Fig. 5

a Comparison of the transcript abundances (y-axis) predicted by RATTLE, FLAIR, StringTie2, and TALON, with the abundances of the SIRV spike-in transcript isoforms (x-axis). Each panel shows the Pearson correlation r for the comparison of the abundance values. Units on the y-axis vary according to the method: RATTLE provides abundances as read counts per million, like TALON and FLAIR. StringTie2 produces a TPM value. SIRV data corresponds to the RNA1 sample. Correlations for other datasets are provided in Additional file 1: Fig. S10. b Distribution of the lengths (y-axis, log10 scale) of the transcripts predicted by each method compared with the transcripts from the reference (Gencode v29) using cDNA and dRNA. StringTie2 was run with (stringtie) and without (stringtie_no) the annotation. The dashed line indicates the median value for the reference transcripts. c Distribution of the proportion of the annotated transcript covered (y-axis) by the best matching predicted transcript from each method. The dashed line indicates 0.75. d Pearson correlation (y-axis) of the predicted transcript abundances in log10(TPM) scale from two replicate cDNA and dRNA sequencing experiments from the Nanopore sequencing consortium. For each method, we give the number of predicted transcripts that were matched to the reference annotation in both replicates (x-axis). Colors are like for b and c. e Saturation plot of transcripts using an incremental number of dRNA reads as input. Left panel: for each input (x-axis), we give the total number of transcripts predicted by each method (y-axis) with an expression of > 5 reads. Right panel: for each input (x-axis), we give the total number of annotated transcripts matched by the predicted transcripts from the left panel with a coverage of at least 75%. The number of transcripts obtained by each method and input is given in Additional file 2: Table S10