AFLP reliable for study of evolution?

[david baum]

The issue of homology assessment is, of course, present with AFLP data, but
no more so than in other data types.  And at least it is easy to assess:
by cutting-out some bands and sequencing them.  Instead, I would highlight
two other problems:

1) As with restriction fragment analysis (that doesn't take the trouble to
map the sites), there is likely to be non-independence between bands.  Thus
a band of length 300bp could be homologous to a band at 400bp without you
knowing it.  This length difference erroneously gets scored as two
presence-absence characters and is, hence, overweighted.  Still, if the
number of variable characters is high (and it can be much higher with AFLPs
than with RFLPs) then the occassional non-independent sites might get
swamped.
2) If, you are working close to the species level then you have a problem
interpreting the resultant tree.  Supposing you had sequenced a particular
gene or had done RFLP analysis on a given part of the genome.  In that case
the tree obtained would be an estimate of the phylogeny of that part of the
genome.  So even if this part of the genome has a different tree to other
parts of the genome (due to reticulation as occur via interbreeding within
populations, lineage sorting, and hybridization) then at least it has a
tree like history.  In the case of AFLPs (or RAPD), which are scattered
through the genome, there may be no underlying divergent tree to be
estimated.  In that case a standard "phylogenetic" analysis is really a
glorified "phenetic" analysis and I, for one, am unsure what the resultant
tree (phenogram) really means.  There may be ways to look at data (post
hoc) and see if it is representative of a real underlying tree.  But to my
knowledge these analyses have not yet been developed.

I have been a skeptic about using AFLP data for phylogenetics but, provided
one is confident that the underlying genomic relationships are divergent
not reticulate, maybe it has potential.

David


>Date:         Fri, 20 Aug 1999 22:02:35 -0500
>Reply-To: Lai-Lai Teo <lailaiteo at HOTMAIL.COM>
>Sender: Taxacom Discussion List <TAXACOM at USOBI.ORG>
>From: Lai-Lai Teo <lailaiteo at HOTMAIL.COM>
>Subject:      AFLP reliable for study of evolution?
>To: TAXACOM at USOBI.ORG
>
>I'm currently working on DNA-typing of plant.  In our laboratory, we use
>Amplified Fragment Length Polymorphism (AFLP), published by Vos et al.
>(1995).
>>From the literature and discussion with some friends, I found that the
>reliability of this technique is possibly questionable.  The main problem
>is the homology of the bands migrating at the same rate.  If the bands
>migrate at the same rate, they are assumed to have the same length and thus
>indicate similarity along evolutionary line.  I don't know much about
>evolution, but I have a few questions here.  Firstly, is two homologous DNA
>fragment necessarily the same product of evolution?  If selective force can
>produce homologous structures like the wings of butterfly and the wings of
>bird through convergent, I'm sure it can also do the same to DNA
>restriction fragment length from different evolutionary line.  Secondly, if
>the above question is not the main problem, are the bands that migrate at
>the same rate truly homologous?  All we can say for sure is that the first
>and last few nucleotides (primer sequence plus 3 selective nucleotides) are
>exactly the same.  how about those in the middle?
>I hope there's someone out there who can help me to clarify my doubt so
>that I can proceed with better condidence.  Your advise and recomendation
>will be much appreciated.
>





David Baum
Dept. Organismic and Evolutionary Biology
Harvard University Herbaria
22 Divinity Avenue
Cambridge MA 02138

Tel: (617)496-6744, -8766
Fax: (617)495-9484
dbaum at oeb.harvard.edu
http://www.herbaria.harvard.edu/~dbaum