"Kevin Karplus" <karplus from cheep.cse.ucsc.edu> wrote in message
news:mailman.146.1172616964.5139.mol-evol from net.bio.net...
>> On 2007-02-26, Ahmed Moustafa <ahmed from pobox.com> wrote:
>> Dear All,
>>>> Generally distance-based tree construction methods begin with multiple
>> sequence alignment for the protein sequences and than measuring the
>> distances between each two sequences in the alignment to fill in the
>> distance matrix.
>>>> My question is why not rather than performing the multiple sequence
>> alignment, simply use pairwise alignments scores to fill in the
>> distance matrix and from there to construct and infer the tree.
>>>> Thanks in advance!
>> The distance matrix can be computed from pairwise alignments, but
> computng n*(n-1)/2 pairwise alignments is often slower than
> constructing a multiple sequence alignment (depending on the MSA
> method) and the resulting distances are less accurate.
Kevin is right. The traditional reason is mainly because of the accuracy
and it applies to more or less all tree reconstruction methods. It should
be more accurate to estimate the distances or any other phylogenetic
information from the multiple alignment. The distances should also be
better behaved. When you compute from separate alignments, you can get odd
results e.g. from fragments that do not overlap much.
Of course, if you want to make the multiple alignment in the first place,
you usually do as you suggest (ironically). You make a quick and dirty
guide tree from separately computed alignments and distances.
> Kevin Karplus karplus from soe.ucsc.eduhttp://www.soe.ucsc.edu/~karplus> Professor of Biomolecular Engineering, University of California, Santa
> Undergraduate and Graduate Director, Bioinformatics
> (Senior member, IEEE) (Board of Directors & Chair of Education Committee,
> life member (LAB, Adventure Cycling, American Youth Hostels)
> Effective Cycling Instructor #218-ck (lapsed)
> Affiliations for identification only.