Which Statistical Congruence Tests Should I Use Before Concatenating Multiple Phylogenetic Trees?

Question

[John]

What are some examples of statistical congruence tests that should be performed to demonstrate whether the trees obtained by different phylogenetic markers are significantly different, and whether it is valid to concatenate the sequences into a single alignment or not?

So far I have found Templeton’s "significantly less parsimonious test" which is implemented in PAUP and MacClade. I'd prefer something implemented in R, but I'm interested in other tests that people use and the best way to implement them.

Answer 1

[2184687-1231-83-]

Not entirely what you are asking, but you could have a look at Ktreedist, which will give you tree to tree distances. You can then use those distances to decide whether to concatenate them or not:

http://molevol.cmima.csic.es/castresana/Ktreedist/Ktreedist_documentation.html

Ktreedist calculate the minimum branch length distance (or K tree score) from one phylogenetic tree to another. The K tree score provides a measure of the difference in both topology and branch lengths between two trees after scaling one of them to have a global divergence as similar as possible to the other tree.

Answer 2

[David W]

The most widely used one it probably (still) the Incongruence Length Difference Test [doi: [10.1093/sysbio/44.4.570] (called the Partition Homogeneity Test in PAUP*). If you really wanted to, you could probably implement it in R (it's just estimating parsimony trees from the various partitions and the whole set) - but what ever the test does, it doesn't really test for incongruence [doi:[10.1080/10635150290102302]

Concaterpillar is another option - but I can't tell you much about that one.

Answer 3

[Botond Sipos]

If your aim is to eventually use Maximum Likelihood reconstruction than probably the best would be to use the test developed by Huelsenbeck and Bull:

• John P. Huelsenbeck and J. J. Bull: A Likelihood Ratio Test to Detect Conflicting Phylogenetic Signal - Syst Biol (1996) 45 (1): 92-98. doi:10.1093/sysbio/45.1.92

I am not aware of any "plug-and-play" implementation, but I think that you can use HyPhy to perform this test without doing any scripting.

Answer 4

[DG]

You can use concatepillar:

Concatepillar

Which uses RAxML and Hierarchical Likelihood ratio tests to determine what genes in a dataset are congruent and can be concatenated for phylogenetic analysis.

The author of that software has developed other congruence tests as well. See:

Let Them Fall Where They May: Congruence Analysis in Massive, Phylogenetically Messy Datasets

and

Evaluating Phylogenetic Congruence in the Post-Genomic Era