PEG Evidence Categories

🌟 Why do we need PEG evidence categories?

Predicted Effector Gene (PEG) lists are built from diverse sources of evidence, ranging from genetic associations and computational predictions to functional assays and literature curation. Each source has its own origin, methodology, and level of confidence.

To ensure clarity and interoperability, PEGASUS group these diverse evidence types into standardised categories that are widely recognised in effector gene exploration.

➡️ Organising evidence into categories helps users:

• Interpret consistently – understand the general type of support behind each gene.
• Compare across studies – align evidence types, even when detailed sources differ.
• Build trust and transparency – see at a glance how predictions were made.

---

PEG Evidence Categories

🔀 Variant-centric 🔀

Evidence categories	Abbreviation	Explanation
Linkage disequilibrium	LD	Assessment of whether variant is correlated with another variant of interest and may act as a proxy.
Finemapping and credible sets	FM	Finemapping results - probability of variant being causal within a credible set, using Bayesian or probabilistic models
Colocalisation	COLOC	Variant affects two traits (typically a complex trait and a molecular phenotype) at the same locus.
Molecular QTL	QTL	Variant affects a molecular phenotype, e.g. gene expression (eQTL), splicing (sQTL), or protein expression (pQTL).
Mendelian Randomization (MR)	MR	Uses genetic variants as proxies for exposures to test their causal effect on outcomes.
Regulatory region	REG	Variant lies in open chromatin or enhancer/promoter elements in relevant tissue (e.g. ATAC-seq, DNase-seq, or histone mark data)
Chromatin interaction	CHROMATIN	Variant lies in a region physically interacting with a gene promoter via 3D chromatin architecture (e.g. Hi-C, Capture-C data).
Predicted functional impact	FUNC	Variant predicted to disrupt gene/protein function or regulatory motifs, e.g. via SIFT, PolyPhen, CADD.
Proximity to gene (distance)	PROX	Assessment of whether variant is within or near gene boundaries.
Genome-wide association (GWAS) signal	GWAS	P-value from source GWAS for association of variant with trait specified in metadata file
PheWAS (Phenome-Wide Association Study)	PHEWAS	Variant is associated with multiple traits, suggesting pleiotropic effects

🧬 Gene-centric 🧬

Evidence categories	Abbreviation	Explanation
Protein–protein interaction	PPI	Gene’s protein interacts with other disease-relevant proteins.
Pathway or gene sets	SET	Gene is part of a known pathway or complex relevant to the phenotype, e.g. results of enrichment analyses using Reactome or KEGG.
Gene-based association	GENEBASE	Aggregated analysis of association of variants in gene with trait (e.g. SKAT, MAGMA, burden tests) .
Expression	EXP	Gene is differentially expressed in relevant tissue or disease e.g. the gene is more highly expressed in phenotype-related tissues compared to others.
Perturbation	PERTURB	Gene perturbation causes phenotype-relevant effects in lab or model organisms (knock out animal/cell line, human organoid).
Biological Knowledge Inference	KNOW	Gene–phenotype relationships can be inferred based on known biology, without providing specific references or direct experimental evidence linking the specific gene to the phenotype.
Genetically predicted trait association (TWAS/PWAS)	TPWAS	Evidence from transcriptome- or proteome-wide association studies showing that gene’s genetically predicted expression or protein level is associated with phenotype,
Drug related	DRUG	Evidence from drug mechanism of action, e.g. gene encodes a known drug target or interacts with targets of drugs used to treat the phenotype, supporting therapeutic relevance.

Variant or Gene centric evidence

Evidence categories	Abbreviation	Explanation
Cross-phenotype	CROSSP	Gene or variant already established in a related phenotype (biologically similar).
Literature curation	LIT	Human-curated gene or variant–disease links from literature.
Association from curated database	DB	Variant or Gene is curated as causal or related to the phenotype from existing database, like ClinVar, ClinGen, OMIM, etc.