Is it possible to predict protein domains?

David Jones jones at
Thu Sep 26 15:01:06 EST 1996

Lluis Ribas (lluis at wrote:
: Cornelius Krasel (krasel at wrote:
: : Hello folks,
: : 
: : I wonder whether it might be possible to predict the domain structure
: : of a protein without having to resort to sequence comparison. Are there
: : preferred amino acids in the linkers between domains, or amino acids
: : which are generally avoided?
: : 
: : Something like this would definitely be useful for approaching several
: : problems -- has anybody ever tried it?
: : 
: : --Cornelius.

:   You are describing fold recognition. It has been tried, it has
:   been used, it has been suffered :-).
:   For more information check the CAME homepage in Austria, or the
:   EMBL homepage in Heidelberg, or look at the list in Pedro's 
:   homepage for other methods (threading, Hidden MArkov, etc.).
:   Does it work ?. Hell, yes!...sometimes.

Fold recognition isn't really a solution to the problem of finding
domains in sequences - it can help, but in general fold recognition is
hampered by the presence of domains, rather than being useful for
finding them. The first problem is that threading methods are generally not
selective enough to clearly recognize domains, the second problem is that
if the domains in question have novel folds then there is nothing to
recognize. Having said that, the new version of our threading package
does have new features for trying to identify domains. Does it work ?
Hell, yes! ... not as often as I'd like, though.

The problem of identifying domain boundaries in sequences ab initio is
a "challenging" theoretical problem - it is also a rather fundemental
problem. I've been tinkering with solutions for ab initio domain
prediction for years, and although I've made some progress recently,
I don't see it producing a practical tool for a while yet.
The problem with domain linkers is that they are really just
non-descript pieces of polypeptide chain - rather than having particular
properties, they stand out by having very few properties. Domain
linkers obviously must not have (for example) strong helix forming
tendencies or be hydrophobic - they are just what's left after the
polypeptide chain has finished folding into separate domains. Imagine
rolling up a length of string starting at both ends - eventually
the two balls of string meet in the middle leaving a short length of
string in the middle - that's a domain linker - simply a bit of
polypeptide which wasn't invited to the folding party! Or perhaps
it left the party early...

This message was written, produced and executively directed by Dr David Jones
Address: Dept. of Biological       |   Email: jones at
Sciences, University of Warwick,   |     Tel: +44 1203 523729
Coventry CV4 7AL, U.K.             |     Fax: +44 1203 523568

More information about the Bio-soft mailing list