|
Protein Surfaces
and Solvent Accessibility:
Definition and Prediction
Jan Hansen, Ole Lund,
Henrik Nielsen, Hans Henrik Stærfeldt, and Søren
Brunak, Center for Biological Sequence Analysis, The Technical
University of
Denmark
The surface of a protein is non-trivial to define, even when the
structure is known. A widely used measure is the relative solvent
accessibility, which denotes how large a part of the van der
Waal's
surface of each amino acid residue is exposed to the solvent
surrounding the protein. Prediction of this quantity is important
for
predicting active sites, antibody epitopes, and
post-translational
modifications.
The value of the relative solvent accessibility, however, is
very dependent on parameters used for calculating it from the
three-dimensional structure, mainly the sphere size. I will show
accessibility distributions for various residues over a range of
sphere
sizes, and I will discuss problems such as multimeric proteins,
internal cavities, and incomplete sequences.
Prediction of solvent accessibility has been performed with
artificial
neural networks and statistical methods. I will describe a neural
network approach currently under development in our group, and
compare
it to previously published methods. In particular, I will discuss
the
question of how much information can be extracted from the local
sequence context, and the issue of single sequence versus profile
prediction.
|
