Fig. 4: Characterisation of predicted non-AUG initiation codons.

a Distribution of start codon types predicted by Trips-viz (RiboSET). b The stability (dG, Gibbs free energy) of mRNA secondary structure downstream of start codons within 22nt window ‘0’ corresponds to the start codon. 390 starts from RiboSET (green line), a sample of 400 AUG starts with high-scored TIS (blue line), a sample of 400 AUG starts with low-scored TIS (orange line). Lines are mean values across genes with 95% confidence intervals. c Cumulative distribution functions of TIS scores for all non-AUG codons in theoretical NTE, RiboSET (green), all non-AUG codons in theoretical NTE, PhyloSET (yellow), Trips-viz predicted non-AUG starts, RiboSET (purple). d TIS sequence logo and frequency plot of non-AUG starts predicted by Trips-viz. e Fraction of genes in RiboSET, PhyloSET and UntranslSET with one or at least two alternative localisations in the Human Protein atlas. f Probability of mitochondrial presequence calculated by TargetP 2.0 for RFK. g Probability of signal peptide predicted by SignalP 5.0, for RAE1. h Domain organisation of proteforms from RiboSET (purple labels) and PhyloSET (black labels) with TM regions found by DeepTMHMM. Yellow region (M) is the TM helix, blue (I) is inner and pink (O) is the outer cell compartment. Red asterisks represent the start of CDS in a proteoform. Source data is provided as a Source Data file.