Supertree Or Consensus Tree?

Question

[Pavid]

Hey,

I previously asked something similar but I've been reading some things to make a good decision and I'm a bit confused. Perhaps silly but I'm relatively new to this field.

I'm working with M. tuberculosis strains, and I have different types of data:
- SNP (Single Nucleotide Polymorphism)
- MIRU-VNTR (Mycobacterial Interspersed Repetitive Units - Variable Number Tandem Repeat)
- Spoligotype
- RFLP (Restriction Fragment Length Polymorphism)

Some samples have been genotyped for all methods but most of them only for one or two. In the case the samples have all methods, sometimes some conflicts appears, one method says it belongs to family A, other to family B and other to family C.

What I was trying to do is to concatenate all trees to have a better look at the phylogeny.

First thing:
Does it make sense what I trying to do?

Then:
Which method would be more appropriate, supertree or a consensus tree is enough?
The problem with consensus tree is that I think I need to have the same samples in all the trees.

Any suggestion is welcome,
Thank you

Answer 1

[Vitis]

If you have a lot missing data in some types of your data, I'd do a very solid phylogenetic analyses based on the best type, for example, SNPs. Then I'd map other types as characters, basically, phylogenetic character mapping onto the phylogenetic hypotheses, to see how you SNP hypotheses got supported.

[Pavid]

I already have a individual phylogenetic tree for all the different methods. what do you mean with mapping?

[Vitis]

So the entire phylogeny reconstruction is about synapomorphy or homoplasy. So if you can use one of trees as a reference, and map your other data onto that tree, to the OTUs or nodes, you'll see how much the reference tree is supported by other types of data. For example, if certain nodes can be united by multiple types of data, those phylogenetic relationship will be well supported.

[Pavid]

so if I correctly understanding, that's basically the same thing that doing a consensus tree, no?

[Vitis]

you're right. But you can get more information about homoplasies. If your goal is a just a tree, this is essentially the consensus method (old school).

[Pavid]

thank you, you have been very helpful. I will read more about it, probably do a consensus tree now since I have samples from different sources and I'm not expecting much similarity, and I will have more perhaps try to do something similar with what you said

[Pavid]

I was trying to use phylip consense to build a consensus tree but it gaves me an error: name not found in first tree. In some trees I have more samples than in others. Do I need to remove the extra samples? If so, do you any other good program to do this? Thank you

[Vitis]

I kinda left the area for a while. Last time, I was using PAUP for consensus trees. Although PAUP is excellent for solid phylogenetic analyses, I had to say PAUP is not very user-friendly and I wouldn't recommend it. Mesquite is also very good, maybe you can give that a try. I think it's more flexible and easier to use. But I don't know whether it also requires every input tree has the same number of OTUs, which your data seems not to be.

[Josh Herr]

You can allow for missing data in your matrix to construct your consensus tree (which can be done in PAUP) and then should be able to map the data type across using MESQUITE.

I've never tried what you are planning but I mapped the consistency of morphological traits across a consensus phylogeny from numerous genes to see which morphological traits supported the "accepted" phylogeny. I used PAUP and MacCLADE (which is now MESQUITE) to do this and it worked well for me.