pseudo-rate matrices valid?
Kevin Karplus
karplus at cheep.cse.ucsc.edu
Wed Apr 21 22:36:36 EST 2004
In article <284b17cb.0404181352.1be8f29c at posting.google.com>,
Andreas Bernauer wrote:
> I've just computed rate matrices and figured that some off-diagonal
> entries are negative (pseudo rate matrix). I am wondering if anybody
> knows, how accepted those matrices are, say for scientifc publication.
> I've searched the web for input on this, but without success on how
> accepted pseudo rate matrices are.
>
> Background: I estimated Q in P = exp(tQ) for a model of amino acid
> replacement, where P is the transition matrix of a Markov model. For
> biologocial interpretation, Q's off-diagonals should be positive,
> otherwise one gets negative probabilities for small t.
I don't know how accepted such pseudo-rate matrices are, but I'd
probably reject a paper that offered them as a solution to an
optimization problem for amino acid replacement, since it is clear
that the optimization was done without the proper constraints.
Perhaps the first question you should ask is why you are getting the
negatives off-diagonal, and whether you can fix your optimization to
eliminate them.
If you can show biologically useful results (better than with standard
rate matrices), then biologists may accept your pseudo-rate matrices
without worrying about whether they make sense or not, but there is
probably a genuine rate matrix that is close to your pseudo-rate
matrix that would work just as well without producing negative
probabilities.
--
Kevin Karplus karplus at soe.ucsc.edu http://www.soe.ucsc.edu/~karplus
life member (LAB, Adventure Cycling, American Youth Hostels)
Effective Cycling Instructor #218-ck (lapsed)
Professor of Biomolecular Engineering, University of California, Santa Cruz
Undergraduate and Graduate Director, Bioinformatics
Affiliations for identification only.
More information about the Comp-bio
mailing list