Predicted Effector Gene Aggregation, Standards and Unified Schema (PEGASUS): A Community Framework for Effector Gene Reporting

Key results

1. ShownPEGASUS splits PEG reporting into three linked products per GWAS trait: a metadata record, an evidence matrix covering every gene and evidence item considered at each locus, and a PEG list of author-prioritised genes linked back to that evidence.

   Each analysed GWAS dataset yields one metadata record, one evidence matrix and a PEG list. Matrix rows are variant-gene pairs at a locus; columns hold evidence types, provenance fields or optional integration scores. The list is pulled from the matrix via an author_conclusion column. The split mirrors full GWAS summary statistics versus lead associations. · Introduction; Overview of the PEGASUS framework; Evidence matrix; PEG list

2. ShownA controlled vocabulary sorts evidence into variant-centric and gene-centric categories. Finer-grained streams sit under each category, so new evidence types can be added without redesigning the schema.

   Variant-centric examples: statistical fine-mapping, credible set membership, chromatin interactions, QTLs, clinically relevant variants. Gene-centric examples: expression, perturbation experiments, gene set associations, gene burden analyses, literature. Category vs stream example: QTL analysis containing eQTL and sQTL streams (Figure 4, Table S1). · Evidence matrix; Organisation of evidence categories

3. ShownA machine-readable metadata standard records the context needed to interpret a PEG output. It has a small mandatory core plus recommended fields that link to external ontologies and resources.

   Fields cover trait ontology IDs plus a readable description; GWAS accession, sample characteristics, genome build and publications; locus definition (method, window size or fine-mapping) and gene-to-locus rules; evidence sources, categories and biosamples; any integration method; column definitions. genome_build is mandatory, variant_type recommended; files in YAML or JSON. · Metadata standard

4. SuggestedIn working-group implementation tests, PEGASUS reportedly fit every tested PEG dataset without changes to the original analyses. Testers judged the conversion effort similar to submitting to established genomic repositories.

   Members reformatted previously published PEG outputs and their own analyses, spanning different evidence types, traits and integration strategies. Their own outputs converted easily; legacy outputs needed extra curation where provenance or context was missing. No dataset counts or timings are reported. · Design principles, usability and implementation…; Implementation outcomes and experience

5. ShownTwo worked examples show the format end to end: a synthetic dataset converted into metadata, matrix and list, and one published locus from Aragam et al. (2022) re-expressed in PEGASUS without re-running its analysis.

   Synthetic data (Supplemental File 2 and online documentation) is traced from raw input to a metadata entity-relationship model for QTL evidence (method e.g. FastQTL, source e.g. GTEx) to the evidence matrix (Figure 3) and PEG list (Figure 2B). The Aragam et al. (2022) locus is in Supplemental File 3; a metadata template is Supplemental File 4. · Worked example with synthetic and real data; Data sharing; Description of Supplemental files; Figure legends