Figure 3

Inferring the length of the regions at the ORF 5'/3' ends that are characterized by different NFC distributions relatively to inner parts of the ORF. (A) - NFC distributions are calculated for each codon type, on windows of 50 codons, for the first and last 200 codons of the ORF. (B) Next, the distance between each pair of NFC distributions originating from different windows is calculated, creating a distance matrix for each codon type. The resulting distance matrices are averaged over all codons, and each column in the averaged matrix is averaged again, overall producing a mean distance vector. Each component in this vector describes the average distance between a NFC distribution calculated in the window it represents to NFC distributions of other windows (across all codon types). To determine at what location relative from the 5'/3' ends the distance between NFC distributions stop to significantly differ, a sliding window of length 10 was applied on the mean distance vector, and the values in and outside the window were compared using a Wilcoxon test. The first test that resulted in a p-value greater than 0.05 defined the location relative to the 5'/3' ends that was characterized by similar NFC distributions. (C) - Each subplot describes the mean distance vector calculated on the first and last 100 windows using the Hellinger metric (dotted graphs). The vertical bars depict the calculated standard deviation for each window. The navy bars beneath mark the regions relatively to the 5'/3' end with significantly different NFC distributions in comparison to subsequent regions on the ORFs. A similar test was directly applied on the averaged RC profiles (instead on the mean distance vector; see Figure 24 in Additional file 1), shown in burgundy bars (absent bars indicate of no such region). To emphasize the difference within each organism, different y-scales were used for each organism. For a comparison between organisms using the same y-scale see Figure 21 in Additional file 1.