Schematic overview of the proposed approach for inferring translation kinetics from RP data.

<p>To obtain a segment tree representation of the RP data (left) mapped ribo-seq (light grey) and RNA-seq (dark grey) reads are assigned to nested segments of decreasing lengths (starting from segments [1, <i>S</i>] equivalent to the full-length CDSes) while there is sufficient data. Ribosome densities <i>ω</i> for each segment are computed for the available replicates and are used to parameterize the log-normal distributions describing measurement error of these segments. To determine per-gene translation initiation rates <i>k</i>₀ and per-codon elongation rates <i>k</i>_AAA, …, <i>k</i>_GGG many candidate sets of translation rates are tested. For every candidate set the TASEP model of translation is simulated with the proposed rates for all genes in the <i>model simulation step</i> (right). Ribosome occupancy, i.e. the relative amount of time ribosomes spend at a particular location on the mRNA, obtained from the simulation (dashed grey) is then aggregated per segment to compute the average occupancies <i>N</i>, which are compared the log-normal distributions of the corresponding segments from the segment tree representation in the <i>model evaluation step</i>. Evaluation results are used by a genetic algorithm to propose new candidate sets of rates and repeat the simulation-evaluation cycle until the search for translation rates converges. To simplify notation, the gene index <i>g</i> is dropped for all gene-specific variables in the figure.