Contributions
BERDL Data Atlas — Inventory, Topic Map, and Cross-Reference Synergies inventoried the KBase Data Lakehouse 1. It catalogued 1,740 deduplicated tables across 119 databases, 17 tenants and 10 funding agencies or programs, and mapped 536 schema-level cross-tenant bridges defined by 29 canonical join keys 1. The BERDL Data Atlas project found that DOE accounts for approximately 78% of the tables and that the largest entity table, kbase_ke_pangenome.gene, holds 1,011,650,903 rows 1. It also sample-validated UC1, a bridge joining FitnessBrowser to AlphaFold through SwissProt best hits, and the report describes itself as a catalog of data systems rather than a test of biological hypotheses 1.
Functional Dark Matter — Experimentally Prioritized Novel Genetic Systems found that 57,011 of 228,709 genes (24.9%) across 48 Fitness Browser organisms lack functional annotation 2. Of those, 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, so 17,344 have measurable phenotypes 2. The same project identified 1,256 organism–pathway pairs across 44 species in which GapMind (a tool that scores pathway completeness from genome sequence) found nearly complete pathways co-occurring with strongly fitness-active dark genes 2. The report presents these as organism-level co-occurrences supporting the hypothesis that dark genes may encode the missing steps, not as gene-to-enzyme assignments 2.
Truly Dark Genes — What Remains Unknown After Modern Annotation? reannotated dark genes with Bakta v1.12.0 3. Of the 39,532 pangenome-linked dark genes, 33,105 (83.7%) were reclassified and 6,427 remained hypothetical, while the other 17,479 could not be assessed 3. The project found that truly dark genes are shorter than annotation-lag genes (median 121 vs 194 amino acids) and less often core (43.1% vs 72.7%) 3. However, the report’s claim that every effect exceeded its pre-registered thresholds is internally inconsistent for the essential-fraction comparison, whose odds ratio (OR) is 1.420 against a threshold of OR ≥ 1.5 3. It also found that 12.0% of truly dark genes lie within 2 genes of a mobile genetic element, a figure limited to the 4,394 genes in organisms with ICA (independent component analysis) fitness-module data 3. That proximity supports horizontal transfer but does not establish it 3.
Gene Function Ecological Agora built an atlas on the Genome Taxonomy Database release r214 (GTDB-r214) spanning 18,989 species representatives, with 13.74M producer/participation scores, which classify clade × function tuples by paralog-expansion (producer) and participation behavior into Innovator-Isolated, Innovator-Exchange, Sink/Broker-Exchange and Stable categories 4. The atlas also holds 17.07M Sankoff-parsimony gain events, Sankoff parsimony being a weighted-parsimony reconstruction of gene gains on a tree 4. Two of its four pre-registered hypotheses were confirmed: Mycobacteriaceae × mycolic-acid as Innovator-Isolated, and Cyanobacteria × photosystem II (PSII) as Innovator-Exchange at class rank 4. One further hypothesis was qualified or reframed, the Alm 2006 point estimate was not reproduced at GTDB scale, and the pre-registered Bacteroidota hypothesis remains falsified at the pre-registered absolute-zero criterion 4.
Polyhydroxybutyrate Granule Formation Pathways — Distribution Across Clades and Environmental Selection found that 21.9% of 27,690 GTDB species carry phaC and 21.7% have a complete polyhydroxybutyrate (PHB) pathway, with the highest prevalence in Pseudomonadota at 60.9% 5. PHB presence was enriched in variable environments (chi2 = 1,656.36), and this held within genome-size quartiles, for example a 4.4x enrichment in Q1 with p = 1.18 x 10^-11 5. The PHB–niche-breadth association (rho = 0.106), however, became partial rho = -0.047 after controlling for genome size, so the report treats genome size as the primary correlate 5.
Prophage Gene Modules and Terminase-Defined Lineages Across Bacterial Phylogeny and Environmental Gradients found prophage-associated gene clusters, 4,005,537 in total, in all 27,702 analyzed species, with 34.9% of species carrying all 7 modules 6. PERMANOVA (permutation-based multivariate analysis of variance) on a 1,773-species subsample gave F=212.99 for genome size, F=30.04 for environment and F=6.17 for phylogeny 6. After controlling for genome size, niche breadth correlated with prophage module count at partial rho=0.468 6. The report notes that its module interpretations are limited by the absence of dedicated prophage detection 6.
Plant Microbiome Ecotypes — Compartment-Specific Functional Guilds and Their Genetic Architecture identified 1,136 plant-associated species and 7,995 genomes 7. On 607 compartment species it found PERMANOVA R² = 0.071 and db-RDA (distance-based redundancy analysis) location-only R² = 0.060, alongside significant dispersion heterogeneity (PERMDISP F = 15.6, p = 0.001), so the PERMANOVA R² is treated as an upper bound and the compartment hypothesis was rated weakly supported 7. Adversarial review showed that the earlier Phase 1 R² = 0.527 fell to 0.072 after excluding the three genome-rich species per compartment 7. Under within-genus label shuffling, only 3/15 markers survived 7.
Genotype × Condition to Phenotype Prediction from ENIGMA Growth Curves assembled 27,632 ENIGMA growth curves across 123 strains and 195 molecules 8. Of these, 15,227 (55.1%) showed no detectable growth, and fit-ok curves had median µmax = 0.028 h⁻¹ 8. The project built a 46,389-pair modeling corpus and distinguished predictable binary growth capability from poorly predictable continuous kinetics 8. It also identified 8 metabolic guilds from KO (KEGG Orthology) profiles 8. Across 587 Oak Ridge communities it reported two anti-correlated genus clusters, whose members averaged pH 6.78 versus pH 5.43 in global 16S samples 8.
ENIGMA Carbon Census — Tiered Knowledge Census of 83 Enrichment Compounds resolved all 83 enrichment compounds to structures 9. It found 9 callable and 74/83 (89%) organism-dark, a resource-defined gap meaning their genetic determinants of utilization were not linkable through the queried the KBase Data Lakehouse and curated resources 9. One of the 9, lauric acid, was callable only through a reference bacterium, and one isolate call, xanthine, was later identified as a carbon-category error 9.
SSO Subsurface Community Ecology — Spatial Structure, Functional Gradients, and Hydrogeological Drivers analyzed 23,458 sediment ASVs (amplicon sequence variants, exact error-corrected amplicon sequences) across 9 SSO wells 10. A Mantel test, which correlates ecological and geographic distance matrices, detected distance-decay (Spearman ρ = 0.323, p = 0.029), and the project found that hydrogeological zone explained 27.5% of variance, while well identity explained 19.2% and was not significant 10. The project’s contamination-plume interpretation of the U3–M6–L7 corridor remains unconfirmed because no geochemistry was analyzed 10.
Regulatory and proteomic architecture of Δfur-permitted lipid A loss in *Caulobacter crescentus* found that the Δfur ΔsspB contrast correlated with a published Δfur signal across 93 DEGs (differentially expressed genes) at Spearman ρ = 0.315, with 71% sign concordance, and that 53 of those genes were buffered 11. Using RB-TnSeq (random barcode transposon sequencing) data, the concordant Fur signature was marginally enriched for envelope-stress phenotypes (1.60×, p = 0.016), while the SspB-buffered set was indistinguishable from background (1.04×, p = 0.515) 11. The mechanistic role of the buffered respiratory program therefore remains a hypothesis 11.
Rational Design of Protective Microbiome Formulations for Competitive Exclusion of *Pseudomonas aeruginosa* in Cystic Fibrosis Airways integrated 4,949 isolates from 175 patient samples 12. It found that the Pseudomonas aeruginosa reporter strain PA14 grew best on amino acids under synthetic cystic-fibrosis sputum conditions, reaching an endpoint optical density (OD) of 0.60 on proline, but noted that other reporter pathogens have distinct profiles, so its design is PA14-targeted and not shown to generalize 12. The report leaves its substrate counts (21, 22, and 20 amino acids plus glucose and lactate) unreconciled 12.
Metagenome-Prioritized Phage Cocktails for Crohn’s Disease and IBD selected four ecotypes from 8,489 metagenomic samples using consensus K = 4, with a cross-method adjusted Rand index (ARI, a chance-corrected measure of clustering agreement) of 0.131 13. It reported that Crohn’s disease in species–metabolite space is dominated by one canonical correlation axis (r = 0.96) and identified six actionable Tier-A pathobionts 13. From these it produced per-patient cocktail drafts for 14 of 23 UC Davis patients 13. Its clinical-covariate classifier reached a macro one-vs-rest (OvR) area under the ROC curve (AUC) of 0.799, but agreed with metagenomic ecotype calls in only 41% (9/22) of cases 13.