Fig. 5
From: Single-cell trajectories reconstruction, exploration and mapping of omics data with STREAM
Comparison with other methods on real datasets. a ENO3, a marker gene for late-stage differentiation, is used to evaluate the myoblast commitment trajectory (highly expressed in fully differentiated cells). b Along the myoblast differentiation trajectory, the scatter plots show the relationship between ENO3 expression and rank-based pseudotime or distance-based pseudotime inferred by STREAM. The dashed lines are fitted curves by generalized additive model (GAM). c Correlation analysis along myoblast differentiation. The bar plot shows for each method the Pearson correlation between inferred rank-based pseudotime and ENO3 expression, Pearson correlation between inferred distance-based pseudotime and ENO3 expression, Spearman correlation between inferred pseudotime, and ENO3 expression and Kendall’s tau correlation between inferred pseudotime and ENO3 expression. Only methods that successfully detect the correct bifurcation are included. d The average rank of three maker genes, i.e., ENO3, MYH2, and MEF2C on four different correlation analysis metrics as described in c. e Using the two marker genes Gata1 and Pax5, which are highly expressed on MEP-committed branch and CLP-committed branch, respectively, and rarely co-expresses in the same single cell, the illustration shows three different reconstruction scenarios: optimal model (high precision and high recall), over-branching (high precision but low recall) and under-branching (low precision but high recall). f The scatter plots show the relationship between precision and recall for each marker gene across methods. Each circle represents one method and its F1 score is reported below. g Stream and subway map plots of the mutually exclusive genes Gata1 and Pax5
