Fig. 7
From: Dictionary learning allows model-free pseudotime estimation of transcriptomic data
Correlations of experimental times with pseudotimes for 6 real-world datasets with multiple subtypes. Subfigures in each row show evaluations for each dataset. Subfigures to the left show results for pseudotimes derived based on the ranks of the values of the coefficient vector for the best scoring component; Subfigures to the right show the correlations for pseudotimes derived based on the entire low-dimensional representation with Monocle’s polygonal reconstruction. For dataset GSE92652 Monocle fails, which is why only the remaining 5 datasets are analysed here. The x-axis displays the number of components the data is reduced to, or for UMAP the number of neighbours and for t-SNE the perplexity. The majority of correlations are higher when pseudotimes are derived based on one atom