How can one improve a phylogenetic tree? If one was to make the 'golden standard' of phylogenetic trees, what features would it have? Would it be highly reliable, well resolved etc? What other features are important to aim to maximise when it comes to improving phylogenetic trees?

  • 5
    $\begingroup$ Can you please merge your two questions into one as these are highly connected? Addtionally you may want to work on your attitude. No one here is obliged to answer, so you probably want to cut down the use of ASAP. Additionally: How does your own research look like, what have you found out? $\endgroup$ – Chris Apr 12 '15 at 10:41
  • 1
    $\begingroup$ I've found that a phylogenetic tree is an estimate of the true evolutionary relationships between species.Thus, we cannot really measure a tree's accuracy because we don't know the 'true value' i.e.what the actual pattern of evolutionary relationships is.I'm aware that trees which are well resolved,constructed using multiple genes & with high reliability/repeatability(indicated by high bootstrap values)are considered to be better than others but I'd like some references to back up my inferences since I'm doing a final year project.An objective list of what makes a good phylogeny may not exist $\endgroup$ – Biomad123 Apr 12 '15 at 11:08

A 'good' phylogenetic tree would be completely refined, properly binary, have high bootstrap support for all its branches, and reconstruct clades we think exist with high support values.

Generally high support values(there are other methods besides bootstrapping but that is a question for another time) is a measure of a good tree. There are other concerns, for example a tree that has no felsenstein zones is probably better than one that has lots.

We can know a lot more about trees than you suppose, though. Through simulation studies we can create datasets with a known phylogenetic history and then see how well the methods reconstruct the history. There are still problems(errors in the model used, just random error, ILS, HGT, other things) but mostly that kind of study tells us what good phylogenies look like.

It's very easy to generate a lot of different phylogenies. Using different methods, different models or approximation schemes, different genes or sequences from each species, or simply bootstrapping "new" samples from your existing dataset. A phylogeneticist usually has dozens of trees to choose from, which is a sort of crude bootstrap method itself. If trees look similar, they're probably pretty good. If they don't, you have a problem.

  • $\begingroup$ Would you please elaborate on what it means for a tree to be 'completely refined'? I am interested, but unfamiliar with what you're referring to here. $\endgroup$ – Galen Dec 1 '17 at 3:46
  • 1
    $\begingroup$ Some methods represent uncertainty by producing internal nodes that have more than three connections, essentially saying "all these groups are about the same amount similar, I can't figure out their arrangement". Finding an arrangement that has fewer of these internal nodes contains more information, and is called a 'refinement' of the tree. The 'star tree' where everything is connected equally is the least refined tree, and contains no information. The 'fully refined' tree is properly binary and has the maximum information. $\endgroup$ – Resonating Dec 1 '17 at 19:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.