[adamjohnwright]

A curator testing DeltaSignal on beta couldn't tell what the result table's Influence column meant, nor quite what the others meant.

The change

• Clearer headers: Entity, Activity (× normal), Change vs baseline, Influence. The headers also get role="columnheader", to go with the table's existing roles.

• A short note above the table, "What do these columns mean?". It opens in place, so it works on touch screens, where tooltips don't. It defines:

  • Activity: the predicted steady-state activity on the input's scale: 1× is normal, 0 is none, and higher is more active. (node_activities × 100, per DeltaSignal's docs/API.md.)
  • Change: activity minus the entity's own baseline. It's a difference, not a ratio, and it's what the diagram's colours show.
  • Influence: how strongly the reactions the entity takes part in respond to a small change in it, at the predicted state. It's a ranking, not a unit. (Σ|∂F/∂input| at the solution, per docs/API.md and compute_influence_scores.)

  The note also says that rows are sorted by the size of the change, that a highlighted row is one the user perturbed, and that an entity made of several logic nodes shows their average activity and change and their largest influence (aggregateRowsByReactomeId).

Text and style only. The pre-push checks passed: types, lint, unit tests, builds and an end-to-end smoke run.

🤖 Generated with Claude Code

[adamjohnwright]

Replaced by the next PR, which corrects the Influence definition.