(a) A diagram showing the three rules used to annotate NMD-elicit mutations. (b) A schematic showing the pipeline used for the prediction of NMD-elicit mutations in TCGA data. (c) A summary of the number of NMD-elicit mutations categorized by their original TCGA classification as indicated. FSD, frameshift deletion; FSI, frameshift insertion; MM, missense mutation; NM, nonsense mutation; Silent, silent mutation; IFD, in-frame insertion; IFI, in-frame insertion. (d) Boxplots comparing the expression levels of genes that harbour NMD-elicit (green) and NMD-escape (red) frameshift indels and nonsense mutations. The horizontal line at 0.5 indicates no differential expression. The expressions in the NMD-elicit group (N=53,406) are significantly lower than those in the NMD-escape (N=19,773), non-PTC-harbouring (N=488,430) or silent (N=202,508) groups (median ratio of REV=0.54; ***P<2.2e−16, one-sided t-test). (e) Scatter plots showing the correlation between the number of all mutations in a sample and the number of NMD-elicit mutations for each tumour type as indicated. The grey dashed lines are the angle bisectors of the first quadrants.