Bio Pathway Wikipathways

Version Compatibility

Reference examples tested with: clusterProfiler 4.18+, rWikiPathways 1.26+, org.Hs.eg.db 3.18+.

Before using code patterns, verify installed versions match. If versions differ:

• R: packageVersion('<pkg>') then ?function_name to verify parameters

If code throws ImportError, AttributeError, or TypeError, introspect the installed package and adapt the example to match the actual API rather than retrying.

WikiPathways is a LIVE, monthly-updated database. enrichWP, gseWP, and gson_WP all download data.wikipathways.org/current/gmt/ at run time, so the SAME code returns DIFFERENT pathways and p-values months apart with no error. current/ is not a version. For a reproducible analysis pin a dated release with downloadPathwayArchive(date='YYYYMMDD', organism=, format='gmt') and report the date. All enrichWP/gseWP/downloadPathwayArchive calls require internet at run time.

WikiPathways Enrichment

"Which community-curated WikiPathways are enriched in my genes?" -> Test WikiPathways gene sets against a gene list (ORA) or a ranked vector (GSEA), pinning a dated GMT for reproducibility - because the live monthly database changes under identical code, and the WP GMT is Entrez-keyed so any other ID type silently overlaps nothing.

• R (ORA): enrichWP(entrez, organism='Homo sapiens', universe=all_entrez)
• R (GSEA): gseWP(named_decreasing_entrez_vector, organism='Homo sapiens')
• R (reproducible): downloadPathwayArchive(date='YYYYMMDD', organism=, format='gmt') -> read.gmt -> split term -> enricher/GSEA

Scope: WikiPathways-specific enrichment - the data model, the current/-vs-dated GMT reproducibility pin, the Entrez-GMT requirement, the term-field split, PFOCR as a noisier complement, and the WP-vs-KEGG-vs-Reactome contrast. The ORA/GSEA method choice and hypergeometric/background theory -> the category README. The DE list and ranking statistic -> differential-expression/de-results. KEGG and Reactome -> kegg-pathways, reactome-pathways. Plot grammar -> enrichment-visualization.

The Single Most Important Modern Insight -- A WikiPathways Result Is a Snapshot of a Live, Community-Edited Database Taken on the Run Date

WikiPathways is a wiki: anyone can create or edit a pathway, content is CC0, and there is NO formal journal-style peer review gating a pathway's publication (Pico 2008 PLoS Biol 6:e184; Martens 2021 NAR 49:D613). The collection is republished as a dated GMT archive every MONTH. Three properties every misuse forgets:

1. current/ is not a version. enrichWP, gseWP, and gson_WP all silently download data.wikipathways.org/current/gmt/ - the latest monthly release. Identical code two months apart returns different pathways and different p-values, with no error and no warning. Reproducibility is NOT a code freeze; it is a dated GMT: downloadPathwayArchive(date='20240310', organism='Homo sapiens', format='gmt'), read it, run enricher/GSEA on the pinned sets, and report the date in methods. gson_WP() freezes only within a session (it snapshots current/), not across time.

2. The WP GMT speaks Entrez, and the wrong ID type fails silently. The GMT is Entrez-keyed via BridgeDb. Passing SYMBOL or ENSEMBL yields near-zero overlap and an empty or misleading result with NO error - convert to Entrez upstream (bitr/OrgDb) before enrichWP. Likewise universe=NULL makes the background "all genes that happen to be in WP" - a small, biased set that inflates significance; pass the assayed/tested Entrez vector as universe.

3. A community pathway is a hypothesis someone drew, not a reviewed fact. The two things that make WP valuable (open CC0 license, anyone-can-edit curation that captures disease/drug pathways KEGG and Reactome lack, e.g. the COVID-19 Disease Map) are the same two things that make its quality heterogeneous. Many WP pathways are also imported from KEGG/Reactome, so "three databases agree" can be circular rather than independent. Treat each hit as a community claim - check getPathwayInfo(WPID) last-edit/curation before leaning on a single WP pathway for a key conclusion - and run WP as a COMPLEMENT to KEGG/Reactome, never a sole peer-reviewed source.

Tool Taxonomy

Decision Tree by Scenario

Over-Representation Analysis (enrichWP)

Goal: Find WikiPathways over-represented in a thresholded gene list, against a defensible background.

Approach: Convert significant genes to Entrez, pass the tested-gene set as universe, run enrichWP, then make the result readable. enrichWP downloads the current/ GMT - acceptable for exploration, but pin a date for anything reportable.

library(clusterProfiler)
library(org.Hs.eg.db)

# enrichWP downloads the current/ WP GMT over the network; symbols/Ensembl must be Entrez first
sig <- bitr(sig_symbols, fromType='SYMBOL', toType='ENTREZID', OrgDb=org.Hs.eg.db)$ENTREZID
all_entrez <- bitr(tested_symbols, fromType='SYMBOL', toType='ENTREZID', OrgDb=org.Hs.eg.db)$ENTREZID

wp <- enrichWP(gene=sig, organism='Homo sapiens', universe=all_entrez,
               pvalueCutoff=0.05, pAdjustMethod='BH', minGSSize=10, maxGSSize=500, qvalueCutoff=0.2)
wp <- setReadable(wp, OrgDb=org.Hs.eg.db, keyType='ENTREZID')   # geneID column -> symbols
as.data.frame(wp)   # ID=WPID, Description, GeneRatio, BgRatio, p.adjust, qvalue, Count

GSEA (gseWP)

Goal: Find WikiPathways whose genes shift coordinately across the full ranking, with no cutoff.

Approach: Build a NAMED Entrez vector sorted DECREASING by the ranking metric, fix the permutation seed, then run gseWP. There is no universe argument - FCS uses the whole ranked list.

gl <- sort(setNames(de$log2FoldChange, de$entrez), decreasing=TRUE)   # named, decreasing, Entrez names
set.seed(123)                                                          # fix permutation reproducibility
wp_gsea <- gseWP(geneList=gl, organism='Homo sapiens',
                 pvalueCutoff=0.05, pAdjustMethod='BH', minGSSize=10, maxGSSize=500)
as.data.frame(wp_gsea)   # NES, p.adjust, core_enrichment (the leading edge)

Reproducible Analysis with a Dated GMT (the correct pattern)

Goal: Make a WP analysis reproducible across re-runs by pinning a dated release instead of pulling current/.

Approach: Download a dated GMT (pass format='gmt' - the default is gpml), split the compound name%version%wpid%org term field into TERM2GENE/TERM2NAME, run enricher/GSEA on the pinned sets, and report the date in methods.

library(rWikiPathways)
library(tidyr)

# downloadPathwayArchive needs an organism to actually download a file (organism=NULL opens the index)
gmt <- downloadPathwayArchive(date='20240310', organism='Homo sapiens', format='gmt', destpath=tempdir())
wp2gene <- read.gmt(file.path(tempdir(), gmt))
wp2gene <- separate(wp2gene, term, c('name','version','wpid','org'), sep='%')   # term is a %-joined compound
t2g <- wp2gene[, c('wpid','gene')]   # TERM2GENE
t2n <- wp2gene[, c('wpid','name')]   # TERM2NAME

wp_pinned <- enricher(sig, universe=all_entrez, TERM2GENE=t2g, TERM2NAME=t2n)   # report date='20240310'

gson_WP(organism) returns a GSON snapshot object, but it still pulls current/ - it freezes a session, NOT a chosen historical date. Only the dated downloadPathwayArchive GMT survives a re-run months later.