Hi Phillip,

First, a 19 taxa data set is really small, and you don't need to use TNT New tech searches, that are strategies originally developed for extremelly large datasets, but with the final purpose of finding some shortest trees, but not all the shortest trees. You can simply do an implicit enumeration (that will find all the shortest trees), or several replications (e.g. 100) using the 'traditional searches'. Perhaps you have more than two most parsimonious trees.

Second, the "additional" politomies are not "created" by TNT. The program handles the trees uncollapsed during the searches (keeping the trees as fully dicotomic, even if some branches don't have characters supporting them). All the branches are collapsed in the tree when doing a consensus. Then, is most likely that the branches you lose (where the politomies are shown) are not supported for any character in your two trees.

You can collapse the branches after the search in New tech. menu, or in any of the searches menues (default option of the program is no collapse after searches).

Hope this helps.

Camilo

...may depend on the type of consensus tree chosen (e.g. strict, majority rule, Adams' consensus). In case of a strict consensus, the presence of one or a few badly defined taxa that are switching positions ('wildcard taxa') can result in wide-ranging polytomies where other types of consensus trees show a better resolution.

Hi Phillip,

First, a 19 taxa data set is really small, and you don't need to use TNT New tech searches, that are strategies originally developed for extremelly large datasets, but with the final purpose of finding some shortest trees, but not all the shortest trees. You can simply do an implicit enumeration (that will find all the shortest trees), or several replications (e.g. 100) using the 'traditional searches'. Perhaps you have more than two most parsimonious trees.

Second, the "additional" politomies are not "created" by TNT. The program handles the trees uncollapsed during the searches (keeping the trees as fully dicotomic, even if some branches don't have characters supporting them). All the branches are collapsed in the tree when doing a consensus. Then, is most likely that the branches you lose (where the politomies are shown) are not supported for any character in your two trees.

You can collapse the branches after the search in New tech. menu, or in any of the searches menues (default option of the program is no collapse after searches).

Hope this helps.

Camilo

19 taxa is small enough for branch-and-bound, a method that is mathematically guaranteed to find all MPTs. Don't use TNT for that, at least not with usual settings. There are spectacular failures of this in the published literature.

How many characters are there in your dataset?

I second Camilo's opinion. You HAVE to do an exact search. It will produce the same tree(s) than would produce PAUP's branch-and-bound or implicit enumeration algorithms (commands 'band' and 'alltrees', respectively). In TNT, an implicit enumeration can be done with the command 'ienum'. Also, be aware of the collapsing rules your are using (i.e. argument 'collapse'). Hope it helps.

Hi David,

We all agree that the size of the dataset certainly allows using a method that can get all the MPTs.

But I don't understand this phrase: "Don't use TNT for that, at least not with usual settings. There are spectacular failures of this in the published literature." Can you explain a little more?

Cheers,

Camilo

It's possible that there are ways to do this right in TNT, I don't know; I'm not sure what exactly implicit enumeration is. But a NT search is likely to fail at finding all MPTs for such a small matrix. I'll have to look up examples (I had misremembered the one I had in mind; it used the parsimony ratchet, not NT, and failed to find any trees less than 35 steps longer than the MPTs). Not being familiar with TNT myself, I would use branch-and-bound in PAUP*; it would take a few minutes, at most half an hour if you have a very complex matrix. NT is designed for very large matrices.

David is right, 'alltrees' in PAUP* does an explicit enumeration.

David,

To do an exact solution (i.e. getting all the MPTs trees through a branch and bound algorithm) on TNT is as easy as using only a single command (ienum) in the coomand line or going to /Analyze/Implicit enumeration in the menues (Windows version). As said before, New Tech search estrategies were developed for other kind of problems (big matrices or extremely noisy ones), and are not recommended for small to medium sized matrices. Implicit enumeration -branch and bound- for small ones, or traditional searches -multiple replications of Wagner's trees + TBR or SPR- when you have more than 25 terminals, are much better options. Parsimony ratchet is included in the New Tech searches in TNT (there are four strategies: drifting, ratchet, sectorial searches, and tree-fusing), and of course, any algorithm if is not used wisely, could produce a suboptimal result (valid for any kind of software). The "failure" is not of the program, but of the users in most cases. If you have the reference of the work you talked about, please share it, and perhaps I can point you out where the mistake was made.

Cheers,

Camilo

Sorry, I didn't see this for half a year. It's clear where the mistakes were made – a sloppy NT search was done on a small matrix with not enough TBR. References and some more details are in my 2018 preprint in the parts on phylogenetics.