In a large and variable dataset such as in this EST study, one can expect that extensive rate heterogeneity that may not be correctly accounted for by using non-partitioned analyses.
If such branches are equally distributed among datasets, at least 35 datasets should have estimated rate variation less than (μ+2σ)/(μ−2σ); when this is not the case, we infer that error in rate estimation is present and causing more extensive rate variation than expected based on the stochastic substitution process alone.
Summarizing rates of evolution of Influenza A viruses and possibly other segmented RNA viruses by a single number might therefore not give a realistic picture of the extensive rate variation found in these viruses.
However, the coefficient of variation in the absence of multiple rate categories (median estimate of 0.42; 95% credibility interval of 0.32 to 0.52) was not found to provide evidence of extensive rate heterogeneity, hence this alternative model was rejected.
The existence of such an extensive rate variation among lineages precludes the application of the linearized tree method [ 45] whose estimates can be highly biased when the molecular clock does not hold [ 46].
A possible reconciliation between the extensive rate of gene duplications and the gene balance hypothesis is that functional divergence directly followed the duplication event, thereby facilitating the retention of both duplicates [ 23].
We recently demonstrated that myeloma cells are highly sensitive towards proteasome inhibitors due to their extensive rate of immunoglobulin synthesis, thereby triggering the terminal unfolded protein response (UPR) and apoptosis via endoplasmic reticulum (ER) stress.
Neighbor-joining trees were constructed (Poisson model) [ 37] and sequences presumed to be non-orthologous, due to extensive rate variation and evidence of gene duplication, were excluded from further analyses.
However, the inclusion of the rate categories did not significantly improve the fit, and the coefficient of variation in the described analysis with a single rate category had a median value of 0.33 (95% credibility interval of 0.22 to 0.45), which suggested there was not strong evidence of extensive rate heterogeneity.
These rates are similar to those estimated in previous studies (e.g., [ 11, 16]), and our results suggest that there is extensive rate heterogeneity between the different segments of the Influenza A genomes of H1N1 viruses, and even within segments as demonstrated in particular by the posterior estimates for M2 and M1.
