Extended Data Fig. 8: Griph robustness analysis and comparison to other methods.
From: Self-organization and symmetry breaking in intestinal organoid development
Comparison of Griph lower dimensional embedding (Griph/LargeVis) to different embedding approaches (PCA, PCA combined with distributed stochastic neighbor embedding (t-SNE), diffusion maps) and analysis of method sensitivity for variable gene selection. Five different sets of variable genes have been selected (using Griph to retain 10%, 25% or 50% of genes per bin, by Michaelis–Menten fitting of the gene dropout rates as implemented in M3Drop, or by the previously described mean-variance fitting procedure48) and analysed. a, b, First and second dimension are shown and results are colour-coded for enterocyte marker genes (as in Extended Data Fig. 9b) (a) and YAP1 target genes (as in Fig. 5b) (b). n = 1,863 cells.
