Running BUCKy on partitions rather than genes?

[Katie Everson]

Hi all,

I was wondering whether it is appropriate to run BUCKy on partitions (identified by PartitionFinder, for example) rather than genes.

I am using a dataset with 11 genes, but there are only 4 partitions in the optimal partitioning scheme identified by PartitionFinder. I have run MrBayes analyses on each individual gene, on each individual partition, and on the concatenated dataset.

Thanks!
Katie Everson

[Cécile Ané]

 Hi Katie,

PartitionFinder will tell you which partitions should be analyzed with different evolutionary models (like HKY versus GTR+Gamma etc.), but assuming that all partitions evolved under the same tree topology. In your case, I suggest that you use your 11 genes separately as 11 different loci in BUCKY, because BUCKy assumes that each locus is recombination-free (same tree topology for all sites within the locus), and that different loci are unlinked. These assumptions are probably reasonable for your 11 genes. But they are probably not met for your 4 partitions, as one or more partition is probably spanning 2 or more genes, with possible recombination within that partition. So the BUCKy analysis on genes would be best.

If PartitionFinder suggests splitting a gene into 3 partitions, for example based on codon position, then it is best to analyze this gene as a single locus, but let MrBayes use the 3 partitions at that locus (MrBayes assumes that the same tree topology is shared by all 3 partitions at that locus). Then later you can combine all the MrBayes results from a bunch of genes to run BUCKy.

I hope that answered your question!
Cecile.

--
You received this message because you are subscribed to the Google Groups "BUCKy users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bucky-users...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.