Five vertically stacked rings represent the major omics layers, ordered along the flow of biological information: Genomics → Epigenomics → Transcriptomics → Proteomics → Metabolomics. Each ring holds 12 feature "modules" (a gene and its downstream products). Every module's abundance value drifts as a mean-reverting stochastic process, mimicking real biological noise:
dv = k(μ − v)·dt + σ·dW (Ornstein–Uhlenbeck)
k = 0.6, μ = 0.5, σ = noise slider
Two integration strategies connect matching modules between adjacent layers with a colored edge:
Correlation (late integration):
sim_i = 1 − |v_L,i − v_(L+1),i|
edge active if sim_i ≥ threshold
Concatenation (early integration):
edge active if v_L,i ≥ 0.7·threshold AND v_(L+1),i ≥ 0.7·threshold
Integration index I = (active edges / possible edges) × 100%
Composite score S = Σ(w_l · mean(v_l)) / Σ(w_l) × 100%
- Correlation threshold — similarity/co-activation cutoff required for an edge to form.
- Biological noise (σ) — volatility of each module's random-walk abundance signal.
- Transcriptomics weight — reweights that layer's contribution to the composite score.
- Correlation / Concatenation — switches between late-integration (similarity network) and early-integration (independent thresholding) strategies, a real distinction in multi-omics pipelines.
- Geno/Epi/Trans/Prot/Metab — show or hide an omics layer and its links entirely.
Real pipelines (e.g. MOFA, iCluster, similarity network fusion) face exactly this trade-off: correlate layers after separate processing, or concatenate them early and threshold jointly — the choice changes which biological modules appear "integrated".