>>>>> "Steve" == <HARDIES at thorin.uthscsa.edu> writes:
Steve> Joe Felsenstein writes (in response to a question by Tom
Steve> Doak):
>> By doing multiple runs with DNAML from PHYLIP, at each at a
>> different value of the transition/transversion ratio, and
>> picking the result that achieves the highest likelihood, and
>> also saving that tree, one will get a joint ML estimate of the
>> ratio and the tree. I think that is what you want.
Steve> Do you think it would be useful (when using the above
Steve> method) to obtain the transition/transversion ratio from a
Steve> subset of the most closely related sequences to suppress
Steve> noise caused by saturation of transitions at greater
Steve> divergence?
Joe can answer this more authoritatively than I, but in principle
there should be no need to restrict the method to a subset of the most
closely related sequences. DNAML includes a model of the
substitutional process that will correct for multiple substitutions.
In practice, the model has to assume the *same*
transition/transversion bias applies across the entire tree. Whether
that's a reasonable assumption will depend on the particular data set.
If you have enough data it would be possible, I think, to estimate the
ratio separately in two or more different sets of taxa and use a
likelihood ratio test to determine whether they are different. DNAML
also assumes that all sequence difference are a result of nucleotide
substitution, not indels. Thus, it should be used only on portions of
seqeuences where the alignment is good.
-- Kent
+------------------------------------------------------------------------+
| Kent E. Holsinger Internet: Kent at Darwin.EEB.UConn.Edu |
| Department of Ecology & Holsinge at UConnVM.UConn.Edu |
| Evolutionary Biology BITNET: Holsinge at UConnVM |
| University of Connecticut, U-43 |
| Storrs, CT 06269-3043 |
+------------------------------------------------------------------------+
