Supplementary Figure 2: Probabilistic mapping approach.

(a) Ribo-seq produces several distinct reads from a given codon (AGU) located in the P site of the ribosome. Besides the canonical read (1), both shorter (2) and longer (5) reads arise due to stochastic mRNA cleavage. Further variability is caused by observed (3) or unobservable (4) untemplated single nucleotide additions. The proportions of these reads are determined by the model parameters. (b) The deterministic approach of mapping reads to P site codons assigns all reads of a certain class (characterized by its length and 5’ mismatch state) to a specific codon at the same position within each read. In contrast, probabilistic mapping solves the inverse problem of the generative model to assign reads to codons. Finally, regularization can be used to overcome the problem of fractional mappings that are due to the over-parameterized problem.