I have obtained distance matrices whose coefficient are related to binding specificity of different molecules. The distance matrices are suitable for the Phylip package and I can get nice trees. However, I would like to have confidence values for the trees. It seems to me that boostrap will do the job but I could not use the Phylip bootstrap program (seqboot) with a distance matrix. Any help to do this will be greatly appreciated.
Seqboot resamples the sequences to give a new set of distance values each time the data is resampled. So the equivalent with your data would be to re-measure the binding specificity many times, to get a better estimate of the accuracy of those values, and hence the trees built from them. As far as I know, there are not many other methods other than neighbor-joining to create a tree from a distance matrix. Maximum parsimony, maximum likelihood and Bayesian methods use a multiple sequence alignment. I am having difficulty understanding how a measurement like binding specificity should be expected to fit a phylogeny in the first place.
Brian, thanks for your response. We published a while ago a method to cluster MHC molecules according to predicted peptide binding repertoires and I wonder how to provide confidence to the trees depicting the clusters
http://www.ncbi.nlm.nih.gov/pubmed/18449999
Certainly the bootstrap is the only way I know to do find any kind of confidence on distance trees. Perhaps PAUP or phylum would work.
I wonder if the data is amenable to parsimony analysis which you could do either in Mesquite, or Phyml or PAUP. Could you send me the matrix so I could try it out?
- Badges
- Science method