Closed roblanf closed 3 years ago
I agree that this is a good thing to add. We do basically prune out the real trash basically immediately after we finish addition. The big concern is that these samples can affect the placement of other (good) samples and shouldn't.
Long Placement Attraction? There's a PhD chapter in that when it's all a bit less full on.
I definitely think LPA is a thing. Would be a fun problem particularly when you consider the effects of correlated error.
I added this option in the latest commit. It should become available via conda once I create a new release. Thanks for the suggestion, Rob!
@AngieHinrichs could we add this filter to the public tree workflow?
Sure. Currently I'm filtering post-placement with matUtils extract --max-parsimony 20 --max-branch-length 30
. So just to confirm, I'll be adding --max-parsimony-per-sample 20
to usher, and then just matUtils extract --max-branch-length 30
.
This option is available in the latest commits. Thanks @yatisht !
Currently there's an option to not place uncertain samples:
My request (as always, I know everyone is busy, I am probably the only person who wants this option, and I am very happy to be completely ignored) is for something that won't place really rubbish samples that are obviously garbage:
This would allow one to try and place many samples, but to ignore samples that were missed in prior filtering steps but that parsimony shows are obviously no good. E.g. those on terminal branches >100, which I sometimes see and then have to prune out later.