Protein loops

Chris Pook cjp at
Tue Jul 20 10:31:10 EST 1993


Has anybody ever tried to fit a new loop onto an existing protein by
searching through coordinate databases for a bestfit? What I really need
is some software that can do this, taking into account not only the
anchor points for the loop but also the accessibility of the side chains
(ie. its going to be exposed so I don't want any buried loops) and the fit
of the loop against the core of the host protein.

I have tried composer, and the version of composer in Sybyl, and they work
fine when it comes to dealing with backbone coordinates but they ask you
to specify what the sequence should be. I suppose that these programmes
are designed for modelling structures of known sequences by homology, but
it doesn't make much sense in my case. If one assumes that certain amino
acids have a propensity for a particular structure (eg a beta turn) then
it is difficult to justify fitting an invented sequence onto backbone
coordinates from another protein. One might as well use poly-alanine!
Having got the backbone the programmes the build the side chains using
ideal geometries.

The result is a loop which fits well enough between the anchor points but
which has no real identity of its own. Furthermore, the side chains are
likely to make bad contacts with the core of the host protein that need to
be alleviated (by mutating the residues?) to avoid disrupting its
structure. Thus the loop becomes even less related to its origins, which I
don't want since I would like to be able to use a real peptide fragment.

I suppose what is really needed is some programme that will look at the
host protein, allow you to select the anchor points and will then in some
way map out the surface between those two points which the loop will
cover, eg VDW surface, H bond donors/acceptors etc. It can then go away
and search for fragments that can best fit onto the surface, with varying
degrees of stringency, but, at the very least, without making bad contacts
with the core. Of course, there may not be any hits, in which case you are
back to square one!

I guess the whole problem at the end of the day is all to do with the
relationship between sequence and structure - how many structures can a
specific sequence form, or how many sequences can fold into a given
structure. There is no clear answer for well-defined secondary structure,
such as helices or strands, so just how important is the sequence of a

Sorry for waffling on so much, but I be glad of any help concerning this
sort of "semi de novo" protein design


Chris Pook                                      cjp at
Dept of Biochemistry
Tennis Court Road
Cambridge University

More information about the Bio-soft mailing list