The Supercomputer as Laboratory

As the name implies, MD is a way of tracking how molecular structure changes over time. In their natural environment inside living cells, proteins are flexible and constantly in motion, almost vibrating -- changing from one picosecond (a trillionth of a second) to the next. MD computes the forces acting on each atom at each instantaneous change of time (using Newton's equations of motion), and gives detailed information about these infinitesimal movements. To get realistic results, it is important to consider not only the protein itself but also the water molecules that surround it and influence its shape. This means that thousands of atoms need to be included, and a single MD simulation can take hundreds of supercomputing hours.

As supercomputing hardware and software have evolved over the last few years, Brooks has begun using MD to simulate processes involved in how proteins unfold, which is believed to mirror the rules of folding. Because some of these events occur on such short time scales, laboratory methods have been unable to study them, and Brooks' computations exploring early stages of unfolding are in many ways like experiments. Using the supercomputer as a laboratory, he gathers data about a fundamental process that isn't well understood.

"We're exploring phenomena that haven't been explored before," says Brooks. "There's not a well defined model in which we provide input and get the answer. We don't know in detail what we're looking for, so we're coming as close to doing experiments as you can without being in the laboratory." Brooks was able to conduct a number of these experiments during the "friendly user" period of testing on Pittsburgh's CRAY C90. He focused on apomyoglobin, a partially unfolded form of myoglobin, an oxygen-carrying protein that occurs widely in muscle tissue.

go back to the main screen