Extended Data Fig. 4: Developing the CytoTRACEKernel.
From: CellRank 2: unified fate mapping in multiview single-cell data
a. Similar to the CytoTRACE publication18, we compute the CytoTRACE score by (i) calculating the number of genes expressed per cell (GEC), (ii) computing each gene’s Pearson correlation with GEC, (iii) mean-aggregating imputed expression of the top 200 correlated genes (Methods). Box plots indicate the median (center line), interquartile range (hinges), and 1.5x interquartile range (whiskers); the shown box plots are schematics. b. Force-directed layout embedding (FLE) of 22, 370 Caenorhabditis (C.) elegans muscle and mesoderm cells undergoing embryogenesis, colored by estimated embryo time78 (left; 130 − 830 minutes), CytoTRACE pseudotime computed using CellRank 2 (middle) and the original implementation (right). c. Quantitative comparison of the two implementations of the CytoTRACE pseudotime on bone marrow77 (using 10x and SmartSeq2), C. elegans embryogenesis78 (subsetted to ciliated neurons, hypodermis and seam, and muscle and mesoderm), and zebrafish embryogenesis79. The x axis (y axis) displays Spearman’s rank correlation between CellRank 2-CytoTRACE (original CytoTRACE) and ground-truth (GT) time labels. Ground-truth labels were derived from either embryo time or stages as in b. (C. elegans and zebrafish embryogenesis) or from manually assigned maturation labels from the original CytoTRACE study18 (bone marrow).
