Shared epitopes prediction?

[Fergus Doherty]

Perhaps someone can help me with a problem.  An antibody detects
procarboxypeptidase B in rat pancreas, but the antibody is to an unrelated
protein.  Comparing the sequence of the protein to which the ab was raised
(165 aa) versus procarboxypeptidase B (>400 aa) with the GCG BESTFIT
program shows 21% identity, 42% similarity.  21% identity is probably not
significant as the identical aa are spread over the sequence of the smaller
protein, there is no string of identical or non-identical aa that could be
a common epitope.  Is there another program that might do this comparison
better, ie look for a small (6-10) run of identical/near identical aa (that
might be the common epitope)?  Could it be an epitope that depends on some
secondary structure (retained or regained following SDS-PAGE and Western
blotting) so that the epitope consists of residues close in space but
non-adjacent?  What are the chances of this? If so is it possible to check
for this with a suitable algorithm?  Any help appreciated.

Fergus Doherty,
dept. Biochemistry,
University Medical School,
Queen's Medical Centre,
Nottingham NG7 2UH
Tel: 602 709366  FAX 602 422225 Internet: mbxfd at unicorn.nott.ac.uk