Hello --
I would like to use the output of a TREE-PUZZLE analysis of nucleotide
data under the Tamura-Nei (TN) evolutionary model as input for the
general time-reversible 6-substitution type (6st) model in PAUP*; i.e.,
an equivialently rate-parameterized TN model. I'm a bit unsure as to
which values to use in the PAUP* "rate substitution" matrix.
TREE-PUZZLE outputs estimates of the transition/transversion (ti/tv)
ratio parameter and the pyrimidine/purine (Y/R) transition ratio
parameter. These are distinct from the *expected* ti/tv and Y/R ratios
which are also output. The parameter estimates are used to compute
values in the instantaneous rate matrix (Q matrix) of the TN model. In
the TREE-PUZZLE documentation, the elements of the "rate matrix" are
defined as the product of the Q matrix and the base frequency matrix
(base frequencies as diagonal elements and zero otherwise). In other
words, each element of the Q matrix is multiplied by the frequency of
the base for the corresponding row. Because base frequencies differ,
the resulting "rate matrix" is asymmetric, corresponding to a 12st
model. This can be converted to a 6st model by averaging cross-diagonal
elements: (a_ij + a_ji)/2.
The resultuing set of values can then serve as input to the 6st "rate
matrix" of PAUP*. The problem as I see it is that here the base
frequencies are already part of the rate matrix, which makes me wonder
what PAUP* is doing if one selects "use empirical frequencies," or is
this ignored under the general time-reversible model?
--
Sean Turner, Ph.D.
NCBI (GenBank)
National Library of Medicine
NIH, Building 45, Room 6AN.12 phone: 301-435-8943
45 Center Drive, MSC 6510 fax: 301-480-2918
Bethesda, MD 20892-6510, USA e-mail: turner at ncbi.nlm.nih.gov
---
