Brooks' objective in his research on apomyoglobin is to extend and complement findings from recent nuclear magnetic resonance (NMR) experiments. Using sophisticated methods of NMR spectroscopy, these experiments give a rough picture of what parts of a protein are folded or not under different conditions. In particular, the NMR experiments indicate that as apomyoglobin is subjected to a progressively more acidic environment it unfolds. It seems to unfold, however, in a particular way -- not in a continuous movement from folded to unfolded states but following a specific pathway, as Brooks puts it, with waystations along the way.
The experiments indicate that there are intermediate, stable states of the protein. Apomyoglobin's pathway ends at about pH 2, where it completely unfolds. But in a less acidic environment around pH 4, it appears to assume a partially folded, compact structure that Brooks refers to as a "molten globule." Part of his objective is to provide more detail about this and other intermediate structures. "Experimentally, people have begun to characterize these protein-folding intermediates as either metastable or transient structures. But there is a large lack of detailed structural models."
Brooks simulated apomyoglobin under five different pH conditions. The number of water molecules included ranged from 3,800 to 6,600 as the protein partially unfolds (requiring more water to surround it), and the smallest simulation included about 14,400 atoms in total. Each of these massive computations tracked the protein for at least 1.5 nanoseconds, with an instantaneous snapshot every two femtoseconds (a millionth of a billionth of a second) -- resulting in at least 750,000 individual "frames" of data for each simulation.
These calculations would not have been feasible, says Brooks, without the "friendly user" opportunity on the CRAY C90. "I would not have been able to explore this system in this detailed way on conventional machines, and in this case 'conventional machines' extends even to the C90 under normal conditions of a large number of multiple users."
Brooks and his colleagues are analyzing this huge collection of data using interactive graphics and other tools, a task that will continue for months and perhaps years. Preliminary results have been encouraging. "The most obvious thing we see," says Brooks, "is strong, direct correlation at nanosecond time-scales between the motion of the three-dimensional structure and the NMR data. This was the first benchmark comparison we wanted to do -- to give us confidence we're on the right track."
go back to the main screen