Figure 3
From: Dinucleotide controlled null models for comparative RNA gene prediction
Key concepts of the algorithm shown on an example alignment of 5.8S rRNA. (A) Distribution of dinucleotide frequencies of 1000 simulated alignments are shown as box-plots (the line in the box indicates the median, the borders of the box the 25th and 75th quartile, and the dotted lines 1.5× the interquartile range). Red circles show the frequencies observed in the original alignment. (B) Relationship between the number of substitutions and observed differences empirically determined by sampling of 25 points. Each point shows the average of 10 simulations. Note that the short distances are sampled more densely. These settings are the default values in our program and used throughout the paper. (C) Distribution of mean pairwise identities for 1000 random samples. The MPI of the original alignment is shown in red. (D) Comparison of site-wise MPIs in the original alignment and the average of the corresponding sites of 1000 random alignments.