Fig. 3: Comparative evaluation of percentile correlation on simulated clusters.
From: acorde unravels functionally interpretable networks of isoform co-usage from single cell data
a Proportion of co-expression values above 0.8 in each empirical cluster, obtained after running the acorde clustering pipeline on simulated clusters using several correlation and proportionality methods as a distance metric. b Jaccard index of simulated vs calculated clusters obtained with each evaluated co-expression method. Simulated clusters were paired with one calculated cluster based on mean profile similarity, and synthetic transcripts in each of the paired clusters were compared. c Percentage of unclustered isoforms generated by each co-expression method. Results are shown before and after re-assigning unclustered isoforms by co-expression with the mean profile of extant clusters (i.e., cluster expansion). d Evaluation metric overview. Metrics are specified in the grid headers. x-axis shows values of the different metrics, y-axis displays evaluated co-expression methods.
