Recent progress in Computational Molecular Biophysics.

Jiri Sponer

J. Heyrovsky Institute of Physical Chemistry, Prague, Czech Republic

Jackson State Unversity, Jackson, MS, USA

1) Advance of high-level ab initio quantum-chemical calculations into biomolecular modeling (1994-1996).

• Advantages and limitations
• Future directions

2) Second-generation of empirical force fields for biomolecular modeling (AMBER, CHARMM...) (around 1995);

• Improvements and limitations.
• Future directions in developments of force fields for biomolecular modeling.

3) Molecular dynamics simulations of fully hydrated biopolymers on a nanosecond scale.

• Particle mesh-Ewald formalism (since 1995).

4) Advances in understanding of protein folding and protein fold prediction.

• Funnel theory vs. transition state theory
• Database potentials vs. molecular mechanics potentials.

High-level quantum-chemical calculations:

Such calculations were not available before 1994-1995 for larger systems.

Essential features

• Inclusion of electron-correlation effects
• Use of polarized basis sets of atomic orbitals
• In contrast to all other techniques, there are no empirical parameters.

Main advantages:

• Complete and physically correct description of the molecular interactions.
• Accuracy and reliability comparable to gas phase experiments, but much more flexible.

Main disadvantages

• Enormous computer requirements, limited to small systems only, "in vacuo" calculations. Ab initio calculations are, among other items, essential for a proper parametrization of empirical force fields for biomolecular modeling.
• essential for proper treatment of important local interaction, where high accuracy and physical correctness are necessary.

Future directions:

• studies on systems containing metal cations, since these cannot be studied by any other technique.
• Development of hybrid quantum mechanics
• classical mechanics techniques.
• treatment of solvent effects

Standard ab initio methods are not so difficult to use, much less tricks than in moledular modeling.

Examples of recent results:

i) First physically correct description of base stacking (Sponer et al., Biophys. J. 73, 1997, 76); nonplanarity of DNA base amino groups. (Sponer & Hobza, J. Am. Chem. Soc. 116, 1994, 709; Luisi et al., J. Mol. Biol. in press)

ii) High-level calculations on amino acids and small peptides. Models of active centers of proteins (H. Lee, J. Am. Chem. Soc. 118, 1996, 3946, A. Silva et al., J. Mol. Biol. 255, 1996, 321.)

iii) Studies of interactions between metal cations and various ligands. (Garmer & Gresh, J. Am. Chem. Soc. 116, 1994, 3556; Burda et al., J. Phys. Chem. B 101, 1997, 9670).

Strong warning with respect to all inexpensive methods (mainly semi-empirical, AM1, PM3)!!!

Density Functional Theory technique (DFT):

• In principle very efficient
• many different variants
• Lack of the systematic improvement characteristic for high-level ab initio method,
• Quality differs from one system to another. (poor prediction of glycine conformations, complete failure for base stacking).
• Aggressively advertised

Force field development

The second generation of force fields is based on the same analytical expressions as the previous force fields. However, the parametrization is much more careful, with inclusion of more experimental data and mainly with aid of quantum-chemistry. Cornell et al., J. Am. Chem. Soc. 1995, 117, 5157; MacKerrel et al., J. Am. Chem. Soc. 1995, 117, 11946

• Performance has been improved.
• Current force field are expected to reach the accuracy limits affordable for pair-additive potentials.

Main drawbacks:

• Constant atomic charges
• No real chemistry
• Atom- atom pair additivity

Future development of force fields - inclusion of induction term, it means nonadditive force field.

Nanosecond MD simulations Current improvement of computer hardware allows for the first time to make extensive (nanosecond) MD simulations on hydrated biopolymers. The particle mesh-Ewald technique is quite essential to provide stable trajectories.

Positive examples: B to A transition of DNA (Cheatham and Kollman, J. Mol. Biol. 259, 1996, 434)

Verification of the Py.Pu.Py triplex architecture (Shields, Laughton, Orozco, J. Am. Chem. Soc. 119, 1997, 1463)

i-DNA: three-dimensional structure with repulsive base stacking (Spackova, Berger, Egli, Sponer, J. Am. Chem. Soc. in press)

Migration of cations into DNA grooves (Young, Jayram, Beveridge, J. Am. Chem. Soc. 119, 1997, 59).

Protein simulations (Perera, Darden, Montoe, Pedersen, Biophys. J. 73, 1997, 1847)

Failures and problems: B to A transition is force field dependent (Feig & Pettitt, J. Phys. Chem. B 101, 1997, 7361)

Difficulties with RNA (Cheatham & Kollman, J. Am. Chem. Soc. 119, 1997, 4865)

limitations: There is no experience with simulations on a scale above 10 ns.

Protein folding and fold recognition

Two basic problems:

How proteins fold?

What is the correct structure for a given sequence, protein fold recognition.

Levinthal paradox: Random search would NEVER find the native state.

Funnel theory: Protein folding is a complex reaction that proceeds via a mechanistically determined pathway from unfold state to the native state.

New view: Transition state theory:

Unfold structure -> (intermediates) -> TRANSITION STATE -> native state.

There is no Levinthal paradox to reach the transition state!