HGF amplifies
PIK3CA-driven pathology.

Each organoid was measured at 13 time points, every 4 hours from 0–48 h. Three narrow annuli were centred at 0.80, 0.85 and 0.90 of the fitted microwell radius. Each circumference was sampled into 360 angular bins (one per degree), with 5° circular smoothing.

At each angle, brightfield edge, local texture and darkness evidence were assessed directly on the annulus. A candidate required either agreement between at least two channels at robust z ≥ 2.5, or one strong channel at z ≥ 4.5. Runs shorter than 12° were removed.

A hit had to overlap an adjacent frame within ±10° and outer-ring hits required supporting evidence on the inner rings. Persistent wall-crossing stitch seams were suppressed. Microwells listed in microwell_exclusions.csv were removed before summaries were calculated.

For consecutive frames, entries are newly contacted angular bins and exits are bins no longer contacted. Turnover is the fraction occupied by either entries or exits. The organoid value uses the maximum turnover across its three rings and is averaged across time. The population chart shows the median organoid value for each condition.

A organoid is active at a time point when contact changes on at least one ring. Its activity frequency is the percentage of measured four-hour steps that are active; missing measurements are excluded. The population chart shows the median across organoids, so it can equal 0% when at least half of the population is Never even though the remaining organoids are active.

Each organoid is classified from its activity frequency: Never: 0%; Occasional: >0–34%; Repeated: >34–67%; Frequent: >67–100%. The stacked bars show the percentage of the condition in each class. These classes describe how often contact changed, not whether an event was migration, growth or budding.