Multidimensional AM-FM models and methods provide us with powerful, image and video decompositions that can effectively describe non-stationary content.
In analogy to the procedure we applied for SNP frequency spectra, we analyzed substitution rates and stationary GC content (GC*, i.e. the GC content toward which sequences are evolving according to measured substitution rates) after controlling for recombination rates or ancestral GC content.
In the next step, we used the estimated substitution rates to calculate the stationary GC content, denoted by GC*, that is, the GC content that is acquired after a long time assuming that the substitution rates do not change (Duret and Arndt 2008).
The above explains the preference for high stationary GC content in tCGI.
To address this, we used the estimated substitution rates in order to calculate the stationary GC content (GC*) along genes.
Once the 12 substitution frequencies are established, the stationary GC content (GC*) can easily be computed [ 44].
Substitution rates and stationary GC content were calculated using a previously developed neighbor-dependent substitution model [ 38, 39].
Furthermore, the stationary GC content (GC*) profiles peak near the TSS, although it does not exceed 50% in most species, while the current GC content is about 60%.
However, the value of the stationary GC content in tCGIs (GC* = 0.41) suggests that these regions are predicted to lose their CGI properties in the long run, whereas dCGI with (GC* = 0.58) will keep them.
Several signatures of GC-biased evolution between species, such as current and stationary GC content and BDS, have been found to be inversely correlated with chromosome length (Fullerton et al. 2001; International Chicken Genome Sequencing Consort 2004; Dreszer et al. 2007).
