7. Further analyses

These checks use only the Twist1 network, so they describe behaviour, not biology. They include a driver-aware baseline (personalised PageRank from Twist1), the strongest agreement with NetPert (rho 0.79).

Figure 8. DIR enrichment on : top-10, 20, 50 and 100 are all direct driver–response connectors (149/200 at 200).

Figure 9. Up- versus down-regulated Twist1 responses (log2FC).

Figure 10. Rank shifts between NetPert and betweenness on the network.

Figure 11. Genes perturbation theory lifts most above shortest-path ranking ().

Takeaway. On the network NetPert's top ranks are all direct driver–response connectors, it agrees with betweenness yet is more robust to missing edges, and directed regulatory edges measurably contribute. The dissemination assays cannot confirm or refute this (n = 18, responses only).

Agreement slices ()

**Responses only**
Both methods score responses low; different floors.

**Intermediates only**
Method agreement on the genes NetPert is built to prioritise.

**Assay rho by method**
No method predicts the 18-gene response-only readout; honest null.