Fig. 6

Performance of SVGs selected by each method for clustering spatial domains in mouse embryos. Proportion spatial locations annotated to one of sixteen tissue domains in the E9.5 mouse embryo. Concordance in the clustering outputs and the pre-defined spatial domains in the mouse embryo was computed across a range of top SVGs (between 100 and 1900 genes) selected by each method. Clustering is performed using two spatial clustering methods (BayesSpace and SpaGCN) and four non-spatial clustering methods (SINFONIA’s Louvain and Leiden, k-means, and hierarchical clustering). Concordance between the clustering outputs and the pre-defined spatial domains is quantified in terms of the adjusted Rand index