Skip to main content
Fig. 2 | BMC Bioinformatics

Fig. 2

From: scEvoNet: a gradient boosting-based method for prediction of cell state evolution

Fig. 2

The development of the specific cell type between frog and mouse. A First UMAP represents highlighted annotated neural crest cells in the whole embryo dataset, second UMAP represents predicted neural crest cells with our classifier, and third UMAP represents predicted scores of our classifier (the predicted scores represent the probability of a given data point belonging to the NC class). B The AUC score for the neural crest classifier is 0.89 (it measures the ability of a model to distinguish between positive/negative classes by calculating the area under the ROC curve, which is a plot of the true positive rate against the false positive rate as the decision threshold is varied). C The confusion matrix for mouse and frog samples (prefix x_ is for Xenopus, prefix m_ is for mouse). The values in the confusion matrix are the correlations between two lists of scores for all cell type models. D Selecting the subnetwork of 300 shortest paths from Xenopus neural crest to mouse neural crest shows characteristic genes that are shared with closely-related cell types, such as mafb or cldn6 (group 3) between frog neural crest and mouse neural tube. It also reveals two groups of genes: genes from group 1 are organism-specific genes (frog neural crest and frog neural plate), and genes from group 2 are important genes for the specific cell type (NC) in both organisms (x-neural crest and m-neural crest)

Back to article page