Department of Chemistry, University of Minnesota
Image Galleries
Linear-scaling quantum gallery
Biochemical reaction gallery
Numerical density-functional theory gallery
The York Group
research
involves the development of a broad range of theoretical methods aimed,
ultimately, at providing the computational biology community with
greatly improved tools for the simulation of biological macromolecules
in solution. The main application focus of our group is on an
extremely interesting and challenging area of biology: the study of the molecular mechanisms of
RNA catalysis. This area is immensely important, not only
from a fundamental biological perspective, but also for the design of
medical therapies that target genetic disorders, and new biotechnology
such as allosteric RNA chips.
Currently, the lab is concentrated on the study of phosphate hydrolysis
reactions in solution, and catalyzed by two experimentally well-studied
prototype RNA enzymes (i.e., ribozymes): the hammerhead ribozyme, and the hairpin ribozyme. Despite the
tremendous amount of experimental effort, a detailed account of the
molecular mechanisms remains a topic of considerable debate. Many
of the structural studies of the hammerhead ribozyme have been pioneered
by Prof. Bill Scott.
These and other key experimental studies are vital compliments to the
theoretical quantum mechanical and molecular simulation studies in our
group.
From a theoretical perspective, ribozymes presents a tremendous
challenge relative to most protein enzyme systems. RNA is
inherently highly charged and interacts strongly with monovalent and
divalent metal ions and solvent. Careful treatment of long-range
electrostatics is critical, as is the inclusion and equilibration of a
sufficiently extensive solvation and ion atmosphere. The
importance of quantum many-body effects such as polarization and charge
transfer, which are neglected in conventional molecular mechanical force
fields, is greatly amplified in RNA simulations. RNA exhibits a
greater degree of conformational variation than most proteins, and
experimental structural data is often more difficult to acquire,
especially in solution. Reliable molecular simulations of RNA,
therefore, need to take into account extensive conformational
sampling. Finally, the chemistry of many RNA-catalyzed reactions,
such as phosphate hydrolysis, involves second row phosphorus atoms that
require an accurate quantum model (including d-orbitals) for a proper
description.
The design of new theoretical methods for RNA catalysis in the
York Group
include:
Extension of theory to macromolecular problems requires computational
methods that are efficient and have well behaved scaling properties.
Conventional density-functional theory (DFT) and Hartree-Fock methods
are limited to fairly small systems since the computational effort
scales as the cubic of the number of atoms (or higher), making
application of these methods to macromolecular systems unfeasible. The
main scaling bottlenecks arise from (1) the order N3
problem of maintaining orthogonality of molecular orbitals (or
equivalently the idempotency condition of the single-particle density
matrix) in accord with the Pauli exclusion principle, and (2) the order N2
problem of calculating long-range Coulomb interactions. The latter
problem can be surmounted with fast-multipole techniques or
linear-scaling Ewald methods that are also applicable in other important
areas of computational chemistry such as molecular simulation and
implicit solvation methods. A focus of the lab is to develop new
"linear-scaling" electronic structure methods that are able to treat
very large systems, and to apply them to exciting new areas such as the
study of quantum mechanical effects on biological macromolecules in
solution.
Recently it has become possible to
perform fully self consistent electronic structure calculations of
biological macromolecules up to 10,000 atoms in solution at the
semiempirical level. It is a primary focus of the lab to extend these
techniques to ab initio (Hartree-Fock and density-functional)
methods, and apply them to important biological and pharmaceutical
problems such as the elucidation of quantum mechanical electrostatic
potential surfaces and reactivity indexes, pKa shifts, density of
electronic states, and the role of solute polarization in the process of
solvation and ligand binding. More complex problems such as enzyme
catalysis and long-range electron transfer events in biological
macromolecules are areas of intense interest, and complimented by the
molecular simulation and hybrid QM/MM component of the group's research.
In order to develop quantum models that are both highly accurate and sufficiently fast for application of linear-scaling and hybrid QM/MM calculations of catalytic RNA systems, a large database of biological phosphates, phosphoranes and phosphoryl transfer reactions was constructed. The database of Quantum Calculations for RNA (QCRNA), has recently come on-line and is open to the public. The database contains over 1,500 molecular structures and complexes and over 200 reaction mechanisms, all calculated with a strict, consistent high-level density-functional protocol. Construction of the database was a tremendous undertaking and an essential step in the development of robust, highly accurate quantum models for RNA catalysis. The model structures, complexes and reactions in the database are of considerable biological importance and interest by themselves, and to date, publications have resulted on the structure and stability of biological phosphates and phosphoranes including binding with divalent Mg(II) ions, pseudorotation barriers of chemically modified phosphoranes, and in-line attack mechanisms of phosphate hydrolysis and thio effects.
Molecular simulation is a powerful tool
for examining the dynamic behavior of molecular systems and chemical
processes. Unfortunately, the computational requirement inherent in even
the most efficient ab initio methods precludes their application
to simulations of very large systems in the near future. For these
systems, approximate methods that are less computationally demanding are
required. An area of active research involves the development of
improved physical models that reliably describe molecular interactions
with minimal parameters and computational overhead.
Recently, a new model for polarization
and charge transfer has been introduced that utilizes the electron
density as the basic variable and takes into account many-body effects.
The generalized chemical potential equalization (CPE) method is a
density-functional based approach for inclusion of many-body
polarization effects in molecular simulations. The method incorporates
quantum chemical properties such as electronegativity and hardness based
on their mathematical definitions founded in density-functional theory,
and requires minimal overhead relative to conventional force field
methods.
A long term objective of the lab is to
develop a new generation force field for molecular simulations of
biomolecules that includes many-body effects and can be rigorously
combined with higher level ab initio methods to produce hybrid
potentials (see below). The force field will involve modification of
standard electrostatic point charge terms to include nuclear terms and
smooth electron density distributions. These distributions can adjust
(polarize) according to the CPE equations to allow a dynamical
description of the charge density as a function of conformation and
molecular environment.
Applications involving complex chemical
reactions ultimately require the rigor afforded by high level
first-principle methods. Fortunately, regions where chemical bond
cleavage and formation occur, such as the active site of an enzyme,
typically account for only a relatively small portion of the total
system. An attractive strategy for attacking very large problems
involves combining quantum mechanical (QM) and molecular mechanical (MM)
models.
Unfortunately, the computational cost
of sufficiently accurate ab initio
or density-functional methods often preclude their application to
complex mechanistic paths that require extensive configurational
sampling. More efficient quantum models, such a semiempirical
models, are sufficiently fast; however, there currently exists no
semiempirical methods that are able to reliably model phosphate
hydrolysis reactions. To remedy this problem, we have constructed
an extensive density-functional quantum database for phosphate
hydrolysis reactions using small molecule models. We have built a
program for non-linear semiempirical parameter optimization against the
database in order to derive reliable semiempirical parameters for
phosphate hydrolysis reactions that can be applied in hybrid QM/MM
calculations to the important ribozyme systems.
Future work involves the design of
new hybrid QM/MM potentials that are based on the chemical potential
equalization and linear-scaling electronic structure methods discussed
previously. Such a method had great potential to overcome many of
the difficulties encountered with conventional QM/MM methods, and is
particularly well suited for systems where the quantum mechanical region
electronically polarizes surrounding residues, or where charge transfer
between the quantum and molecular mechanical regions occurs.