AIM model

[chris blair]

Hi all, 

I just saw the new paper about the AIM addition to starBEAST2. I am interested in trying it with my data. Is there a tutorial available anywhere?

Chris

[Alexei Drummond]

I think that Nicola (cc’ed) and Huw are working on one at the moment. Should be available very shortly.

Alexei

--
You received this message because you are subscribed to the Google Groups "beast-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to beast-users...@googlegroups.com.
To post to this group, send email to beast...@googlegroups.com.
Visit this group at https://groups.google.com/group/beast-users.
For more options, visit https://groups.google.com/d/optout.

[chris blair]

Sounds good. I just came across the SpeciesNetwork package as well. What are your recommendations regarding which package/model to use to estimate a species network? I am assuming that starBEAST2 provides a bit more speed and it is able to estimate some additional parameters. I have thousands of loci, so I will have to prune for either analysis. Would one be more suitable for NGS data? Thanks.

Chris

[Huw A. Ogilvie]

Hi Chris,

The latest version of StarBEAST2 (v15) available in BEAUti includes an AIM template.

AIM and the multispecies network coalescent (MSNC, implemented in the SpeciesNetwork package) are different models. The AIM model is based on a species tree, with continuous gene flow occurring between all pairs of species that exist at the same point in time - it is an approximation of the IM model where the species tree is jointly estimated with the migration rates, but the migration histories of the gene lineages are integrated out. The MSNC model is based on a species network, where reticulate nodes represent instantaneous transfer of genes from one species to another, including transfer from species from lineages which may have no living or no sampled descendants in the present.

Both of these models and methods are very new. My personal opinion is that we don't know at the moment which model is a better representation of gene flow. It could easily be the case that some biological systems better fit the AIM model, and others better fit the MSNC. We also don't have any kind of systematic study of the computation scaling of the models and methods, as has been done for species tree methods (e.g. Ogilvie et al. 2016). Hejase and Liu published a paper on the scaling of non-Bayesian methods of species network inference ( https://bmcbioinformatics.biomedcentral.com/articles/10.1186/s12859-016-1277-1 ) but AIM and SpeciesNetwork are both Bayesian methods.

- Huw

[Graham]

Chris,

You might also be interested in my DENIM package (https://academic.oup.com/sysbio/advance-article-abstract/doi/10.1093/sysbio/syy041/5003138).