Msprime 1.4

Msprime introduces the SMC(k) model, significant performance improvements, and uses tskit 1.0.

SMC(k)

Now, species with high scaled recombination rates (large Ne and large ρ), like Drosophila, can be simulated efficiently with the new SMC(k) model. The SMC(k) model approximates the Hudson model, and the level of approximation is set by the user as the K parameter. For example, K=0 simulates an SMC model, SMC(1) simulates an SMC’ model (for discrete genomes), and SMC(L) simulates a full Hudson model.

Using the SMC(k) model is as easy as:

msprime.sim_ancestry(2, model=msprime.SMCK(1))

The SMCK model is an approximation to the Hudson model, and it is optimised for small samples in the Drosophila like regime. We recommend using the Hudson model outside of these cases.

For more information, check the msprime documentation, and keep an eye out for a publication that’s coming soon.

Performance Improvements

Simulations of the standard coalescent model are now much faster for large sample sizes (1.6-fold faster for 1 million samples in this example).

Other Notable Changes

• For the ARG-heads, simulations above the local MRCA are now supported.

  • To simulate above the local MRCA (until all local MRCAs are reached), set the stop_at_local_mrca to False

  

  • For example, here we see that node 14 is a unary node at the third tree, which would have been impossible to observe otherwise.
• Support for internal samples for the FixedPedigree simulation model.
  • Previously, samples that are internal nodes in the pedigree were not allowed to have children as a simplification. This simplification is now lefted.
• Demography objects can now be created from provenance entries.
  • With this, it is now possible to plot a demes demography from a tree seq generated by stdpopsim, without having to install stdpopsim, know the model etc.
• msprime 1.4 requires tskit >=1.0 (breaking change)
  • tskit 1.0 marks the point at which backwards compatibility is guaranteed for users. It also marks a significant milestone with many changes and features involved.

Feedback Etc.

Please post a discussion or an issue on the msprime GitHub repository if you have any problems. We hope the update to msprime version 1.4 is useful in your research.