💡 Illustrative Examples of Evidence Matrix Columns

The tables below show examples of how PEG evidence matrix columns can be named and formatted.
These examples are not mandatory fields — they are provided to demonstrate recommended naming patterns, data formats, and reporting styles.

• Projects may define additional or alternative columns, we recommend following these general conventions.
• Metadata should provide comprehensive information to understand the data type, provenance, and scale used for each column.

Variant-centric evidence example

Evidence Category	Column header	Data Format	Description	Requirement	Example data
GWAS	GWAS_pvalue	Exponent or −log10	P-value of the primary variant in the source GWAS. Specify whether exponent (e.g. 4×10⁻⁹) or −log10 scale in the metadata file.	optional	4×10⁻⁹
Proximity	PROX_nearest_gene	boolean	Indicates whether the variant is the nearest gene. Details on how distance is derived (e.g. to TSS, to gene footprint) should be documented in the metadata.	optional	N
QTL	QTL_eQTL_pancreas_pvalue	exponent or -log10	Significance value for eQTL association in pancreas tissue.	optional	0.01
QTL	QTL_eQTL_pancreas_CI	range	Confidence interval for the eQTL effect. Define confidence level (e.g. 95%) in metadata.	optional	[1.2, 2.5]
Functional	FUNC_CADD	float	CADD functional prediction score. Specify genome build and release in the metadata.	optional	15.62
Fine-mapping	FM_credible_set_ID	string	Identifier of the credible set variant from fine-mapping.	optional	chr10:114754071:T:C
Fine-mapping	FM_PIP	float	Posterior inclusion probability (PIP) from fine-mapping.	optional	0.98
Coloc	COLOC_PPH4	float	Colocalisation posterior probability that both traits share a causal variant (PPH4).	optional	0.85

Gene-centric evidence example

Evidence Category	Column header	Data Format	Description	Requirement	Example data
TWAS	TPWAS_TWAS_pvalue	float	Transcriptome-wide association study (TWAS) p-value linking gene expression to trait.	optional	1×10⁻⁷
Expression	EXP_Adipose_TPM	float	Expression level of the gene in adipose tissue, reported as Reads Per Million per Kilobase (RPMK) or Transcripts Per Million (TPM).	optional	0.8
Expression	EXP_pancreas_TPM	float	Expression level of the gene in pancreas tissue, reported as RPMK or TPM.	optional	—
Perturbation	PERTURB_mouse	Free text / ontology terms	Observed phenotype in mouse perturbation models (e.g., knockout, overexpression). Terms can be free text or ontology labels, defined in metadata.	optional	hypoglycemia | increased insulin secretion | impaired glucose tolerance
Knowledge	KNOW	Narrative text	Expert or knowledge-base curation describing gene function and its relationship to phenotype or disease.	optional	ANGPTL4 inhibits lipoprotein lipase (LPL), increasing circulating triglycerides and reducing fatty acid uptake.
Literature	LIT	Narrative text	Human-curated evidence from published studies linking the gene to relevant traits or disease mechanisms.	optional	Zebrafish Tcf7l2 mutant shows hyperglycemia, pancreatic and vascular defects, reduced regeneration.
Literature	LIT_PMID	PMID list	PubMed identifiers (PMIDs) supporting literature evidence for the gene–trait association.	optional	PMID_28851992 | PMID_31829936
Drug	DRUG	Drug name(s)	Drug(s) known to target or modulate the gene, separated by | (pipe). Reference databases (e.g., DrugBank) can be cited in metadata.	optional	METFORMIN | CYCLOSPORINE

Integration example

Column header	Format	Description	Requirement	Example data
INT_pops	float	Population count or weighted metric used in integration scoring. Define precise meaning and provenance in the metadata file.	optional	9
INT_Combined_prediction_author_score	any	Author-provided integrated prediction score. Units, scale, or categories (e.g. STRONG, WEAK) must be described in the metadata file.	optional	STRONG