IUBio Biosequences .. Software .. Molbio soft .. Network News .. FTP


Wolfgang Wuster bss166 at bangor.ac.uk
Fri Aug 25 04:38:36 EST 1995

On Fri, 25 Aug 1995, Mark Siddall wrote:

> In article <41hlip$du1 at mserv1.dl.ac.uk> Javier Quintero Vazquez <bnjquint at usc.es> writes:
> >
> >Greetings,
> >
> >Can anyone tell me:
> >
> >a.-In your opinion, the transition/transversion rate that would be used 
> >in DNADIST program to the Cytochrome B sequence of fishes.
> Ummm..... in the first place I would argue that you should not use distances 
> as they violate triangle inequalities and assume constancy.  Second, 
> estimates of TV:TS are usually between 1:2 and 1:10 I think.  If you
> don't have a good way of telling... do you really think you should just pull a 
> ratio out of the sky?

Try to find out the ranges of TS:TV ratios you have in your data. Also,
try plotting the pairwise TS:TV ratio between your terminal taxa against
the total %age base pair difference, that will tell you whether you have a
problem of transition sites becoming saturated. If there is no saturation,
then there is no real basis for weighting one type of substitution over
another. In my reptile cytb sequences, saturation usually started to
become obvious at pairwise divergence levels of 13-14%. 

> >b.-Previously, can I expect many differences between: a tree building with
> >distances based methods/a tree building with parsimony/cladistic methods.
> Yes.
> The distance-generated tree will have _no_ logical basis, whereas
> the parsimony-generated tree will have.  In the distance tree you will not
> be able to reconstruct evolutionary changes whereas in the parsimony
> tree you will.

Depends what you are looking for. If you are interested in the branching
order of your tree, but not character state transformations, then a number
of different algorithms are roughly equally likely (i.e., not very) to
give you the correct tree - try neighbour-joining, Fitch-Margoliash,
maximum likelihood, max. pars. Simulation experiments give no real reason
to prefer one over the other. Those clades which are supported by all
methods are probably the ones you can have most confidence in, since they
have then been shown not to be dependent on the assumptions of the
algorithm used.

Hope this helps.
Wolfgang Wuster
School of Biological Sciences, University of Wales, Bangor, UK
e-mail: bss166 at bangor.ac.uk

Thought for the day: If you see a light at the end of the tunnel,
it is probably a train coming your way.

More information about the Mol-evol mailing list

Send comments to us at biosci-help [At] net.bio.net