search

blab / mers-structure

MERS-CoV spillover at the camel-human interface
Creative Commons Attribution Share Alike 4.0 International
7 stars 3 forks source link

Essential revision 5 #23

Closed evogytis closed 6 years ago

evogytis commented 7 years ago

More generally, the model choices need better explaining. Why delve into a structured coalescent in BEAST2 for the ancestral reconstruction, but go back to the Skygrid in BEAST1 for computations of Ne? We assume this is a pragmatic choice, and for the latter you carefully reduced the human clusters to reduce bias, but we think the rationale for your choices need laying out more clearly. Even if pragmatic rather than principled, (e.g. there are no structure coalescent options in BEAST1), we think it still needs to be stated why you made the choices you did. Especially since there are other recently-developed BEAST2 packages that could be used to fit the same structured coalescent model: BASTA & MASCOT, as well as the very flexible PhyDyn package (which might offer improvements in computation time).

evogytis commented 7 years ago

This one's easy, can do today even. We had good rationale against using BASTA (poor sequence sampling in demes), not sure MASCOT was around when we started. Sad to say I'm not familiar with PhyDyn or its applications.

trvrb commented 6 years ago

Agreed that this is resolved.