Testing multiple branches in one run using different tags

[Nathanael Walker-Hale]

Hi Iakov,

How should one go about specifying multiple branches to test in the same run? Do I have to use the branch number, or can I, analogously to PAML, label them #1, #2, #3, etc.?

By the way, I asked over on Bioinformatics, but I’ll put it here as well just in case. Is it possible to fit the branch models (e.g. Yang and Nielsen 1998) as opposed to the branch-site models. That is, the model that averages over all the sites for a particular lineage?

Thanks again for your help,

Best,

Nat

[Iakov Davydov]

Hi Nat,

Apologies for not responding on bioinfo SE, apparently email notification wasn’t configured properly. Should work fine now and I replied to your question there.

Regarding the original question, there’s currently no way to run godon branch-site model test on an arbitrary set of branches. That said, there are the following possibilities:

• run test on all the branches --all-branches
• all branches excluding leaves --all-branches --no-leaves
• a set of branches jointly (via marking all of them with #1). but that is a different test of course.

If that does not fulfill your needs I have the following suggestions:

• if tree dataset is relatively small, test everything and discard tests you do not need
• on a moderate-size dataset generate multiple trees with the branch-of-interested marked with #1. Perhaps, this script can help you to generate multiple trees from a single master tree with all the branches marked with #1. I think you need options --only-labeled and --deroot (useful if your original tree is rooted)
• in case you are using --m0-tree (usually a good idea), you can either run M0 branch-length optimizations in a separate run and then fix branch-lengths. or (slightly more cumbersome) use this option only for the first test, and then extract M0 tree branch-lengths from the json output. in any case this is useful if M0 branch-length optimization is slow.
• finally, it is possible to implement this logic in godon (somewhere here), but I cannot give you a timeline for this. pull requests are welcome of course

Hope this helps.

‌

[Nathanael Walker-Hale]

Hi Iakov,

Thanks so much for your reply, that’s really helpful. I had already started doing your second suggestion (generating multiple trees with differently positioned #1 branches), but hadn’t thought to fix the branch lengths to save time. Thanks for the suggestion.

It would be a nice test of my golang, so if I get a chance I may have a go and send a PR.

Thanks again,

Nat