Using a number of SNPs reasonably obtained through next-generation sequencing approaches (10,000 - 50,000), accurate parameter estimates and model selection were possible for models with more ancient demographic events, even given relatively small numbers of sampled individuals.
For example, under the IM model, the median Akaike weight reached 1.0 (100% support for the true model) at a sample size of n = 3 per population for models with more ancient divergence.
As such, the recent parameterizations for single-population models included more relaxed population size changes, but the same θ values used for models with more ancient demographic events.
As expected, RMSE for parameters in simulations with recent demographic events were often larger than for data simulated under the same model with more ancient demographic events.
As expected based upon the short sequences being used [41], phylogenetic analyses of the TGFβ superfamily have limited power and we were unable to exclude alternative models with a more ancient origin of the α-subunit lineage (Fig. S2A).
With more ancient events and excluding the BG model, 95% CI performed well, containing the true parameter value for all model parameters in more than 90% of replicates for datasets composed of 10 or more diploid individuals (Additional file 1: Table S1).
Although the results from Naeglaria could be explained by a more ancient primary endosymbiosis [ 32], no current model of plastid evolution predicts EGT into Monosiga, a member of the opisthokonts.
As a result, classes of human disease that contain many vertebrate-specific genes are unlikely to be as highly conserved in certain mammalian models as disease classes of more ancient origins.
Other well-characterized arthropod species, such as Cupiennius salei (Damen et al., 2000; Stollewerk et al., 2003) and Strigamia maritima (Chipman et al., 2004; Chipman and Akam, 2008), provide additional models of interest for understanding more ancient evolutionary history.
We expected that these datasets would be more difficult for parameter estimation and model selection, due to the substantially more ancient timing of the simulated events.
