1 of 14

Moving from guidelines to standards for PEG data

Laura Harris

GWAS Catalog

2 of 14

Major evidence types – variant & gene centric

(Costanzo et al, 2025, in press)

Gene-Centric

Biological features

Top-down

Similarity based

Variant-Centric

Genomic Features

Bottom-up

Locus-based

3 of 14

PEG evidence matrix vs PEG list

Categorisation

Strong evidence

Moderate

Weak evidence

​

Effector Gene

Gene 2

Evidence matrices. PEG lists

Gold standard list

Effector Gene

Gene 2

Effector Gene

Gene 2

4 of 14

What needs to be defined for a variant-centric evidence matrix?

  • Data content

​

​

​

  • Metadata content

​

​

​

  • Data structure e.g standard headers & layout

​

​

​

  • File format

​

​

txt

csv

xlsx

meta

data

meta + data

+

*Sarsani et al 2023

What content should be mandatory?

5 of 14

High level�requirements for �Metadata

Must include:

  • the GWAS the list was derived from
  • standard terminology for evidence types
  • criteria for significance for each evidence type
  • whether the list contains de novo wet-lab evidence
  • details of method for prioritization

Individual publications & pipeline-generated data should be interoperable

​

​

6 of 14

Metadata content

    • GWAS data

​

​

​

​

​

    • Method

​

​

​

    • Evidence

​

    • Trait description
    • Ontology mapping
    • GWAS source (accession ID) OR GWAS sample metadata
    • Genome build

​

    • Free text description of gene prediction methods

​

    • Annotation
    • eQTL
    • Tool

​

7 of 14

Metadata content

  • Evidence
    • Annotation: name
    • eQTL: source
    • Tool: name
    • Fine_mapping: LDref, software, CI
    • 3DGenomics: type, source, threshold
    • Colocalisation: trait 1, trait 2
    • Expression: source, cutoff
    • Related phenotype: GWAS evidence source
    • Animal phenotype: source
    • Drug database: name
    • TWAS: software, source
    • Other: description

​

Should a minimum number of evidences be mandatory?

Should any specific evidence type be mandatory?

Can be used to indicate de novo evidence

Evidence types could be defined via an ontology

8 of 14

Data standards

  • Single table containing all evidence types
  • Plain text file not including graphics
  • Present evidence for all genes considered for each locus
  • Use standard identifiers for genes
  • Include sentinel variant
  • Prioritisation score

High level requirements for

Data

9 of 14

  • Include all sentinel variants (all tested or all in the GWAS?)
    • Variant identifier: chr:bp:REF:ALT
    • GWAS p-value

​

  • All genes tested for each variant
    • Ensembl ID and gene symbol
    • “Nearest gene” annotation Y/N

​

  • All evidence for each gene-variant pair
    • Standard column headers
    • Refer to metadata

​

  • Score
    • Author’s own
    • Universal score?

Data content

10 of 14

  • 1 row per variant-gene pair

​

  • Standard formats for evidence column headers
    • To follow pattern “evidencetype_detail”?
      • e.g Expression_adipose, Finemapping_PIP

​

  • Validation criteria for each evidence type? Or allow flexibility?

​

​

  • Authors own score in any format

​

​

Data structure

11 of 14

​

​

File format

​

​

  • Plain text file

​

​

  • Metadata in same file or linked elsewhere?

​

​

  • List (top genes) presented separately from evidence matrix

​

​

12 of 14

PEG evidence matrix vs PEG list

Categorisation

Strong evidence

Moderate

Weak evidence

​

Effector Gene

Gene 2

Evidence matrices. PEG lists

Gold standard list

Effector Gene

Gene 2

Effector Gene

Gene 2

13 of 14

What needs to be defined for a PEG list?

Summary of results including only the top gene for each variant

​

  • Top gene for each variant – with ticklist of evidence – plus categorization or score

​

​

​

​

  • Top gene for each variant – +/- – plus cat/score

​

​

​

​

  • Only include genes passing a threshold – not necessarily present every gene if insufficient evidence

​

14 of 14