Destabilizing mutations could be removed both by removing sites with high statistical correlations to other positions and by removing nearly invariant positions at which "hidden correlations" can occur.
In all cases, a discrete gamma-distribution model with four rate categories plus invariant positions was used, estimating the gamma parameter and the fraction of invariant positions from the data.
The HMM logo [17] of the alignment (Fig. 2) reveals highly variable zones and a few invariant positions.
The distributed and largely invariant positions of DIMM cells in the CNS suggest their derivation from numerous, different NBs, but this supposition awaits future experimental analysis.
Another observation on the alignment in Figures 1A and 2 is the remarkable tendency of invariant positions to contain Gly, Pro and Asn.
The distribution of such invariant positions and the spacing between them appears mostly conserved in the natural sequences contained in the Rd.HMM search results.
In fact, an analysis of the structural features of these invariant positions confirmed the occurrence of strained dihedrals (this is shown below under "Rd.HMM of highly mobile proteins").
Hence for each nucleotide position in a primer site, there were only three possible variants of the initial sequence, and a scale of 0 to 3 was used; invariant positions scored zero.
The tree was constructed using a Maximum Likelihood approach as implemented in the PhyML program [29], using a discrete gamma-distribution model with four rate categories plus invariant positions.
The "WAG+I+Γ+F" model, in which among-site rate variation was approximated by a discrete gamma distribution plus the proportion of invariant positions, and aa frequencies were estimated from the data, was selected for the GapC1 analyses according to Akaike Information Criterion.
In addition, those invariant positions in the Rd.HMM for the X-ray structure, where the natural sequence was coincident, concentrated on border zones of helical and extended structure, or in the zone of turns (see 1-letter amno acid codes in Figure 6B and 6D).
