I agree that determining possibly different gene trees is of primary interests here as well. And for visualising the differences the software is very good.

If the trees differ (and if this result represents the truth) then evolutionary congruence

does not hold and does not have to be tested by other means of course.

But even if the tree topologies are identical for different genes, bad things can

happen if the evolutionary models differ. It could even be that the disagreement between the models leads to a wrong topology if two or more genes are

pooled in one data set. (See e.g. the heterotachy problem for different partitions.)

The best thing would be to do both, gene trees and partitioning when analysing

the complete data set. PartitionFinder will help to join genes that can be joined.

Its more work though.

My impression was that Arif asked for a test which determines whether it is OK to pool genes to one data set. This is what the IDL test was designed for. When going from MP to ML I would go from IDL to PartitionFinder. Arif actually mentions "congruence of the constructed trees", but that is not what the IDL test does.

I assume this is the underlying difference in what software to use.

Congruence fo trees -> A Congruence Index for Testing Topological Similarity between Trees

Congruence of molecular evolution - tree and model -> PartitionFinder

PAUP* can do these tests but it's not free.

Therefore I'd recommend TNT which is free: you can download it here http://www.zmuc.dk/public/phylogeny/tnt/

A script by Mark Siddall to perform an ILD test with TNT is here: http://tnt.insectmuseum.org/index.php/Scripts/ild.run

thanx a lot, Ross.

This is really helpful. I will certainly try this. ..:)

I know it is in Paup where it is called hompart. It is based on the parsimony method and has been found to be not very powerful. I site from "When Does the Incongruence Length Difference Test Fail?"

Pierre Darlu and Guillaume Lecointre, Molecular Biology and Evolution, 2002

" We conclude that the ILD test has only limited power to detect incongruence caused by differences in the evolutionary conditions or in the tree topology, except when numerous characters are present and the substitution rate is homogeneous from site to site."

The Templeton Test can be done with Paup but requires that you evaluate the data by your self or write

a script to do this.

If your have molecular data, I would go for the ML method.

Thanks a lot, Christoph...

Actually I have already decided to go with the ML method. But, I have molecular data of several genomic regions. So, I have to give some statistical proof to justify their congruence. What is your suggestion to do that ?

Hi Arif,

To my knowledge, the best (old) way to run an ILD test is to use PAUP. You can have a look on this "how to do" from the Farrel lab: http://insects.oeb.harvard.edu/farrell_lab/techniques/pa_homogeneity.html or on the PAUP FAQ http://paup.csit.fsu.edu/data/MP_analysis.nex (but, unfortunately, PAUP server doesn't seem thttp://mbe.oxfordjournals.org/content/19/4/432.fullo work today!).

However, keep in mind that the ILD test is a very conservative test. See this 2002 paper: http://mbe.oxfordjournals.org/content/19/4/432.full and use and comparison with other tests in my 2006 paper: http://www.sciencedirect.com/science/article/pii/S105579030600128X

If you use the MP method, TNT is the best choice.

Its also much faster than Paup. Im am not sure it can do an IDL test.

If you use ML, I would partition the data set anyway. Only if the partitions are too small this can lead to a problem. But in this case, all tests will fail to determine whether they have the same

history.

Hi Arif,

"But, I have molecular data of several genomic regions. So, I have to give some statistical proof to justify their congruence."

You can always do a SH test to check if the data from any of your genomic regions gives you "significantly" worse trees than the best tree (which you could set a priori or get from the data it self). You can find an example in http://ijs.sgmjournals.org/content/57/2/276.short , but this was made with PAUP.

It is also possible to use a novel congruence test of de Vienne

http://max2.ese.u-psud.fr/bases/upresa/pages/devienne/index.html

My personal favourite would be the Partition Finder software if you want to test the congruence of different genes. It will tell you which partitions can be pooled and which should be evaluated independently.

Hmm, I am not sure Topological Similarity is what Arif wants to know. This test would be carried out as follows: Compute trees for different partitions and determine the degree of "similarity". That does not determine evolutionary congruence.

Evolutionary models might differ among partitions in which case the data set

should not be analysed in one junk in an ML analysis. PartitionFinder is maybe the best tool we have to estimate evolutionary congruence among genes.

I agree that determining possibly different gene trees is of primary interests here as well. And for visualising the differences the software is very good.

If the trees differ (and if this result represents the truth) then evolutionary congruence

does not hold and does not have to be tested by other means of course.

But even if the tree topologies are identical for different genes, bad things can

happen if the evolutionary models differ. It could even be that the disagreement between the models leads to a wrong topology if two or more genes are

pooled in one data set. (See e.g. the heterotachy problem for different partitions.)

The best thing would be to do both, gene trees and partitioning when analysing

the complete data set. PartitionFinder will help to join genes that can be joined.

Its more work though.

My impression was that Arif asked for a test which determines whether it is OK to pool genes to one data set. This is what the IDL test was designed for. When going from MP to ML I would go from IDL to PartitionFinder. Arif actually mentions "congruence of the constructed trees", but that is not what the IDL test does.

I assume this is the underlying difference in what software to use.

Congruence fo trees -> A Congruence Index for Testing Topological Similarity between Trees

Congruence of molecular evolution - tree and model -> PartitionFinder