The main focus of our research group is to develop computer models of
very large biological molecules and parts of cells. Although much
knowledge has been gained by performing conventional laboratory experiments,
they have limitations that computer models, or computer
experiments, have the potential to overcome. An ideal computer model
will allow one to see and understand what is happening in the cell even at the
atomic level, it will be easy to set up and perform (perhaps with a few mouse
clicks), and it will be very flexible, allowing the researcher to test
hypotheses that are inaccessible to laboratory experiments. Also, the ideal
computer model will useful for suggesting new laboratory experiments, and will
be useful as a guide for making changes to the physical system being studied. In
the years to come, we hope to develop our own computer models to the point where
we can suggest changes to be made to malfunctioning cells and molecules in order
to stop diseases.
Unfortunately, such computer models do not yet exist for anything larger than
small molecules. For somewhat larger molecules, such as enzymes, simple computer
models are starting to be useful for designing new drugs. We want to move beyond
small molecules and enzymes. Specifically, we are currently interested in the
following physical systems:
Present computer models are limited in two ways. First of all, biological
molecules have thousands of atoms, so the amount of information that needs to be
processed by the computer is very large. Also, simulations must be run for very
long times to see interesting things happen. For example, computer simulations
of what a typical enzyme molecule does in a billionth of a second take weeks of
time on the largest supercomputers here at the San Diego Supercomputer Center.
Most of the molecules in the systems mentioned above are larger, and one will
need to simulate up to a hundredth of a second to get any useful information.
So, the only way to accomplish this is to take shortcuts. For
example, we don't really need to know what every single atom is doing in the
molecules and the surrounding water, but we do need to know how the shapes and
positions of the molecules change as they carry out their duties. Much of our
research is devoted to developing models that keep only the important
information, and by using such shortcuts we hope to reduce supercomputer-sized
problems to PC-sized problems.
Active research is being carried out on three short-cut methods in
particular:
- Hierarchical Collective Motions Method
- Faster Computations for Intermolecular Forces
- Weighted-Ensemble Methods
Another issue that computer modelers face is
that of software complexity. Most current computer models of
molecules are locked up in big, "black-box" computer programs that are very
difficult to modify or combine with one another. Thus, new ideas for better
models take years to make it into publicly available software, and to become
available on state-of-the-art multiple-processor supercomputers. To get around
this problem and make our research and that of others easier, we are developing
molecular modeling software and methods of writing such software, based on object-oriented
and generic programming techniques.
At present, our research is focused on the algorithm and software
development, with the long-term aim of successfully modeling the
biological systems mentioned above and further understanding their function. We
also aim to make our algorithms and software accessible to the larger scientific
community and general enough to tackle a wider variety of systems beyond what we
will explore in our lab.
