Supplementary Figure 3: Start-codon prediction.
From: Improved Ribo-seq enables identification of cryptic translation events
(a) Potential start codons are predicted by comparing the activity of each codon to the mean activity of downstream codons. Data from several different kinds of samples such as translation start site profiling employing cells treated with Lactimidomycin or Harringtonine prior to cell lysis can be included into the start codon prediction using a logistic regression approach. A start codon score (s) is computed via logistic regression for each codon in an ORF candidate. Three examples (x, y and z) are shown. (b) For all annotated ORFs with at least 25 reads, start scores (see Figure 1e) were predicted for all codons. We computed the relative ranks for all codons (i.e. the maximal scores get rank 0) and plotted the distribution of ranks for the annotated start codon. For both the HCMV and HSV-1 data set, about three quarter of the start codons are top ranked.
