Pathway Over-Representation Analysis (2D)

Correlation Preset

Parameters

Significant genes K 30

Pathway gene-set size M 24

Live Statistics

Overlap k

Expected overlap

0.0

Fold enrichment

–

Hypergeometric p (exact)

–

Monte Carlo p (empirical)

not run

Background N / K / M

180 / 30 / 24

How it works

This is over-representation analysis (ORA) — the classic 2D counterpart to the 3D running-sum GSEA sim: instead of walking a ranked list, it asks a simpler question with an exact answer. Out of N background genes, K are "significant" (differentially expressed) and M belong to the pathway gene set. If they were independent, how surprising is the observed overlap k?

P(X = x) = C(K,x)·C(N-K, M-x) / C(N,M)   (hypergeometric pmf)
p-value  = P(X ≥ k) = Σ P(X = x) for x = k..min(K,M)

This is exactly Fisher's exact test on the 2×2 contingency table (significant × pathway), the standard ORA method used by tools like DAVID, PANTHER and clusterProfiler — a genuinely different statistic from GSEA's rank-based running sum, testing the same biological question of pathway enrichment.

Top field — a real 2D force simulation: every gene is a particle spring-pulled toward a target inside the "Significant" circle, the "Pathway" circle, their overlap lens, or the background, then jostled by pairwise repulsion so genes never fully stack — the Venn diagram emerges from the physics, it isn't drawn as static shapes.
Bottom chart — the exact hypergeometric distribution P(X=x) for the current N,K,M; the gold bars from x=k onward are the p-value's tail region, and white dots (after running Monte Carlo) show the empirical frequency from thousands of random draws landing on the same curve.
Presets — "Enriched" weights the random pathway draw toward already-significant genes (mimicking a real enriched pathway); "Depleted" weights it away; "Random" draws independently, so on average the overlap should sit right on the expected value K·M/N.

Significant only Pathway only Overlap (both) Neither