My knowledge on this is probably quite out of date by now,
but some years ago there was a lot of research on this topic
because such surfaces are important in electrostatics and
implicit solvation models (calculating surface area) as well
as molecular graphics.
I think the most widely-used definition of a
solvent-accessible surface is Lee-Richards surface in which
a solvent-sized sphere is rolled along the surface of the
protein. Surface is therefore rigorously defined as a
piecewise collection of convex and concave patches of
spheres and tori. It was Connolly who implemented (and sold)
a practical algorithm for computing these surfaces. They
were even known as Connolly surfaces and rendered as dots
before modern computing hardware allowed for rendering
surfaces. Several groups have developed high-efficiency
versions of the calculation. Harold Scheraga's group, for
example, has some FORTRAN code for this. Fred Brook's
virtual reality group also developed a high-effeciency
parallel version (Varshney was the guy's name I think) in C.
There have been many approximations over the years I think
... but you asked about analytical models.
The these algorithms are non trivial. That's a
understatement. And there is actually a mathematical
ambiguity in the surface definition itself.
The Varshney code is freely available ... I received email
permission from both Varshney and his thesis advisor to
freely distribute the code. I even offered it to Warren
Delano years ago when he was writing Pymol, but he refused
to include it because he felt there still might be legal
issues that would effect Pymol. So ... Pymol contains only a
somewhat improvised an non-rigorous surface algorithm (last
time I looked). Fine for graphics of course.
en.wikipedia.org/wiki/Accessible_surface_area
Richard
On Jan 13, 2011, at 1:00 AM, Francois Berenger wrote:
Hello,
Does someone know some good articles
on this particular topic?
I'd like to implement the thing
myself, however if there is
a good software doing the job (with
readable source code),
I might use and cite it.
Best regards,
Francois.