Assembler performance on simulated RNA-seq data. (a) Overall sensitivity, defined as the fraction of simulated transcripts that were predicted correctly. (b) Overall precision, defined as the fraction of predicted transcripts found in the set of simulated transcripts. (c, d) Sensitivity and precision as functions of simulated and predicted abundance, respectively. (e) Transcript length distributions of predicted and simulated transcripts (logarithmic binning). (f) Mean absolute difference between the number of predicted and the number of simulated splice variants for each predicted gene. (g) Spearman’s rank correlation between predicted and simulated abundances for transcripts predicted correctly by all five assemblers.