Computational “Hive Mind” Helps Scientists Solve an Enzyme’s Cryptic Movements

Share
VIDEO | 01:01
A visualization of the enzyme SETD8 in action.
Video Details

A goal of chemical biology is to characterize the movements of proteins as they twist, turn, clamp, and squeeze other molecules during chemical reactions.

It’s currently not possible to watch these acrobatic gestures in real time — microscopes aren’t yet powerful enough — so instead scientists must take a different route.

One common approach is to bombard a protein with X-rays, then analyze the traces that the rays leave on a piece of film. This technique is called X-ray crystallography. From one or more of these pictures, scientists can try to deduce the protein’s movements.

The challenge is that typically just a few poses in the protein’s complete routine are known. Trying to get a full view of all the movements with a handful of still pictures is like assembling a feature-length film with some snapshots.

Now, for the first time, scientists at the Sloan Kettering Institute have used sophisticated chemical techniques and powerful machine-learning algorithms to piece together the complete movements of an enzyme called a methyltransferase.

The breakthrough is the result of a collaboration between scientists in the chemical biology lab of Minkui Luo and the computational and systems biology lab of John Chodera — with invaluable assistance from a worldwide network of citizen scientists running software on their computers. The results, which were published on May 13, 2019, in the journal eLife, will be of broad interest in the field of chemical biology and to scientists interested in drug design.

Capturing Hidden Structures

Dr. Luo’s lab has a long-standing interest in methyltransferases. These are enzymes that move a methyl group from one molecule to another. They play an important role in regulating which genes are on or off in a cell by altering the packaging of DNA in a process called epigenetics.

The specific methyltransferase that Dr. Luo and his colleagues focused on in this study is called SETD8. It transfers a methyl group to a protein called a histone, around which DNA is wrapped like thread around a spool. Mutations in this protein have been linked to different types of cancer.

Trying to get a full view of all the movements with a handful of still pictures is like assembling a feature-length film with some snapshots.

When the scientists started the project, only one conformation — one pose — of SETD8 was known. This was not enough to deduce the protein’s full range of motion. So the first thing they did was try to capture several of the enzyme’s other poses. To do that, they trapped it in different conformations using a variety of small molecules that fit into the enzyme’s grooves. Then they used X-ray crystallography to deduce the shape of the atoms in each conformation. They did this several times until they had captured five individual poses of the enzyme.

“To assemble a complete view, we wanted to know what exactly we missed between these five pictures,” Dr. Luo says. “It could be a straight path between them or it could be a very complicated movement. That’s why we needed computer modeling to help us make predictions.”

Assembling an algorithm that could handle these predictions involved more than 20 years’ worth of algorithm development by a collaborative community of computational scientists. Running the resulting algorithm requires massive computational power — so much power, in fact, that no one computer can handle it.

Citizen Science Aids Computational Biology

To access the required computational power, the SKI team enlisted the help of a distributed computer network called Folding at Home. This massive, interlinked network of computers — a kind of hive mind — has computational prowess that rivals that of national supercomputers.

“No one institution has this amount of power,” says Rafal Wiewiora, a graduate student in the Tri-Institutional PhD Program in Chemical Biology who helps manage the network and is a joint first author on the paper. The computer cluster at Memorial Sloan Kettering has roughly 400 graphics processing units, but, Mr. Wiewiora explains, “Folding at Home has 15,000.”

I think the reason why people like this work is because from the very beginning we set a pretty high bar for both the biochemistry and computational parts of the study.
Shi Chen graduate student

Individual members of the network donate their computers to solve small parts of a data problem. They aren’t paid, but they do compete with one another for points.

Many of the members are motivated to participate for personal reasons. “I had someone contact me recently who said, ‘I’m a cancer survivor, I have the hardware, and I’m really committed to this,’ ” Mr. Wiewiora recalls.

There are currently more than 33,000 members of Folding at Home.

A Synergy of Talents

In addition to computational power, the project also required the perfect blend of complementary expertise. Shi Chen, also a graduate student in the Tri-Institutional PhD Program in Chemical Biology and the paper’s other first author, brought the biochemical proficiency needed to ground the computational methods.

“I think the reason why people like this work is because from the very beginning we set a pretty high bar for both the biochemistry and computational parts of the study,” Mr. Chen says.

The complete picture of the protein’s movements establishes what the researchers call the conformational landscape of the protein. It illustrates how a change in the shape of one part of the protein can alter the shape of another part, like pulling a thread in a game of cat’s cradle.

It also shows how certain cancer-associated mutations can disrupt the protein’s actions, even when far away from the business end of the enzyme.

Movies that the team generated have begun to make the rounds on Twitter. An artist colleague in England even set one to music.

“Every scientist secretly dreams of the day when their work is featured in a groundbreaking new musical composition,” Dr. Chodera tweeted.

This research was supported by the National Institutes of Health, the National Cancer Institute, Starr Cancer Consortium, Mr. William H. Goodwin and Mrs. Alice Goodwin Commonwealth Foundation for Cancer Research, Tri-Institutional Therapeutics Discovery Institute, K. C. Wong Education Foundation, Chinese Academy of Sciences, National Natural Science Foundation of China, the Tri-Institutional PhD Program in Chemical Biology, the Department of Defense, and the Marie-Josée and Henry R. Kravis Center for Molecular Oncology. Dr. Chodera was a member of the Scientific Advisory Board for Schrödinger, LLC during part of this study. He is a current member of the Scientific Advisory Board of OpenEye Scientific Software. A complete funding history for the Chodera lab can be found at http://choderalab.org/funding.