Fig. 5
From: Simulation-based benchmarking of isoform quantification in single-cell RNA-seq
Effect of dropouts on isoform quantification. a Impact of removing isoforms with more than a threshold number of dropouts on Spearman’s rho and the NRMSE for the BLUEPRINT B lymphocytes (left) and the Kolodziejczyk et al. ES cells (right). The x-axis gives the threshold percentage of zeros above which an isoform is removed from the analysis. For example, a threshold percentage of 80% would result in isoforms with zero expression in 80% or more of cells being removed from the analysis. Each colored line is a linear fit for visual guidance and it represents a different isoform quantification tool. b Relationship between how highly expressed an isoform is and the percentage of cells in which it has zero expression. The relationship is considered in RSEM simulated BLUEPRINT B lymphocytes (top left), the real BLUEPRINT B lymphocytes (bottom left), the simulated Kolodziejczyk et al. ES cells (top right), and the real Kolodziejczyk et al. ES cells (bottom right). Each point represents an isoform and the points are colored according to density