MOLECULAR MODELING OF BIOMOLECULES
Institute of Macromolecular Chemistry, Academy of Sciences
of the Czech Republic, Heyrovského nám. 2, 162 06 Praha 6, Czech Republic
This contribution gives a short review of molecular
modeling methods suitable for proteins and their interactions with ligands. The
basic information about molecular modeling of biomolecules is possible to find
in books [1]-[3].
Molecular modeling of
biomolecules is used to complete our knowledge of a molecular system by
information which is not possible to receive from an experiment. The usual aim
of the modeling is either a prediction of unknown molecular structure (protein
folding, docking of a ligand to the protein active site) or complementation of
an experimental knowledge of a system (ionization of the protein active site,
interaction energy evaluation) or a time dependent simulation of a process
(HIV-1 protease flap opening).
Molecular modeling of
proteins has its specifics based on the fact that biomolecules are too large to
use quantum mechanical approach (even small proteins are composed of several
thousand of atoms). Use of classic mechanics is allowed because of usually non-covalent
character of interactions between proteins and its ligands. Specialized types
of force fields for proteins take advantage of composition of proteins - 20
standard types of amino acid residues forms building blocks of protein chains.
Two specialized tasks of molecular modeling of
biomolecules will be discussed in detail: Protein structure prediction and
structure based drug design.
Protein folding is a slow and complicated process
lasting seconds or minutes. Therefore a prediction of protein structure is not
usually based on dynamical simulation of folding process, but on comparison
with proteins with similar sequences and known structures. When the sequence of
the protein is very similar to the sequence of the protein with known
structure, then the structure of selected protein can be predicted by so called
homology modeling [4]. When similarity of sequences is not so high, at least
attempt to estimate a type of fold can be done [5], [6].
Structure
based drug design has the aim to reduce amount of chemical synthesis by
computational prediction of binding properties of non-synthesized compounds.
There are several types of computer based drug design tasks. When 3-dimensional
structure of receptor is known from X-ray crystallography or any computational
model of the structure is reliably predicted, new drug can be designed by so
called docking – by building the compound in the binding cavity of the rigid
receptor, testing sterical correspondence and interaction energy between the
ligand and receptor. Pharmacophore modeling is a method of drug design in the
case when 3-dimensional structure of receptor is not known. It is based on
determination of typical properties of the ligands with good affinity to the
receptor (typical groups and their distances), and then computational design of
new compounds fulfilling pharmacophore criteria.
When
lead compound is known and we want to improve it by slight modification, then
more detailed and time consuming analysis can be appropriate – analysis of
interaction energies between the receptor and ligand, molecular dynamics or
free energy perturbation.
[1]
W. F. van Gunsteren, P.
K. Weiner, A. J. Wilkinson: Computer simulation of biomolecular systems, volume
3, Kluwer Academic Publishers, 1997.
[2]
R. Leach: Molecular
modelling, Longman, 1999.
[3]
A. K. Rappe, C. J.
Casewit: Molecular mechanics across chemistry, 1997.
[4]
M. A. Marti-Renom, B.
Yerkovich, A. Sali, Modeling protein structure from its sequence, manuscript at http://salilab.org/modeller/user_manual.shtml),
Current Protocols in Bioinformatics, submitted.
[5]
http://speedy.embl-heidelberg.de/gtsp/index.html, www pages of Rob Russell, EMBL,
Meyerhofstrasse, 1, D-69117 Heidelberg, Germany.
[6]
http://www.expasy.org/swissmod/SWISS-MODEL.html, SWISS-MODEL web pages.