₁

The Fitness Browser’s 77% Pseudomonadota composition (37/48 organisms) limits cross-phylum inference. The extended covering set, which incorporates 25 non-Fitness-Browser organisms from Bacillota, Actinomycetota and Campylobacterota, improves coverage from 4 to 6 phyla, but non-Fitness-Browser ortholog-group coverage is estimated at the genus level and may overestimate individual-organism coverage; these organisms also lack Fitness Browser condition profiles and therefore support broad screens rather than condition-specific experiments. Bacillota organisms (B. subtilis, S. aureus) were not selected by the covering-set algorithm because their ortholog groups are subsets of coverage provided by Pseudomonadota, despite their value for studying genes in native Gram-positive contexts. ₁

Future Directions

The report proposes a dual-route experimental campaign, none of it yet run. Route A (NB09, evidence-weighted) identifies 42 organisms covering 95% of composite priority and would start with MR-1 stress/nitrogen screens and E. coli CRISPRi for targeted hypothesis testing. Route B (NB11, conservation-weighted) identifies 42 organisms covering 95.6% of importance-weighted priority and would start with S. meliloti, P. putida and MR-1 broad phenotypic screens for novel function discovery. The two routes share 39 organisms; the report calls them complementary, with Route A producing mechanistic insights for genes with hypotheses and Route B discovering functions for broadly conserved true knowledge gaps. These priority percentages use different denominators from the 98.7% ortholog-group coverage of the extended covering set. ₁

For targeted validation, the report says the top 5 fitness-active candidates (4 in MR-1, 1 in P. putida N2C3) are immediately testable via RB-TnSeq under their predicted condition classes. The specific targets are AO356_11255 (D-alanyl-D-alanine carboxypeptidase prediction, EamA domain), to be tested under nitrogen limitation; MR-1 202463 (YGGT domain, T5 Dawn, composite score 0.698), to be tested under multiple stress conditions; and the MR-1 K03306 paralog family 199738/203545/202450, to be tested under nitrogen limitation with single/double mutant comparisons. The functional labels are predictions to be tested. ₁

For CRISPRi validation of essential dark genes, the top 50 essential candidates (NB07) include specific CRISPRi experiment designs. The highest-priority targets are Keio:14796 (YbeY domain, score 0.875, predicted ion transport), MR-1:200382 (RimP_N/DUF150_C, score 0.874, predicted ribosome assembly) and Koxy:BWI76_RS08540 (OmpA/TIGR02802, score 0.865, predicted cell division). The covering set assigns 8,900 essential genes to specific organisms for CRISPRi-based approaches. ₁

The report proposes AlphaFold2 structure prediction for the top 100 candidates and especially the 1,853 T5 Dawn genes (nearly characterized), saying it could confirm functional hypotheses and distinguish between DUF families. It says the 4,273 T1 Void genes, which lack all evidence, would particularly benefit from structure-based function prediction as the only remaining computational inference strategy; such predictions would remain computational, not experimental validation. ₁

The NMDC dataset includes proteomics (346K observations) and metabolomics (3.1M observations) that were not used in this project. The report proposes correlating dark-gene carrier abundance with metabolite or protein profiles as potentially more direct functional evidence than abiotic correlations alone; this is a proposed integration, not a reported result. ₁

The report’s future-directions list describes expanding the covering set beyond Fitness Browser organisms as still pending: the current covering sets draw only from 48 Fitness Browser organisms, which are 77% Pseudomonadota, and the extended tractable-organism list (data/extended_tractable_organisms.tsv) curates 25 additional organisms with published TnSeq/CRISPRi resources from Bacillota (6), Actinomycetota (2), Campylobacterota (2) and additional Pseudomonadota (15). Incorporating them into Route B’s conservation-weighted covering set would enable experimental characterization of kingdom-level OGs (55.9% of dark gene OGs) in their native Gram-positive, Actinobacterial and Campylobacterial contexts, and the passage says this requires a pangenome Spark query mapping each organism’s genes to dark-gene root_ogs, an extension of the NB11b pipeline. This conflicts with the report’s own output inventory, which lists the genus-level query and an extended covering set (11c_extended_covering_set.ipynb, data/extended_covering_set.tsv); the report does not reconcile the two. ₁

The report also proposes publishing the darkness-spectrum census (dark_gene_census_full.tsv) and the experimental action plan as a community resource for bacterial functional genomics, so other labs could target specific organisms, tiers or condition classes; publication is proposed, not reported as completed. ₁

Slots Into

• gene-essentiality — essential dark genes, neighborhood-based inference, CRISPRi prioritization, and the 9,557-gene essentiality analysis. ₁
• condition-specific-fitness — strong fitness phenotypes, condition classes, cross-organism concordance, and targeted RB-TnSeq experiments.
• pangenome-integration — pangenome links, full GTDB r214 conservation, ortholog-group propagation, taxonomic tiers, and covering-set optimization.
• metabolic-model-gapfilling — GapMind pathway gaps and domain-compatible candidate gene–pathway matches.
• environment-embedding-geography — carrier-versus-non-carrier environmental enrichment, AlphaEarth divergence, and lab–field concordance.
• multi-omics-integration — NMDC abiotic and trait-feature validation and proposed proteomic and metabolomic follow-up.
• cofitness-network-architecture — fitness modules, co-fitness validation, conserved neighborhoods, and double-validated operon predictions.
• cross-tenant-data-bridging — integration of Fitness Browser, pangenome, GTDB, GapMind, NMDC, and Bakta evidence into one catalog.
• experimental-prioritization-of-functional-dark-matter — dark-gene census, candidate scoring, the experimental roadmap, and the evidence-weighted (Route A) versus conservation-weighted (Route B) organism selections.
• research-attention-inequality — uneven dark-gene fractions across organisms, attributed to differences in annotation depth.
• structural-annotation-gap — annotation-depth variation that inflates the hypothetical-gene fraction.
• computational-pathway-prediction-validation — GapMind gaps are organism-level co-occurrences that need EC, structural or experimental confirmation.
• biosynthetic-prototrophy-and-auxotrophy — asparagine-biosynthesis gaps across diverse phyla.
• evidence-triangulation-for-functional-annotation — tiered EC, Pfam and keyword domain matching for gap candidates.
• cross-species-fitness-transferability — ortholog groups whose fitness phenotypes agree across species.
• comparative-conservation-metric-calibration — the coarse eggNOG breadth classes versus species-count scoring.
• ontology-and-category-schema-sensitivity — almost every cluster falls into the universal breadth category.
• confirmatory-exploratory-ecological-association-discordance — few carrier enrichments, and pre-registered versus exploratory tests.
• lab-field-fitness-concordance — pre-registered lab–field concordance and carrier environmental enrichment.
• fitness-condition-coverage-prioritization-bias — stress-dominated phenotypes and MR-1’s deep condition coverage driving its projected value.
• pathway-versus-reaction-evidence-resolution — GapMind gaps resolved only at organism level, not gene-to-step.
• genetic-perturbation-coverage-bias — taxon-coverage gaps in the Fitness Browser collection and proposed new RB-TnSeq libraries.
• genomic-under-representation — absent Actinobacteria and Epsilonproteobacteria and sparse Archaea and Firmicutes coverage.
• core-gene-annotation-paradox — widespread dark-gene domain families such as TauE, DUF444 and EamA, and the kingdom-level dominance of dark-gene ortholog groups.
• genomic-dispersal-functional-coupling — dark genes co-localize with other dark genes, and dark-gene–operon-partner neighborhoods are conserved across organisms.
• two-speed-bacterial-genome — the rejected H1b prediction that stress-condition dark genes are more accessory than carbon/nitrogen dark genes.
• fitness-matched-null-models — dark and annotated genes show statistically indistinguishable cross-organism concordance.
• sampling-depth-and-downsampling-effects — ortholog-group species-count ranges widen sharply when moving from Fitness Browser-only to full-pangenome sampling.
• transposon-callability-bias — essential dark genes have no genefitness rows, so fitness-centric scoring cannot rank them.
• selection-on-outcome-leakage — shared genera feed both carrier abundance and community trait scores in the NMDC analysis.
• taxonomic-resolution-dependent-functional-inference — genus-level NMDC validation may miss species-specific dark-gene signals.
• cultivation-collection-bias-in-ecological-genomics — sparse AlphaEarth coverage and inconsistent isolation-source metadata limit biogeographic tests.