SATe fails to decompose small datasets

[Nam Nguyen]

Hi all,

I found a bug in SATe 2.2.5.  According to the paper, SATe will always perform at least 1 decomposition step, regardless of the size of the dataset.  However, when I was aligning datasets with 100 sequences and using a minimum alignment subset size of 200, SATe did not decompose the tree, rather it just ran Mafft on the entire dataset.  Reproducing this is easy, just set the alignment subset size to be greater than the number of sequences to be aligned.

Cheers,
Nam

[Jamie Oaks]

Hi Nam,

Thanks for your post. Mark and I discussed this issue and decided that the current behavior of SATe is preferred. The maximum subproblem size is used as a stopping criterion for the recursive tree decomposition. If the user specifies a value such that the stopping criterion is met at the outset, the expected behavior is for no recursion to occur. Otherwise, SATe will be doing something it was not "told" to do by the user.

If you want to have SATe perform a single round of decomposition, this behavior can be obtained by specifying the maximum subproblem size to be one less than the number of of taxa. This will force SATe to perform one, and only one, decomposition.

Thanks again,

Jamie