Inner Space

Anton Simulations Uncover Importance of Empty Space for Protein Function

Experiments can detect chemical changes on a scale as short as about a thousandth of a second. Most supercomputers can only simulate complex biomolecules for as long as a few millionths of a second. Scientists at the University of California, San Diego used the D.E. Shaw Research Anton supercomputer at PSC to understand what happens in that thousand-fold gap, between the snippets we understand from computer simulations to the longer responses we see when a protein, say, encounters a drug.

Why It’s Important

A big limitation to our understanding about how drugs work—and how our bodies carry out day-to-day functions—lies in a “blind spot” in our ability to study them. Experiments can detect chemical changes on a scale as short as about a millisecond—a thousandth of a second. Most supercomputers can only simulate complex biomolecules for as long as a few hundred nanoseconds—less than a millionth of a second. That’s more than a thousand-fold gap in our understanding about how important proteins work in, for example, brain function, heart activity and cancer. Postdoctoral fellow Jamie Schiffer, working in the laboratory of Rommie Amaro at the University of California, San Diego (UCSD), wanted to understand what happens in that gap, between the snippets we understand from computer simulations to the longer responses we see when a protein, say, encounters a drug.

“The average person measures time in minutes, even days. Experimentalists measure protein motions … on the millisecond time scale … We [can] simulate proteins at the nanosecond time scale. Then there’s Anton, which can get up to the millisecond time scale.”—Jamie Schiffer, UCSD

How PSC Helped

The UCSD scientists turned to Anton, a D.E. Shaw Research (DESRES) supercomputer then hosted at PSC. Anton is a super-specialized system that only does “molecular dynamics,” simulating the motions of the atoms that make up large biomolecules. Anton enabled Schiffer to simulate the folding and unfolding, or “dancing,” of the protein T4 lysozyme L99A for timescales as long as 30 microseconds. Such simulations could have taken hundreds of days on another system.

The Anton simulations revealed an unexpected shift of the protein’s shape from the “ground” and “excited” states that were previously revealed by test-tube experiments. In this “intermediate” state, big gaps opened up in the protein, which “let in” molecules roughly the size of drug molecules. Experiments had previously shown that these molecules could get into the protein’s structure—but not how. In order to move between the folding states, lysozyme in effect had to “pull itself apart.” The rare intermediate state couldn’t have been discovered without Anton’s long simulation times. Schiffer and her colleagues reported their findings in the Biophysical Journal late last year. 

The Amaro group next plans to continue their work to understand protein function in health and disease on the new, more powerful DESRES Anton 2 system now hosted at PSC.

“For me [an important lesson from this work] would be to not just focus on the [atoms] in these simulations but also to look into the empty space [between them]. Proteins don’t fit together in a perfect puzzle; there are also these gaps … [We need to] map the empty spaces to … figure out which way these [proteins] might move.”Jamie Schiffer, UCSD