Taxonomic nomenclature and identifier reconciliation across databases

Taxonomic nomenclature and identifier reconciliation is the process of aligning names, accessions, taxonomy versions, and database-specific identifier scopes before comparing or joining microbial records. ₁ The central risk is that a syntactically successful join can connect biologically different organisms or silently discard valid matches when databases use different naming conventions, taxonomic ranks, or identifier namespaces. ₁ This concept extends provenance-aware-resource-discovery and taxonomic-resolution-dependent-functional-inference by treating identifier provenance and taxonomic scope as analytical evidence rather than mere preprocessing details. ₁

Why reconciliation is necessary

Short strain names are not globally unique across resources; ENIGMA field isolates often carry short names such as MT20, MT42, GW460-LB6 and FW507-14TSA that can match completely unrelated organisms when matched against kbase.ke_pangenome.gtdb_metadata.ncbi_strain_identifiers. ₁ In one documented example, ENIGMA MT20 referred to Rhodanobacter glycinis, whereas GTDB MT20 referred to Streptococcus pneumoniae. ₁ The resulting erroneous match involved 8,434 genomes, including 1,751 clinical genomes, and contaminated genus–environment profiles with spurious clinical data. ₁ Among 32 pangenome linkages made through ncbi_strain_identifiers, 12 were incorrect genus matches. ₁ The digest’s fix is to cross-check the genus from the source database (e.g., enigma.genome_depot_enigma.browser_taxon) against the genus of the GTDB match and to reject linkages where the genera disagree. ₁ These results directly support using assembly accessions such as GCF_* instead of short strain names whenever possible. ₁

The originating project report independently states that matching ENIGMA strains by ncbi_strain_identifiers caused 12 of 32 genus-level mismatches, which supports the figure carried by the pitfalls digest. ₂ In that report, MT20 (ENIGMA: Rhodanobacter glycinis) matched unrelated NCBI organisms (Streptococcus pneumoniae), injecting 1,751 spurious clinical genomes into environmental profiles. ₂ The fix used CORAL brick 522 (GTDB-Tk assignments) for authoritative taxonomy and cross-checked genus consistency; it reduced verified linkages from 32 to 20 but eliminated all false matches. ₂ The report frames this strain-name collision as a systematic data-integration hazard for any project linking field isolates to reference databases by short strain identifiers. ₂

Species names also diverge through taxonomic reclassification and database-specific lineage formats. ₁ MetaPhlAn3 cross-cohort analyses encountered short names, full lineage strings, GTDB-renamed Bacteroides vulgatus and Phocaeicola vulgatus, Eubacterium rectale and Agathobacter rectalis, Ruminococcus gnavus and Mediterraneibacter gnavus, plus multiple Clostridium, Lactobacillus, and related reclassifications. ₁ These examples show why simple string normalization is insufficient for cross-cohort taxonomic integration. ₁

The CrohnsPhage mart record separates this problem into three taxon-key incompatibilities. ₁ First, format: short names such as Faecalibacterium prausnitzii versus full lineage strings (k__Bacteria|p__Firmicutes|...|s__Faecalibacterium_prausnitzii) from the same database in a different export format. ₁ Second, GTDB r214+ genus renames: Bacteroides vulgatus (CMD_HEALTHY lineage) and Phocaeicola vulgatus (CMD_IBD short name) share an NCBI taxid but were treated as separate species, and the same affects B. dorei / coprocola / plebeius / salanitronis / sartorii / massiliensis / coprophilus, all now Phocaeicola. ₁ Third, other reclassification splits: Eubacterium rectale → Agathobacter rectalis, E. eligens → Lachnospira eligens, E. hallii → Anaerobutyricum hallii, Ruminococcus gnavus → Mediterraneibacter gnavus, Clostridium bolteae / clostridioforme / symbiosum / innocuum / asparagiforme / hathewayi / citroniae / aldenense → Enterocloster / Hungatella / Erysipelatoclostridium, and Lactobacillus mucosae / ruminis → Limosilactobacillus / Ligilactobacillus. ₁ This refines the list above by distinguishing export-format mismatches within one database from genuine nomenclatural change. ₁ The ibd_phage_targeting report likewise carries R. gnavus as an alias of mediterraneibacter gnavus. ₃

GTDB names can also diverge from NCBI names in ways that encode distinct clades rather than synonyms: the cf_formulation_design report uses gtdb taxonomy throughout and notes that pseudomonas e denotes a non-aeruginosa Pseudomonas clade distinct from P. aeruginosa, so the two names must not be merged. ₄

Evidence for analytical consequences

In the CrohnsPhage mart, failure to reconcile taxonomic names produced a log₂FC approximately 28, equivalent to approximately 2.68 × 10⁸ fold, because one cohort’s abundance collapsed to a pseudocount. ₁ This supports the interpretation that apparent biological extremes can be generated by unmatched taxon names rather than by measured abundance differences. ₁ The digest attributes the artifact to CMD_IBD using the modern names while CMD_HEALTHY uses legacy names, so a naive pivot gives log₂FC ≈ 28 (≈ 2.68 × 10⁸ fold) for every renamed species, not only for a single taxon. ₁ The committed ref_taxonomy_crosswalk does not fully resolve this. Legacy Bacteroides_X rows have ncbi_taxid = NaN, so joining through taxid leaves them orphaned, and the crosswalk lists B. vulgatus and P. vulgatus as two separate rows with different canonical_name values. ₁

The digest lays out three tiers of reconciliation. ₁ The first tier is format normalization: parse the s__Genus_species component from full lineage strings and strip brackets from short names to get a canonical “Genus species”. ₁ The second tier is a hand-curated synonym map, which is enough for a targeted species battery; NB00 in ibd_phage_targeting ships a 23-entry SYNONYM_MAP covering the major renames. ₁ The third tier is systematic reconciliation: an NCBI-taxid-backboned synonymy layer with GTDB-version-aware rename tables. This tier is mandatory before any cross-cohort aggregation over the full taxonomy rather than a curated battery, and it is an NB01 dependency in ibd_phage_targeting. ₁

The ibd_phage_targeting project built such a layer: data/species_synonymy.tsv maps 2,417 aliases to 1,848 canonical species, grounded in ref_taxonomy_crosswalk NCBI taxid matching with GTDB r214+ genus renames (Bacteroides → Phocaeicola etc.) supplemented. ₃ The digest notes that the committed crosswalk leaves legacy Bacteroides_X rows without taxids. Read together with that note, this refines the design: taxid matching through the crosswalk alone is insufficient, which suggests the hypothesis that the GTDB rename supplement carries much of the Bacteroides → Phocaeicola reconciliation. _3,1 It was motivated by a failure mode discovered in NB00: fact_taxon_abundance contains three divergent taxon-name formats between cohorts (CMD_IBD short names, CMD_HEALTHY full MetaPhlAn3 lineage, KUEHL_WGS Kaiju), and a naive pivot splits the same species into multiple zero-overlap rows, producing log₂FC ≈ 28 artifacts. ₃ This supports the pitfalls account of the CrohnsPhage artifact and provides a built instance of the NCBI-taxid-backed, GTDB-version-aware design that the digest says full-taxonomy aggregation requires. _3,1 The layer was built once in NB01b and joined against by NB00, NB01, NB02, NB03 and every downstream notebook; the report describes it as the project-wide reconciliation backbone that made CMD pooling and Kaiju → MetaPhlAn3 projection tractable, and proposes the NCBI taxid + GTDB-version-aware rename-table pattern as a candidate convention and reusable artifact for any multi-cohort microbiome project. ₃ That generalization is proposed from one project rather than tested across others. ₃

Downstream uses show what the layer enables and where it stops. HMP2 MetaPhlAn3 profiles pulled from curatedMetagenomicData v3.18 comprised 1,627 samples, 130 subjects and 582 species, with 255/335 training-feature overlap after synonymy; the report does not state whether the non-overlapping features reflect residual naming gaps or true absence. ₃ Its A6 biosynthetic gene cluster (BGC) criterion used synonymy-inverted matching against ref_bgc_catalog so that pre-GTDB names like “Ruminococcus gnavus” match canonical Mediterraneibacter gnavus, scoring 1.0 if ≥ 1 BGC contains CD-enriched CB-ORFs at effect > 0.5 and FDR (false discovery rate, the expected share of false positives among calls) < 0.05. ₃

Taxonomic rank and identifier scope must also be checked inside a single database family. ₁ Pangenome taxonomy tables must be joined through genome_id, not gtdb_taxonomy_id, because the genome table stores a genus-level taxonomy string while gtdb_taxonomy_r214v1 stores a species-level string; joining on gtdb_taxonomy_id returns zero rows. ₁ A gtdb_species_clade_id has the form s__Genus_species--{representative_genome_id} (e.g., s__Escherichia_coli--RS_GCF_000005845.2), and how the embedded -- is handled depends on the query method. ₁ gapmind_pathways.clade_name uses the full gtdb_species_clade_id, including the representative accession, rather than the shorter GTDB_species value (e.g., s__Rhizobium_phaseoli--RS_GCF_001234567.1 rather than s__Rhizobium_phaseoli). Joining it against a temp view populated from gtdb_species_clade.GTDB_species therefore returns 0 rows. ₁ In addition, 12 pangenome records reference species clades not present in gtdb_species_clade, so even a correctly formatted clade join can leave orphan pangenomes. ₁ These constraints refine pangenome-integration by making the join key and taxonomic level explicit. ₁

Genome-accession prefixes form a separate identifier-format layer. ₁ The phylogenetic_tree_distance_pairs table uses bare NCBI accessions (GCA_001038305.1, GCF_000005845.2), while the genome, gene and gene_genecluster_junction pangenome tables use GTDB-prefixed IDs (GB_GCA_001038305.1, RS_GCF_000005845.2). Joining phylogenetic distances with presence matrices therefore produces zero matches. ₁ Stripping the three-character GB_ or RS_ prefix from GTDB genome IDs before joining gave 100% overlap for all tested species (399/399 for Koxy, 287/287 for Btheta). ₁ kbase.ke_pangenome.gapmind_pathways likewise stores genome IDs without the prefix (e.g., GCF_001307155.1 versus RS_GCF_001307155.1), so direct equality matching against pangenome genome IDs returns zero rows. For this case the digest recommends a fallback chain: the original ID, then the prefix-stripped ID, then an accession-based partial match. ₁

The plant_microbiome_ecotypes project shows that this format mismatch can masquerade as a biological null. ₅ NB12’s null result for H7 (5 species tested, 0/5 significant) was caused by a genome-ID format mismatch between phylogenetic_tree_distance_pairs (bare NCBI accessions) and genome_environment.csv (GTDB-prefixed IDs). The is_plant_associated column was entirely NaN after the merge, and in the Phase 2 pass, which used 22.6M phylogenetic-distance rows, all 5 species showed 0 testable subclade × plant-association enrichments. ₅ NB13 fixed this by prepending GB_ to GCA_ accessions and RS_ to GCF_ accessions, giving 100% overlap (1306/1306 genomes) and recovering 599 plant-associated genomes that NB12 had silently dropped. ₅ The repair fixed the plumbing but left the biological signal modest. The original 5-species set showed 1/5 significant under all metrics, and H7 remained weak (1/5 species) after statistical validity checks. ₅ The full scan was then re-run on all 65 plant-associated species with ≥20 genomes, but only 18 of 65 candidate species had any pairwise phylogenetic-distance data in phylogenetic_tree_distance_pairs, and 5/17 testable species were significant after Bonferroni correction. ₅ The other 47 species are absent from the tree; the report calls this a database-coverage limitation that bounds H7 power irrespective of methodology, so the true H7 effect across all 65 candidate species cannot be estimated and the 5/17 result should not be generalized to them. ₅ An H6 host × subclade extension found 2/9 species with robust pathovar-host segregation (xanthomonas campestris × Brassica, xanthomonas vasicola × Zea mays). ₅ This supports the rule that unmatched or empty joins be audited before being read as biological absence. It also refines the pitfalls account by showing the direction of error: the prefix mismatch turned a result with 1/5 significant species into an apparent 0/5 null. _5,1

Pangenome gene-cluster identifiers are species-specific and cannot be compared directly across species. ₁ Cross-species comparisons should instead use shared representations such as COG categories, KEGG orthologs, or Pfam domains. ₁ eggnog_mapper_annotations.query_name joins to gene_cluster.gene_cluster_id, not gene.gene_id, and Bakta and InterProScan annotations likewise join at the gene-cluster level. ₁ This supports evidence-triangulation-for-functional-annotation and evidence-triangulation-for-functional-annotation by showing that both the biological representation and its join level constrain valid comparisons. ₁

Reconciliation problems extend below the taxon level to gene symbols. In caulobacter-crescentus, NCBI annotation has false negatives for genes annotated under non-canonical names: MsbA (NCBI Cc=0) is present as CCNA_00307 (“phospholipid/LPS ABC transporter”), and LptA (NCBI Cc=0) is present but not annotated with the “lptA” symbol in Caulobacter NA1000, because the genome is annotated primarily with CCNA_* locus tags rather than canonical Gram-negative gene symbols. ₆ The report classifies these as naming-convention false negatives, not true absences, which supports auditing unmatched records separately rather than reading them as biological absence. ₆

Species-name bridges between phenotype and pangenome databases

A bacdive–GTDB species-name bridge linked 42,227 BacDive strains (43.4% of 97,334) to metal tolerance scores across 6,426 unique GTDB species. ₇ Matching used exact species-name agreement for 33,535 strains (34.5%) with GTDB suffix removal as fallback for 8,692 additional strains (8.9%), for example matching BacDive “Pseudomonas fluorescens” to GTDB “Pseudomonas fluorescens A”. ₇ The central discoveries digest records the same bridge (exact species-name matching plus GTDB suffix removal) as linking 42,227 BacDive strains (43.4% of 97K) to 6,426 GTDB pangenome species, enabling BacDive phenotypes to be cross-referenced with pangenome-derived predictions without Spark queries; it reports that the 56.6% unmatched strains reflect GTDB’s different species boundaries from LPSN/DSMZ taxonomy. ₈

The bridge is lossy: 56.6% of BacDive strains did not match any GTDB species, which the project report attributes primarily to GTDB using different species boundaries than LPSN/DSMZ; genome accession-based matching (GCA→pangenome genome_id) would improve coverage but requires a Spark query. ₇ The loss propagates to small validation sets: the research plan referenced n=31 BacDive heavy-metal isolates, but only 10 matched to pangenome species with metal scores after the species-name bridge (see phenotype-database-coverage-bias). ₇

The pitfalls digest supports this account and adds mechanism. ₁ GTDB, the taxonomy used by the pangenome, renames many species with suffixes such as _A and _B (e.g., Pseudomonas fluorescens A), adds genus-level splits (e.g., Pantoea A), and redefines species boundaries differently from LPSN/DSMZ, the taxonomy reference used by BacDive. ₁ Direct species-name matching captured only 34.5% of strains, and base-name matching with GTDB suffixes stripped added another 8.9%, for 43.4% total. ₁ The digest attributes the remaining 56.6% to genuinely different species concepts or to genera not yet in GTDB r214. This adds GTDB release coverage as a second explanation alongside the project report’s species-boundary explanation. ₁ The recommended fix is two-pass matching (exact species name, then base name with the GTDB suffix stripped). For higher coverage, the digest recommends matching via GCA genome accession → pangenome genome_id, which requires GB_ prefix and version-suffix handling. ₁ The digest also records that only 10 of the 31 BacDive strains tagged cat3=#Heavy metal contamination matched pangenome species with metal tolerance scores. It advises that research plans not assume the full BacDive category size will survive matching, and that post-matching sample sizes and power be computed before results are interpreted. ₁

A second BacDive project reports a lower coverage figure: 37,368 BacDive strains (38.4%) matched to the pangenome with a metal score. ₉ The same report elsewhere states that BacDive species-name matching achieves only 38.4% (5,647/27,702 GTDB species); that percentage does not agree with the species numerator and denominator it is paired with, but matches the report’s strain-level figure, so the species-level match rate in this report remains unresolved. ₉ The report proposes that adding GCA accession matching could recover 10–30% more species; this gain is prospective, because accession matching was not implemented. ₉

Erratum. The report gives 38.4% for two different denominators. 37,368 of 97,334 matched strains is 38.4%; 5,647 of 27,702 GTDB species, the figure this sentence uses, is about one fifth, roughly half what the percentage suggests. ₉

Looser matching rules carry their own risk: the metal_cross_resistance project matched BacDive by genus+species substring search, which is imprecise for organisms identified only to genus level (e.g., “Acidovorax sp.”), and the report says these fuzzy matches should be treated with caution. ₁₀

Bridges across classifiers, ranks and identifier types

Classifier-to-pangenome bridges can achieve high coverage. The nmdc_community_metabolic_ecology project mapped NMDC Centrifuge taxa to GTDB pangenome species and computed per-sample bridge coverage; mean bridge coverage was 94.6%, all 220 samples passed the 30% QC threshold, and 92% of samples mapped ≥85% of community abundance to GTDB pangenome species. ₁₁ Species-level ambiguity remained: for ~1,352 Centrifuge taxa that match multiple GTDB clades (same genus, multiple species), one representative clade was selected by alphabetical tiebreaking on gtdb_species_clade_id, and these genus-proxy-ambiguous taxa account for ~6.5% of mapped abundance. ₁₁ The report describes the sensitivity of community completeness scores to this choice as bounded rather than tested; alphabetical tiebreaking is a convention, not biological evidence. ₁₁ See classifier-database-compatibility-in-taxonomic-quantification.

Genus-level 16S bridges can be far less complete, and the microbeatlas_metal_ecology project shows this in two distinct analyses that should not be conflated. Its global MicrobeAtlas artifact from NB03 is an OTU → GTDB genus mapping of 22,357 rows; no read-coverage figure for that global mapping is given in this evidence. ₁₂ Separately, in its re-analysis of ENIGMA ORFRC groundwater 16S amplicons (PRJNA1084851) classified with SILVA 138, only 16.8% of reads came from OTUs whose genus-level SILVA taxonomy could be matched to GTDB AMR annotations, and the remaining ~83% of reads were excluded from the community-weighted mean (CWM). ₁₂ For that ORFRC analysis, if uncovered genera have systematically different metal-resistance profiles, for example if they are predominantly metal-sensitive, CWM would be biased upward; robustness to coverage-fraction variation has not been formally tested, and the report recommends correlating per-sample CWM with per-sample covered-read fraction before further use. ₁₂ The ORFRC 16.8% figure contrasts with the 94.6% mean coverage of the NMDC Centrifuge bridge and suggests the hypothesis that bridge completeness depends on the source classifier, reference nomenclature and rank rather than being a fixed property of GTDB linkage. _12,11

In the same ENIGMA ORFRC amplicon analysis, Citrobacter and Thermodesulfovibrio were not detected; the report offers either detection limits or mismatched genus names under SILVA nomenclature as explanations, so these are not established biological absences. ₁₂

The plant_microbiome_ecotypes project reports two further bridges. Its NMDC taxonomy bridge matched 80.7% of genera, and the sensitivity of results to the unmatched genera was deferred to future work. ₅ Its MGnify bridge linked 20,473 MGnify species across 4 biomes to 1,185 pangenome genera (9,603 MGnify species in matched genera). By biome, this comprised soil 19,472 species (4,229 genera), tomato rhizosphere 579 species (309 genera), maize rhizosphere 336 species (212 genera) and barley rhizosphere 86 species (52 genera). ₅ These figures are counted in different units (genera or species) from the 94.6% abundance-based and 16.8% read-based bridges above, so they refine rather than directly rank against those results. They are consistent with the hypothesis that bridge completeness depends on the source resource and rank. _5,11,12

Rank mismatches recur across external resources. NMDC covstats and Planet Microbe taxonomy.name are both species-level, so any genus-level abundance claim requires species→genus aggregation first. ₁₃ In the other direction, the enigma_contamination_functional_potential genus-to-clade mapping showed a long right tail, with a maximum of 433 clades per genus, so a single genus label can fan out to many GTDB clades. ₁₄ These refine the rank-checking requirement stated above for pangenome tables by showing it applies to abundance resources outside the pangenome as well. _13,14

Some bridges stop at genus level. ₁₅ In webofmicrobes_explorer, web-of-microbes (WoM) ↔ pangenome linkage is available at genus level for all WoM organisms. This gives pangenome context (gene clusters, conservation, functional annotations) for every WoM organism genus, but species-level matching requires strain-to-genome mapping, which the project did not attempt. ₁₅ The project’s kescience-fitnessbrowser matches mix resolutions; each pair of figures below gives the table’s gene and experiment counts. Pseudomonas sp. FW300-N2E3 (pseudo3_N2E3, direct strain; 5,854 and 211) and Pseudomonas sp. GW456-L13 (pseudo13_GW456_L13, direct strain; 5,243 and 106) match directly at strain level. E. coli BW25113 matches Keio as the same strain (4,610 and 168). Synechococcus PCC7002 matches SynE (PCC 7942) at genus level only (2,722 and 129). ₁₅ This refines the rank-checking rule: each linkage should carry its resolution (direct strain, same strain, genus only), and a genus-only match should not be read as strain-level evidence. ₁₅

Identifier columns must be type-checked, not only named correctly. An attempt to link Fitness Browser organisms to GapMind species clades via NCBI taxonomy IDs from kbase_ke_pangenome.gtdb_metadata returned zero matches because the relevant column contained boolean strings rather than numeric taxids; downstream analyses therefore used organism-level fitness aggregates without explicit clade-level linkage, which may reduce the precision of the H2a conservation comparison. ₁₆ This refines the taxid-validation step of the workflow below: an NCBI taxid is a stable identifier only if the column actually holds taxids. ₁₆ The pitfalls digest supports this diagnosis. It records that joining kescience.fitnessbrowser.organism to kbase.ke_pangenome.gtdb_metadata via NCBI taxonomy IDs returns zero matches because the taxid column holds the string values “t” / “f” rather than numeric IDs. ₁ The digest separately warns that NCBI taxid linking to the Fitness Browser resolves at species level rather than strain level, which creates many-to-one ambiguity even when the taxids are valid. ₁

Some cross-taxonomy disagreements remain open questions rather than measured results. The berdl_data_atlas project proposes a use case (UC3) joining kbase ↔ refdata through 11 shared keys to ask where GTDB clades and KBase species pangenomes disagree and what that implies for gene flow; no answer is reported in this evidence (see cross-tenant-data-bridging). ₁₇

Reconciliation workflow

1. Discover the live schema, catalog address, taxonomy version, and identifier columns before reusing historical queries. ₁
2. Prefer stable accessions such as GCF_* over short strain labels, and retain the original identifier, source database, taxonomy version, and reconciliation rule for every match. ₁
3. Build a synonymy layer that records renamed taxa and lineage aliases rather than applying unconstrained string normalization. ₁
4. Validate candidate matches with genus consistency and, where available, NCBI taxids, assembly accessions, or other stable identifiers. ₁
5. Confirm the taxonomic rank represented by each column and use the database-specified join key, such as genome_id for the cited pangenome taxonomy tables. ₁
6. Keep species-specific gene-cluster identifiers within species and translate to shared orthology or domain representations before cross-species analysis. ₁
7. Audit unmatched, multiply matched, and biologically implausible records separately instead of treating them as biological absence. ₁
8. Normalize genome-accession prefixes (RS_/GB_) and clade-ID formats before joining, and treat any zero-row or all-NaN join as a plumbing failure to diagnose before reading it as a null result. _1,5

Tensions

A short-name workflow is convenient for exploratory analysis, but the documented MT20 collision and the 12 incorrect genus matches among 32 pangenome linkages show that convenience can contradict identifier validity. ₁ A hand-curated 23-entry synonym map can resolve a targeted battery, whereas full-taxonomy aggregation requires a broader NCBI-taxid-backed, GTDB-version-aware layer; the appropriate reconciliation scope therefore depends on the analysis scope. ₁

Database names and schemas can change over time, so archived reports and notebooks should not be treated as automatically valid executable instructions without live inspection. ₁ This supports provenance-aware-resource-discovery and makes taxonomy-version recording part of reproducibility rather than an optional annotation. ₁

The two BacDive bridges report different strain-to-pangenome coverage: 42,227 strains (43.4%) via exact matching plus GTDB suffix removal in one project ₇, versus 37,368 strains (38.4%) matched to the pangenome with a metal score in another ₉. The assigned evidence does not state whether the second bridge used the suffix-removal fallback, so the gap cannot be attributed to method from these records, and the second report’s own species-level percentage is internally inconsistent. ₉

GTDB suffix removal increases matches by treating BacDive “Pseudomonas fluorescens” and GTDB “Pseudomonas fluorescens A” as the same species ₇, whereas the cf_formulation_design report warns that GTDB-specific names such as Pseudomonas_E denote clades distinct from the similarly named NCBI taxon ₄. Suffix stripping may therefore merge GTDB-separated clades; how many of the 8,692 suffix-fallback matches do so has not been measured. ₇

Open Directions

• Combine ncbi_strain_identifiers, assembly accessions, NCBI taxids, and GTDB taxonomy versions, then measure how many historical pangenome links change genus or species assignment under a validated reconciliation graph. ₁
• Apply the synonym layer to all MetaPhlAn3 cohort taxonomies and compare abundance, prevalence, and differential-effect estimates before and after reconciliation to determine which findings depend on naming mismatches. ₁
• Test genome_id, gtdb_taxonomy_id, and accession-based joins against manually reviewed genus-consistent matches, using rank-specific precision and recall to identify safe join keys for each pangenome table. ₁
• Quantify cross-species agreement when gene-cluster identifiers are replaced by COG categories, KEGG orthologs, or Pfam domains, and determine which representation best preserves the biological question without species-specific identifier leakage. ₁
• Re-run the BacDive bridge with GCA accession → pangenome genome_id matching and compare it with the 42,227-strain species-name bridge, separately checking the 8,692 suffix-fallback matches for merged GTDB clades. ₇
• Recompute both BacDive bridges from one BacDive snapshot with a shared matching rule to reconcile the 43.4% and 38.4% coverage figures and restate the species-level denominator. _7,9
• In the ENIGMA ORFRC amplicon analysis (PRJNA1084851), correlate per-sample CWM with per-sample covered-read fraction to test whether reads excluded by the SILVA-to-GTDB genus bridge bias metal-resistance estimates; report the covered-read fraction of the global 22,357-row MicrobeAtlas OTU → GTDB genus mapping separately. ₁₂
• Replace alphabetical tiebreaking for the ~1,352 multi-clade Centrifuge taxa with all-clade averaging and measure the change in community completeness scores. ₁₁
• Repair the taxid column used for Fitness Browser-to-GapMind clade linkage and rerun the H2a conservation comparison with explicit clade-level linkage. ₁₆
• Execute UC3 over the kbase ↔ refdata keys to count GTDB clade versus KBase species-pangenome disagreements. ₁₇
• Backfill ncbi_taxid for legacy Bacteroides_X rows in ref_taxonomy_crosswalk and merge the separate B. vulgatus / P. vulgatus canonical rows, then re-run the cross-cohort pivot without the hand-curated map to test whether the log₂FC ≈ 28 artifacts disappear. ₁
• Resolve the 12 orphan pangenome records against current GTDB species clades to decide whether they need remapping or exclusion. ₁
• Run the deferred sensitivity analysis for the 80.7% NMDC genus match in plant_microbiome_ecotypes by recomputing downstream results under alternative assignments of the unmatched genera. ₅
• Re-run the H7 subclade scan for the 47 of 65 candidate species absent from phylogenetic_tree_distance_pairs once tree coverage improves (or with an alternative per-species phylogeny), to test whether the 5/17 result generalizes. ₅
• Build the strain-to-genome mapping for Web of Microbes organisms that webofmicrobes_explorer did not attempt, enabling species-level genome-content ↔ metabolite-output tests. ₁₅