Comparison of observed and predicted coverage patterns. Comparison of observed and predicted coverage patterns. (A) Correlation between observed and predicted number of reads spanning each junction for the HAP1 sample. The left column (“Predicted coverage”) shows the actual number of reads predicted by alpine and the respective transcript abundance estimation method, whereas the predicted values in the right column (“Scaled predicted coverage”) are scaled to sum to the same number as the observed number of uniquely mapping junction reads within each gene. Scaling improves the correlation between observed and predicted junction coverages for all included methods. Axes are square root transformed for better visualization. Red points indicate junctions where less than 75% of the spanning reads are uniquely mapping. (B) Overall distribution of the gene-wise JCC scores for each method in the HAP1 sample, as well as the association between the JCC score and the total number of reads for the gene and the number of uniquely mapped junction reads in the gene. Charlotte Soneson et al. LSA 2019;2:e201800175 © 2019 Soneson et al.