Making the Data Take Shape
Phil Bradley, Computational and Structural Biologist
Proteins make our cells run: They give cells their shape, help them move and divide, and ferry key molecules from one spot in the cell to another. Our cells and bodies wouldn’t work without these strings of amino acids, which twist and curl into amazing shapes that enable each protein to fulfill its biological role.
Math can reveal the secret patterns that give a protein its shape and allow it to perform its job. That’s what Dr. Phil Bradley loves about computational biology: the opportunity to use math to puzzle out the biological rules, hidden within large datasets, that shape how proteins’ form and function intersect.
“I’m just a data person. I could stare at colorful data plots all day long,” he said. “It's like a puzzle: Somebody hands you a high-dimensional dataset with a lot of different features and correlations and hidden structure. Well, what's going on?”
Bradley didn’t start out expecting to make a career of decoding biology. At the Massachusetts Institute of Technology, he began a graduate degree in mathematics, and “almost all the way through the end of my Ph.D., I was planning to be a math professor,” he said.
But math wasn’t Bradley’s only interest; biology had also attracted him, though he’d been stymied by the accompanying laboratory requirement. Near the end of his graduate degree, Bradley began to learn more about how math and biology can intersect — a discipline often dubbed computational biology — from his wife, who was studying biology at MIT. In his final year, he wrapped up the pure math portion of his Ph.D. and started tackling some real-world biological questions that needed mathematical answers.
Bradley fully committed to computational biology with his postdoctoral training, joining Dr. David Baker at the University of Washington to build computer models that predict how proteins fold.
It was a steep learning curve for a mathematician who hadn’t yet learned the names of all the amino acids, but Bradley was hooked. When he started his own lab at Fred Hutch, he expanded his research scope to investigate how to model and predict interactions between proteins.
Bradley, holder of the Bob and Pat Herbold Computational Biology Endowed Chair, is especially interested in how a protein on immune cells called T cells interacts with its targets. Each developing T cell has a unique T-cell receptor, or TCR, that can bind with specialized protein complexes that stud the cells of the body. These complexes hold information about a cell’s state of health. T cells use their TCRs to survey the body’s tissues, hunting for infected and diseased cells (like cancer cells) to eliminate.
Once a T cell has encountered its target, it persists in the body, a living record of a person’s immunological history. But this history is written in a language that scientists are still deciphering.
“I've just been fascinated by T cells in that system and how it works,” Bradley said. “It’s mind-boggling that it works at all.”
If Bradley could read the immune system’s language, he would be able to learn more about T cell function, predict a T cell’s target from its TCR genes (which encode its TCR proteins), and deduce the infections and mutations in a person’s past. But he’s contending with fantastically rich, complex datasets made possible by technologies that allow researchers to uncover the history written in the genes of thousands of T cells within even a single person.
“Getting a new dataset — it’s just like Christmas.”
That’s where Bradley’s specialty — math — comes in. It takes equally complex and sophisticated computation to reveal the patterns within these datasets, which can include information about many genes within each T cell, not just its TCR genes.
“We can figure out in some cases what the target of that T cell response is,” he said. Someday, information like this could be used to determine a patient’s prior exposure to infections or, perhaps, even cancer.
Using natural principles to design proteins
As scientific projects tend to do, Bradley’s aims have also morphed with time. In addition to developing methods to predict T-cell targets and functions, he now develops algorithms and computer models to design proteins to perform new functions not found in nature.
The work grew out of his prediction work and a serendipitous connection with a Hutch colleague, structural biologist Dr. Barry Stoddard.
“It was a historical accident,” Bradley recalled.
Both he and Stoddard were working to figure out the structure of a protein called a TAL effector. Released by certain bacteria that infect plants, these proteins are made up of repeating segments that bind to plant DNA and help enable bacterial infection. Researchers hoped to develop DNA-editing tools based on TAL effectors’ ability to target specific DNA sequences, but first they needed to know their structure.
Stoddard was working to determine TAL effectors’ structure by creating crystals of the proteins, while Bradley attempted to predict their structure using computational methods. Stoddard had a rough structure, but needed a little more information to complete it. Bradley in turn had a rough prediction, but needed real-world data to corroborate it. Together, their methods created a refined, validated structure.
Just in time for the precise DNA-editing tool CRISPR to take the wind out of TAL effectors’ sails. But it was the start of a fruitful collaboration between the two scientists.
“The beautiful thing about this TAL effector is the way its structure is really intimately related to its function,” Bradley said. “So the structure looks kind of like if you took DNA, and then you put clay around it to make kind of a mold, and then you took away the DNA. It perfectly complements it.”
The repetitive nature of this remarkable intersection of form and function got the two scientists thinking: Could they create new and interesting protein functions by starting with a simple protein building block, and then repeating it?
That question led to the design of tiny, self-assembling donut-shaped proteins that have a lot of potential. Various other molecules can be attached to each protein segment, making the conceivable uses of the minuscule rings nearly endless. Add the right molecules, and they could offer new molecular scaffolds for cancer immunotherapy or deliver drugs tailored to a specific disease. Bradley and Stoddard are currently refining the proteins’ design and testing possible applications.
“I do like just data and complex data sets in general, but I do especially like problems where protein structure has some kind of role,” Bradley said.
Data analysis isn’t just his career — it’s his love. If he hadn’t made a career of it, Bradley said, he’d still do it as a hobby, perhaps as an amateur cryptographer.
“Getting a new dataset — it’s just like Christmas,” he said.
— By Sabrina Richards, updated August 14, 2024
RELATED RESOURCES
