Sometimes I have to put text on a path

Sunday, May 25, 2008

Computational Methods and Tools

Computational Biochemistry is Not a Set of Black Box Techniques

A good part of gas phase organic and inorganic chemistry can be described well using standard methods of quantum chemistry and statistical mechanics. There are two fundamental reasons why quantum chemistry and statistical mechanics are so successful in describing the chemistry and thermodynamics of molecules in the gas phase chemistry. First, the molecules of interest to the gas phase chemistry are typically small, and thus amenable to highly accurate quantum chemical calculations. Second, isolated molecules or molecular complex in the gas phase have only a few energetically accessible states, and statistical averaging over these states is often feasible. The situation is very different in biochemistry, where one typically studies heterogeneous polymers, such as proteins or nucleic acids, that interact with other molecules in aqueous environment via multitude of weak interactions. Such molecules and molecular complexes are often too large for quantitative description by even the crudest quantum chemical methods, and possess so many energetically accessible states that their complete enumeration is prohibitive. It is clear that research problems in computational biochemistry cannot be solved by simply applying "black box" techniques that have been so successful in characterizing the gas phase chemistry.

Quantum Chemistry of Biomolecules

PES One powerful research strategy to make computational analysis of biomacromolecules feasible involves the combination of reductionism and systems approach. In the first, reductionist stage, a necessary description of components of a biological macromolecule is obtained using quantum mechanics. Then, during the systems approach stage, molecular simulation techniques based on a simplified Hamiltonian are used to describe the structure, motions, and thermodynamics of the whole system. For example, quantum chemistry allows determination of charges, dispersion coefficients (van der Waals parameters), and barriers for torsional motion (flexibility) of components of macromolecules. Quantum chemistry also provides information about the structure, reactivity, and spectroscopic properties of biological molecules. Such information may be needed as input when applying molecular simulation techniques. In summary, quantum chemistry provides an excellent starting point in understanding many biochemical phenomena.

A large number of powerful quantum chemistry programs are available. Below is a list of some quantum chemistry programs that I find useful in my research:
  • Dalton
    • Molecular polarizabilities
    • Dispersion coefficients (van der Waals parameters)

  • DIRCCR12
    • Explicitly correlated calculations
    • Highly accurate energies

  • Gaussian
    • All the usual: geometries, isotope effects, conformational search, reaction paths, solvation
    • Trickier things: QM/MM via ONIOM, NMR properties, UV-vis

  • Gamess
    • Environment description via effective fragment potentials
    • Renormalized coupled cluster energies

  • Jaguar
    • Conformational analysis
    • Solvation free energies

  • SAPT
    • Intermolecular forces
    • Dispersion interactions

  • Molpro
    • Multireference systems
    • Prediction of UV-Vis spectra

  • Spartan
    • Electronic structures, conformational analysis
    • Visualization of molecular properties

Molecular Simulations

Description of the dynamic behavior and thermodynamic properties of biological macromolecules molecules in solution requires application of molecular simulation techniques. Simulation techniques such as molecular dynamics and Monte Carlo sampling are well established for description of physical properties of fluids. Molecular dynamics simulations have been employed also to describe dynamical structure of proteins, nucleic acids, and their complexes. Free energy simulations provide a theoretically rigorous way to evaluate binding free energies, opening up possibilities to predict efficacy and safety of potential drug candidates.

The field of molecular simulations is plagued by its own problems. For example, it is unlikely that molecular dynamics simulations with currently accessible timescales in the order of 10-100 ns can capture the full dynamics of biological macromolecules. Similarly, it is not clear if currently available molecular mechanics force fields and sampling methods are sufficient for reliable description of binding affinities in novel systems.

Despite these problems, molecular simulations have been proven valuable in understanding chemical reactivity in condensed media and appear promising in understanding the mechanism of enzyme action. Some of the molecular simulation programs that I have used are:
  • BOSS
    • Parameterization of OPLS-AA force field and conformational analysis
    • Monte Carlo simulations of pure liquids, calculation of free energies of solvation

    • Protein structure and dynamics
    • Correlation of atomic motions in enzymes

  • MCPro
    • Monte Carlo simulations of proteins
    • Free energies of binding of drug candidates

  • NAMD
    • Protein structure and dynamics
    • Conformational free energies

    • Reaction free energies via thermodynamic cycles
    • Parameterization of polarizable force fields

Molecular Visualization


Biological macromolecules are a lot more complex than typical organic molecules. The complexity arises from the large number of atoms in a biological macromolecule, and from the possibility of relatively free rotation around many covalent bonds in a macromolecule. The low rotational barriers give macromolecules flexibility and high conformational complexity. The number of theoretically possible three-dimensional structures that any macromolecule can take is enormous. However, each biological macromolecule adopts a distinct three-dimensional conformation, called the native conformation. Prediction of the native conformation of a unique biomacromolecule based on its covalent structure is a challenging problem, and most information about macromolecular structures is obtained from analysis of X-ray diffraction data from single crystals or from analysis on NMR data of dissolved macromolecules. Further analysis of these structures can be performed using molecular simulation techniques. For example, molecular dynamics simulations allow studying enzyme-substrate complexes, which cannot be studied experimentally due to rapid catalytic turnover under experimental conditions. Visualization tools are indispensable for analysis and presentation of macromolecular structures. Some that I find most useful in my research are:
  • PyMol
    • Raytraced publication-quality images and movies
    • Complex scenes thanks to a powerful command language

  • gOpenMol
    • Analysis of molecular dynamics trajectories
    • Visualization and analysis of macromolecular structures

    • Creation and editing of coordinate files
    • Analysis of outputs from quantum chemistry programs

  • VMD
    • Analysis of molecular dynamics trajectories
    • Interactive molecular dynamics

    • Visualization and analysis of macromolecular structures
    • Visualization of molecular properties from quantum chemistry programs

No comments:

Post a Comment