Supplementary Figure 1: Deterministic versus probabilistic mapping strategies.

(a) The read length distribution of the HSV-1 Ribo-seq experiment (see Supplementary Table 1) shows that cleavage is stochastic. Single 5 mismatches are frequent and shift the read-length distribution by +1. (b) In the HSV-1 data set the most abundant read class is 28 bp long and the annotated codon triplets start at positions divisible by three within the read (Frame=0). Shorter or longer reads with frame 0 are also highly abundant, indicating high variability of cleavage distances downstream of the P site. Additionally, a substantial fraction of 29 bp long reads is in frame 1, and many 27 bp long reads are in frame 2. This indicates that cleavage upstream of the ribosome exhibits significant variation, albeit less than downstream. A substantial fraction of reads has a mismatch at their 5’ end, which can be attributed to untemplated nucleotide additions during cDNA synthesis. (c) For each of the data sets (see Supplementary Table 1), the fraction of reads is shown that exhibit the most frequent read length in CDS mapped reads. In all data sets, only a minority of all reads have canonical length. (d) Only a fraction of all canonical reads correspond to footprints with the P site codon at a defined distance from their respective start position in the read alignment (e.g. position 12 within the read).