Matching cells between apical and mid-basal layers

We also wanted to characterise the 3D geometries of cells within small domains, using the cell shapes and arrangements in the two depths we had tracked. To do this, we needed to correctly match cells between apical and mid-basal depths. We manually seeded 3–5 apico-basal cell matches per placode, then used an automated method to fill out the remaining cell matches across the placode and over time.

We did this by sequentially looking at all unmatched apical cells that were next to one or more matched cell. The location of the unmatched basal centroid was predicted by adding the vector between matched and unmatched apical cell centroids to the matched basal centroid. The nearest actual basal centroid to this predicted basal centroid location was chosen as the match, on condition that the prediction accuracy was within 0.25 of the apical centroid distance. Information from multiple matched neighbours was used, if available, improving the basal centroid prediction. Progressively matching cells out from known matched cells filled out placodes for all embryos. We visually checked apical-basal matches for each embryo in movies of an overlay of apical and basal cell shapes with apico-basal centroid connections drawn (Video 3).