Remove taxa

Dirk Husmeier, Biomathematics and Statistics Scotland (BioSS)

After an MCMC simulation, taxa can be removed from the tree topologies by editing the topology strings. Assuming these topology strings are in Newick format, I am using the follwing algorithm.

Hence note that in order do remove taxa, you need not repeat your PDM simulation; you only need to keep the intermediate result file resultsAllTopos.out, which actually has to be moved to file resultsAllTopos.bak.

The taxa-pruning algorithm has been implemented in Java class PruneTaxum. To apply it, proceed as follows:

Move resultsAllTopos.out to the following output file: resultsAllTopos.bak.
Create a new file called prune_taxum.in, which contains one integer number: the number of the taxum to be removed.
Give the command java PruneTaxum > resultsAllTopos.out (assuming you have compiled PruneTaxum.java with javac PruneTaxum.java. This command creates a new file resultsAllTopos.out, where the topology specified in prune_taxum.in has been removed.
Continue with the analysis as usual, for example by using class JambeAnalyseTopos, which reads in resultsAllTopos.out.

Shortcomings

The following two topologies result from pruning a tree of 8 taxa down to 4 taxa by subsequent application of PruneTaxum

((2,(3,4)),7)
((2,7),(3,4))

The strings are different, but the unrooted tree topologies are the same. Consequently, a topology change is erroneously indicated where, in fact, there is none. Now assume you want to carry out one more pruning step, and remove taxum 2. This gives:

((3,4),7)
(7,(3,4))

Now, even the rooted tree topologies are identical, but the strings are different. These examples suggest that a repeated application of PruneTaxum to remove several taxa may not be advisable.

I assume that this problem is less severe when a single taxum is removed from a tree with many taxa.

Back to the previous page.