MetabDB Guide
Entities, evidence, fields and website use.
1. Entities in MetabDB
MetabDB organizes source data into five main entity groups.
| Entity | Meaning | Example |
|---|---|---|
| Variants / genomic contexts | Variants, representative variants, genomic position groups and genetic evidence sets. | rs560887 |
| Genes | Gene identities linked to genetic or gene-level evidence. | GCK |
| Protein targets | Protein targets represented in molecular QTL evidence. | APOA5 protein |
| Metabolic measures | Source-defined compounds, lipids, lipoprotein measures, ratios and derived measurements. | Glucose, GlycA |
| Diseases / GWAS traits | Diseases, clinical outcomes and other genome-wide association study (GWAS) traits. | Type 2 diabetes, BMI |
These entity groups provide the common structure used for search and navigation, while original source labels and definitions remain available in the underlying records.
2. How entities are connected
Stored links describe how records are connected. Evidence layers describe why a connection is supported.
The dashed protein connection represents supporting protein-QTL (pQTL) evidence.
| No. | Stored link | Meaning |
|---|---|---|
| ① | Variant–measure | A source variant linked to a metabolic measure. |
| ② | Representative variant–measure | A recorded representative variant linked to a measure; not necessarily a GWAS lead. |
| ③ | Position group–measure | A coordinate-based genomic group linked to a measure. |
| ④ | Evidence set–measure | A recorded genetic evidence set linked to a measure. |
| ⑤ | Gene–measure | A gene linked to a metabolic measure. |
| ⑥ | Disease / trait–measure | A disease or trait linked to a metabolic measure. |
The six stored links are database structures, not six biological mechanisms.
3. Association evidence
Association evidence describes observed relationships between a metabolic measure and a clinical outcome or trait.
| Endpoint context | Typical effect | Meaning |
|---|---|---|
| Incident | HR; sometimes RR or OR | New-onset disease. |
| Prevalent | OR; other effects as reported | Existing disease. |
| Mortality | HR or RR | Death or cause-specific mortality. |
| Other / source-specific | HR, OR, RR, beta or P-only | Other source-defined clinical or analytical contexts. |
HR, OR, RR, beta and P-only results remain separate. Estimates retain their model, endpoint and exposure scale and are not automatically pooled.
A disease page can contain several source endpoint definitions. Incident, prevalent and mortality contexts are not assigned to purely genetic evidence.
4. Genetic evidence
| Evidence layer | What it shows | Main entities |
|---|---|---|
| GWAS / metQTL | Variant association with a metabolic measure; metQTL means metabolic quantitative trait locus. | Variant ↔ measure |
| Fine-mapping | Posterior inclusion probability (PIP) for variants within a source-defined signal; displayed records use PIP ≥ 0.8 where documented GRCh38 coordinates are available. | Variant / evidence set ↔ measure |
| Rare variant / burden | Gene-level association from rare variants or aggregate tests. | Gene ↔ measure |
| Colocalisation | Support for a shared genetic signal; curated PP.H4 (posterior probability of hypothesis H4) records are displayed at ≥ 0.8. | Measure ↔ disease / trait / molecular QTL target |
| MR / cis-MR | Mendelian randomisation (MR) results using source methods and instruments; cis-MR is labelled separately. | Measure ↔ disease / trait |
Metabolic measure–disease relationships
| Evidence class | Evidence type | Endpoint / direction |
|---|---|---|
| Association | Direct association | Incident / prevalent / mortality / other |
| Genetic | Colocalisation | No inferred direction |
| Genetic | Forward MR | Metabolic measure → disease |
| Genetic | Reverse MR | Disease → metabolic measure |
| Genetic | cis-MR | Source-defined direction |
Forward, reverse and cis-MR remain separate. MR-A/B/C describe linked support: A, sensitivity analyses and colocalisation; B, one of these; C, neither. Analyses without linked sensitivity or colocalisation support are retained as reported and should not be interpreted as failed analyses.
5. Key fields
| Field | Meaning |
|---|---|
| HR | Hazard ratio. |
| OR | Odds ratio. |
| RR | Risk or rate ratio, as reported by the source. |
| Beta | Source-reported regression coefficient. |
| SE | Standard error. |
| P | Source-reported P value. |
| FDR / q | False discovery rate adjusted value, reported by the source or derived from a documented complete test family. |
| PP.H4 | Posterior probability of a shared genetic signal under the source colocalisation model. |
| PIP | Posterior inclusion probability for a variant under the source fine-mapping model; credible-set membership is a separate field. |
| CLPP | Colocalisation posterior probability; in eCAVIAR, support for a variant being causal in both traits, distinct from PP.H4. |
| Variant / SNP | Variant is the general term; SNP means single-nucleotide polymorphism when that variant class is established. |
| Lead variant | A representative variant designated as a lead by the source analysis. |
| Genomic position / locus | A position is a chromosome/base-pair coordinate; a locus is a source-defined genomic region with its own boundaries and model context. |
| N | Sample size for the source analysis. |
| Events | Number of outcome events, when reported. |
| Studies | Number of contributing MetabDB source studies. |
| Source records | Retained source-level analytical rows, within the labelled scope. |
| Position groups | Coordinate-based genomic groups in a stated genome build. |
| Direction | Positive or negative source effect when the exposure orientation is known; the null is 1 for ratios and 0 for beta. |
FDR and adjusted P
Source-reported FDR and BH-FDR derived from complete original P-value families are labelled separately. Displays use FDR when available, otherwise adjusted P with its correction method. A dash means neither value is available. Bonferroni-adjusted P is not FDR.
FDR-supported direction counts require q ≤ 0.05 and known exposure orientation. Nominal P alone does not establish FDR support; corrections remain specific to each evidence layer.
Statistical display
HR, OR, RR and their confidence intervals use two decimal places; beta and SE use three significant digits. PP.H4 and PIP use three decimals. P, FDR and q use three significant digits, with scientific notation below 0.001. Unrecoverable source zeros display “Below reporting limit”; a dash indicates missing information. Downloads retain source precision. Adjustment badges distinguish source-reported values, MetabDB BH and other documented methods.
Exposure scale
Measurement unit and exposure scale are separate. Equal “per SD” labels do not guarantee the same concentration increment across studies. Numerical ranges describe source values; compare effect magnitudes only within compatible models and exposure scales.
6. What counts mean
| Count | Meaning |
|---|---|
| Entity | One standardized entity in the named public registry; a metabolic measurement is not necessarily a unique chemical compound. |
| Source record / source row | One retained source-reported analytical result after within-source deduplication. |
| Canonical pair | One deduplicated entity pair for a specified stored-link type. |
| Support block | A source/study/result-type summary supporting one canonical pair; several source rows may contribute. |
| Evidence card | An interface summary of a scoped entity pair or genomic context. |
| Endpoint context | A source-defined clinical outcome or analytical context; several definitions can map to one disease. |
| Study | One MetabDB source study; studies may share cohorts or upstream data. |
These counts use different denominators and should not be added together unless explicitly stated. Evidence layers can overlap.
7. Using MetabDB
| Page | Purpose |
|---|---|
| Search | Find an entity and its linked evidence. |
| Browse | Explore standardized entities by category. |
| Colocalisation | Explore curated shared-signal evidence. |
| Features | View disease-association profiles, case studies and gene evidence breadth. |
| Studies | Review source cohorts, assays, coverage and overlap. |
| Summary | View overall entity, relationship and study-coverage statistics. |
| Download & API | Access source-level and standardized data. |
| Help | Read definitions, fields and interpretation guidance. |
Search
| Search type | Example |
|---|---|
| Metabolic measure | Glucose |
| Disease / trait | Type 2 diabetes |
| Gene | GCK |
| Variant | rs560887 |
Use the search-type selector when a name is ambiguous.
Browse
- Metabolic measures — assay, measurement type and biochemical class.
- Diseases / traits — diseases/outcomes and other GWAS traits are separate.
- Molecular QTL targets — gene/protein targets by QTL type.
Source collections are separate from entity counts. Assay filters can overlap; NMR (nuclear magnetic resonance) and MS (mass spectrometry) are methods, not biochemical classes.
Colocalisation
- Metabolic measure → diseases, traits or molecular QTL targets.
- Disease / trait → metabolic measures.
- Molecular QTL target → metabolic measures.
Partner entities, studies, position groups and source records are counted as labelled.
Features
- Disease-association profiles — clustering, PCA, stability and sensitivity analyses.
- Case studies — selected source-level disease and genetic evidence.
- Gene evidence breadth — linked measures and supporting evidence layers across genes.
Feature analyses are descriptive and exploratory; evidence layers are not combined into a single score.
Studies
- Cohort and ancestry
- Assay and biospecimen
- Measurement coverage
- Evidence contribution
Measurement and evidence overlap are reported separately from participant overlap.
8. Source definitions and mapping
Standardisation is applied to entity identities so that records can be searched and compared across studies. Original source labels, definitions and analytical results remain accessible.
How identities are standardised
- Start from the source record. Use its source key, study-specific code or complete original label, retaining the publication, supplementary-table or assay-documentation locator.
- Match to reference identifiers. Validate external identifiers against versioned reference dictionaries. New or revised mappings require a complete-label or exact-synonym match, or an explicit official equivalence with source support. Similar names and ordinary cross-references alone do not establish identity.
- Confirm entity type and scope. Keep diseases, other traits, genes, proteins and metabolic measurements distinct. Clinical endpoint contexts are stored separately; ambiguous candidates and conflicting definitions remain pending.
- Record the final mapping. Accepted existing mappings and audited revisions are recorded in the versioned entity registry, with the source identity and mapping provenance preserved.
| Public state | Meaning |
|---|---|
| Mapped | A reliable external identity mapping has been accepted. |
| Source-defined validated | Identity and entity type are supported by the original publication, supplementary material or assay documentation, without a safe external canonical mapping. Some views shorten this label to “Source-defined”. |
| Unresolved / pending | Identity or type is insufficiently resolved for the public catalogue and entity counts; original records remain in provenance. |
Mapped and Source-defined validated entities are both displayed and counted. Mapping status and review status are separate. Standardisation does not make source case definitions, cohorts, assays or exposure scales equivalent.
Release 1 mapping coverage
These counts reflect the current public views, with splicing QTL evidence excluded from results and statistics. Original source records are retained.
| Public entity group | Mapped | Source-defined validated | Total |
|---|---|---|---|
| Diseases / outcomes | 771 | 2,753 | 3,524 |
| Other GWAS traits | 92 | 3,290 | 3,382 |
| Molecular QTL targets | 1,798 | 227 | 2,025 |
| Metabolic measures | 0 | 2,663 | 2,663 |
The 863 mapped disease/trait entities use MONDO, EFO or MeSH; the remaining 6,043 retain validated source identities. Mapped molecular QTL targets include 110 UniProt protein targets.
All 2,663 metabolic measures retain source-defined measurement identities. Accepted ChEBI/RefMet concept annotations do not by themselves establish a unique chemical structure, resolve isomers or justify merging measurements across platforms.
Gene and variant reference identifiers
In the broader gene-evidence collection, 6,536 gene records are linked to reference identifiers: 6,535 to HGNC and one to NCBI Gene. All 110 protein targets map to UniProt. These protein targets are already included in the molecular QTL counts above.
Of 2,553,037 variant source contexts, 2,073,433 (81.21%) have accepted rsID mappings after coordinate and allele checks against dbSNP build 154 / GRCh38. These are source contexts, not distinct variants. The unordered allele check does not establish effect-allele orientation.
Entity mapping preserves source estimates, P/FDR values, model descriptions, measurement units and exposure scales. Gene records and variant source contexts have separate denominators and are not added to public catalogue entity totals.
9. Interpretation notes
- Association is not causality.
- Colocalisation indicates a shared genetic signal, not mediation or causality.
- MR remains source- and method-specific; support classes do not grade causal certainty.
- Fine-mapping probabilities depend on the source model and priors.
- Studies may share cohorts, assays or upstream GWAS; shared UK Biobank samples are not independent replication.
- Equal effect labels do not guarantee comparable exposure scales.
- Coordinate groups are not assumed to be independent loci or allele-resolved variants.