Fig. 4

Integrating spliced and unspliced counts improves disease state classification. Label propagation was used to classify cells according to patient disease status from (A) spliced and unspliced or (B) moments of spliced and RNA velocity integrated features generated from ten integration approaches. The boxplots represent classification accuracy according to area under the receiver operator curve (AUC) and the asterisk represents the method with the highest median AUC. Across all three datasets, spliced and unspliced integration achieves increased classification accuracy over unintegrated data