You could also try FASTS (FASTA with multiple short peptide sequences) at
http://fasta.bioch.virginia.edu/, but FASTS is really meant for finding
matches with more than one peptide at a time. With only one peptide it's
not much different from blast/sw/etc. Your main desire is going to be to
use a scoring matrix with high information content (bit score per
residue). FASTS uses a PAM-20 like matrix ... BLOSUM 80 is *not* good
enough for very short sequences. You should also set your gap penalties
to be fairly high (or very very high to disallow gaps completely). Both
of these things will maximize your chances of finding a significant hit.
Lastly, try searching a smaller database, i.e. if the proteins you're
interested in are mammalian, try searching only the mammalian subset of
nr, or the mammalian-only subset of swissprot to begin. Searching a
smaller database will improve the significance of any hits you find.
On 9 Aug 2001, Darek Kedra wrote:
> Dear All,
>> I've got short, 9-15 AA long peptide sequences from phage display.
> These are "random" sequences which happen to interact with a target.
> The next step is to find if something similar exist in nature.
> I wonder if you can recommend "the best" program/algoritm for doing
> I've used blast (blastp / nr database and tblastn / est db) but
> finding anything meaningful in 500+ weak matches is a tall order.
>> Thank you in advance.
>> Darek Kedra
>> d a r k e d at m y -d e j a . c o m
o ~ ~ ~ ~ ~ ~ o
/ Aaron J Mackey \
\ Dr. Pearson Laboratory /
\ University of Virginia \
/ (434) 924-2821 \
\ amackey at virginia.edu /
o ~ ~ ~ ~ ~ ~ o