Limitations

Limitations of coala

• Supported mutation models: Infintite sites (IFS) and the finite sites model Hasegawa, Kishino and Yano model (HKY85) and general time reversible model (GTR).
• All summary statistics but sumstat_trees need mutation present in the model. When using finite site mutation models, they also need an outgroup present in the model (with exception of ``sumstat_dna`).
• The sumstat_dna statistic is currently not compatible with any other summary statistic and needs to be the only statistic present in the model. You can repeat the simulation with the same seed once with sumstat_dna and once with the remaining statistics as workaround.
• When simulating selection, coala can incorporate only one selected site per locus due to the same limitation in msms.
• As it incorporates only coalescent simulators, coala relies on the assumptions of the coalescent. It particularly assumes that the sampled sequences are only a small part of each population and the populations are randomly mating.
• It is not suitable for simulating many sequences (>= 10k) in a single model, as it supports no simulator that is optimized for large sample sizes. Take a look at msprime for such models.
• It can employ scrm to simulate long loci with a high recombination rate, but not for models that contain selection or a fixed number of mutations, as these features are not supported by scrm.
• Coala creates temporary files with the simulation results when conducting simulations with ms and msms. It requires enough free disk space on the partition that contains tempdir() in that case.
• Parallelization of simulations on multiple CPU-cores does not work on Windows, and should not be done when using Graphical R Interface like RStudio.