Is anyone familiar with using Bootstrap Consensus Tree in MEGA software?

A bootstrapped tree cannot have meaningful branch lengths because the bootstrap consensus tree is the topology of the most frequently appearing branch groupings.  For example if you have a branch with three taxa that has a bootstrap value of 50% the length of the branches and the length of the root of that branch can't really be calculated because half of the time one or more of those taxa is not even in that clade.

What is typically done for publication is to display the maximum likelihood (or Bayesian or whatever) best tree and then add the bootstrap values for the nodes above some cutoff (such as 75% or 80%) to this tree using Adobe Illustrator or similar graphics editing software after drawing the tree with FigTree or similar tree rendering software and exporting the tree as Postscript or other graphical image format.

MEGA can/does produce a tree with meaningful branch lengths and the bootstrap values at the nodes, if you ask it to.

Hi Brian, thank you for clarifying this to me.

Best regards.

MG

Hi Michela, the bootstrap consensus tree isn't particularly useful but you can actually show the bootstrap values for your best tree in MEGA. Choose the "Original tree" tab then click on the Tree Options icon (the tools just above). Click on the box to display statistics / frequency, choose the placement in relation to the node (trial and error to minimise clutter) and click on the box to hide support values below your chosen cutoff value (Brian's 75% suggestion is good). Because bootstrapping depends on character support for versus support against a clade all well supported groups in your bootstrap consensus will also be represented in your best tree.

Brian, I agree with your interpretation and advice but want to quibble with your first statement. It is possible to have branch lengths on a bootstrap consensus tree. You can optimise character changes and branch lengths on any topology, although this is complicated by polytomies, particularly in a partially collapsed (> 50%) bootstrap consensus. I guess your keyword is "meaningful" because of the difficulty in interpreting branch-lengths averaged across what might be quite different contributing topologies but this is not so different to separate optimization of branch-length and topology parameters by averaging across MCMC weighted tree sampling. 

Hi, your question and your answers were a lot of usefully for my work, I have one question:

if I want to show the relationship by the r16S, can I use the bootstrap tree? or I have to show the "most ML tree", It's only because I think that the bootstrap one is more organise for show the information in the paper.

Thanks for your attention and nice week. 

there are 2 trees( original and bootstrap) which one is useful to use in the publication? Thanks

Angie Perk, you should use the original tree which will have meaningful branch lengths, if the consensus tree is a cladogram with equal branch lengths. You can annotate the nodes that have high bootstrap values on the original tree. I am attaching a cladogram and a phylogenetic tree with meaningful branch lengths here so you can compare them. They are both from the same publication.

Brian Thomas Foley Thank you, I have constructed a phylogenetic tree from yeast sequences. I know that there are 2 different yeast species but they seem under the same branch , it does not make sense. I there is a way to make it correctly? Do you check the branches after creating the original tree? Thanks