Issue on re-rooting trees and networks

[Matheus Salles]

Dear all,

I've got a question about rooting trees in PhyloNetworks.

Basically, in my dataset, I've got two lineages that can be considered outgroups along with about 30 terminals of my focal group.

When I run PhyloNetworks without rooting, I can get a network with up to 8 reticulations (and it's got the best pseudo-likelihood value). But when I try to root it, I can only do that up to 3 reticulations. After that, the direction of those hybrid edges probably starts messing things up, and I end up with the error described in the program manual.

Given this context, my question is as follows. If I'm only interested in obtaining estimates of introgression between lineages (and not in the topology obtained through PhyloNetworks), do I really need to root the trees to get accurate estimates? Or can I trust the estimates I get from the unrooted networks, even though I know there's an error when I try to root them?

Please let me know if you need more info.

Best regards,

Matheus

[Cécile Ané]

I would not trust the estimated networks that cannot be rooted by your outgroup clade. These networks are incorrect, based on the external information (not used to infer the networks) that your outgroup clade is in fact an outgroup, so I would not trust these networks (beyond 3 reticulations in your case).

But you could check the alternate networks in files ".networks": These alternate networks have the same unrooted topology as the inferred network, except that the direction of reticulation is different. They are sorted based on their score, and their score is indicated for each. So if one of them can be rooted by your outgroup clade and has an "okay" score, then you could use it in place of the "best-score" inferred network. By "okay" score, you could consider scores that are at least as good or better than the score of the best network with h=3, for example.