Invited
Speaker
"In Silico" Drug Design Based on the Modified Semi-Empirical
Quantum Chemical PM6 Method
Jindrich Fanfrlik, Agnieszka K. Bronowska, Jan Rezac, and
Pavel Hobza
Czech Republic
Ligand-protein docking is broadly used to predict the binding modes
and binding affinities of novel drug candidates and as such plays
an increasingly important role in rational drug discovery. Protein
docking produces a pose generation, i.e. a list of various positions
where the ligand can be bound. In the second step of the docking procedure,
the binding free energies are estimated by using scoring functions.
The design of the scoring functions represents without a doubt a critical
step in the whole process, where a deep understanding of noncovalent
interactions (1) plays a key role. Without knowledge of this function,
which should not only differentiate between binders and non-binders
but also rank the binders, successful "in silico" drug design
would not be possible. Most of the scoring functions are based on
empirical functions, also called molecular mechanics (MM). The use
of empirical functions is fully understandable, because the protein
targets are extended systems with several thousands of atoms. On the
other hand, the limitations of these functions are also well known;
among them, the inability to describe quantum effects becomes the
most serious. Because of the size of the systems being considered,
the nonempirical ab initio quantum mechanical (QM) approach
is not feasible, and also the use of the QM/MM approach is limited.
A compromise would be the use of semiempirical QM (SQM) techniques.
Considering the complexity of protein-ligand binding, it is understandable
why these methods have been used only rarely up to now. The way toward
their use has been paved only recently when the new and very efficient
SQM method appeared (PM6, Ref. 2) and when the first studies improving
its efficiency by adding the correction terms for dispersion and H-bonding
energies were published (3). The second generation of these corrections,
very recently introduced in our laboratory (4), improves the H-bonding
considerably, and the PM6-DH2 method provides excellent results (with
an almost chemical accuracy) for different types of noncovalent interactions.
This is no doubt the prerequisite for any further theoretical
activity in the field. In addition, due to the introduction
of the linear scaling procedure and of the reaction field (2), the
technique can be used for extended systems having several thousand
atoms in a water environment, which allows us to consider the whole
protein.
Analyzing the binding process, we realize that the second critical
step after binding is ligand relaxation and desolvation. The solvation
energy was evaluated using the advanced techniques of a self-consistent
reaction field (5), which is known to provide reliable energies for
not only neutral but also charged systems.
The total binding free energy of a protein-ligand complex is calculated
in the present scoring function as a sum of the binding free energy
of the complex and the relaxation and desolvation free energy of its
subsystems. All of these terms have been shown to be important, and
none of them can be neglected.
Our docking scheme was tested on the HIV-1 protease and bovine carbonic
anhydrase II (6). The results were compared to the structural crystallographic
data and the experimental binding data (ITC, SPR). The PM6-DH2 scoring
improved the docking results dramatically and reproduced the experimental
results correctly. Not only were the ten binders and ten non-binders
in our training-ligand sets correctly assigned, but also the proper
order of all the binders was achieved.
The PM6-DH2 scoring thus provides a novel, valuable and very promising
tool for rational drug discovery and de novo design.
1. P. Hobza, K. Müller-Dethlefs, Noncovalent Interactions.
The Royal Society of Chemistry, Cambridge, 2010.
2. J. J. P. Stewart, J. Mol. Model, 2007,
13 (12), 1173-1213.
3. J. Rezac, J. Fanfrlik, D. Salahub and P. Hobza, J.Chem.Theory Comput.,
2009, 5, 1749-1760.
4. M. Korth, M. Pitonak, J. Rezac and P.Hobza, J.Chem. Theory Comput.,
accepted.
5. Gaussian09.
6. Jindrich Fanfrlik, Agnieszka K. Bronowska, Jan Rezac, Jiong Ran
and Pavel Hobza, submitted.
|
