Fig. 5: Self-organizing map summary figures.

Self-organizing map (SOM) of log counts per million for 9433 genes across cattle, deer, and pig transcriptome analyses, showing mapping quality, distance, and clustering metrics. a heat scale is quality scores based on the average sum of squares distance of the count profile for each gene in the SOM unit to the “codebook” vector of the unit; lower scores (lighter colors) represent shorter distances between a gene’s expression profile and the codebook vector for the unit; b heat scale is average sum of squares distances of each SOM unit from its neighbors; lighter regions show units that are closer to their neighbors, and thus have genes with more similar gene expression patterns; c heat scale is the number of genes mapped to each SOM unit; this metric is expected to be similar between units across the map, and most units on our map had 20–30 genes. d The map showing the four k-medoid clusters of SOM units. Genes in clusters 1 and 3 exhibited patterns that support cranial appendage homology: Cluster 1 genes tended to have average log counts per million higher in cattle (Bovidae) and deer (Cervidae) cranial appendage tissues than in pig outgroup tissues, and genes in Cluster 3 tended to have lower average counts in cranial appendage tissues than in pig outgroup tissues. The contribution to the SOM by each tissue and taxon is in Fig. 6.