Skip to main content

Simulations, machine learning to speed biological discoveries: Omar Demerdash

Omar Demerdash is using computational biophysics and machine learning to simulate the interactions between proteins and molecules in order to speed the development of new cancer treatments and other biological discoveries. Photo by Carlos Jones, Oak Ridge National Laboratory.
Topics: Clean Energy Biological Systems

Attracted to biology, math, and physics as a young student, Omar Demerdash decided that when the time came to narrow his academic interests he wouldn’t pick and choose: he’d pursue them all. Today he’s using his expertise in computational biophysics to model and analyze how molecules interact with proteins to hasten the development of drug therapies for cancer. 

“Almost any medication you take involves a small organic molecule modulating the function of a protein,” said Demerdash, a Liane B. Russell Fellow in the Biosciences Division at Oak Ridge National Laboratory. “I’m running calculations at the atomistic scale to determine whether a potential therapeutic molecule will actually target that protein of interest.” The work entails calculations of energies and simulating how the protein shape and dynamics will change, which indicate whether the therapy will be effective. 

He is also applying machine learning to improve on the predictions that can be made with first-principles physical chemistry alone. The research can greatly speed the development of cancer treatments by winnowing down the most promising drugs for experimentation.

When the time came to set a career course, Demerdash’s father, himself a professor of electrical engineering, encouraged his son to pursue a degree in an applied field such as engineering or medicine rather than physics. “I was interested in human health, but I eventually realized I was more interested in medical research than interacting with patients or even teaching full time,” Demerdash noted. 

He also developed an interest in the field of neuroscience early on, owing to research experiences as an undergraduate student as well as a more personal story: a good friend of his older sister was severely injured in a car crash at the end of her sophomore year of college, leaving the young woman a paraplegic. “The incident left an indelible impression,” Demerdash said. “Injury can occur at any age and leave a person debilitated for many years. I like the idea of tackling those clinical problems, especially in terms of tissue regeneration.”

Exploring biology at its fundamentals

He cites his early interest in neuroscience for pushing him in the direction of computational physical chemistry. “Specifically, what molecular signals in the central nervous system will induce neurons to be born? Wouldn’t it be great to have computational predictions of which proteins or small molecules are going to have the effect of regenerating tissue in the brain or the spinal cord? That was the impetus that led me to this field and away from doing experiments—of going at biology at a more fundamental level,” he said.

After earning his undergraduate degree in molecular biology, Demerdash entered the Medical Scientists Training Program at the University of Wisconsin-Madison. It was around that time that he experienced an epiphany about the interrelated nature of what would become his scientific career. “I fundamentally felt that biology should be studied at the level of physics. Biology is ultimately chemistry, and chemistry is ultimately physics. I began to think we can approach biological problems from the level of first-principle physics, and from there we can really target any therapeutic challenge of interest. Any disease can be treated as a fundamental problem of modeling the physics of the key players, like proteins interacting with other proteins or with drug molecules, or binding to DNA and regulating gene expression.”

He went on to earn both a doctoral degree in biophysics and a medical degree from UW-Madison, and then continued his research as a postdoc at the University of California-Berkeley’s College of Chemistry.

In addition to the Russell Fellowship work, Demerdash is applying his expertise to other projects at ORNL, including the plant-microbial interface science focus area and the application of machine learning to neutron scattering results in order to improve the lab’s efforts to model intrinsically disordered proteins.

The best outcome? Models that can be leveraged by others

He describes the best outcome of his work as the development of methods that not only yield good results, but also can be used by others when making predictions that guide experiments.

His advice to young scientists is to pursue a strong grasp of higher mathematics, including modern mathematics like topology and non-Euclidean geometry that are not part of the standard curriculum. “These higher maths are a key component of string theory in theoretical physics, so it wouldn’t surprise me that they should have important applications to, for instance, the physics of molecules, which is the scale of physics that my work encompasses.”

Math requirements for the life sciences need to be increased far beyond what is usually required today, he stressed.  “Typically, it is only physical scientists who are exposed to higher maths, and this leads to a fundamental schism between the physical and life sciences.” He also emphasizes the need for scientists in all fields to have good computer programming skills, to foster increasingly important work in machine learning and data mining.

He is interested in pursuing neuroscience research again in the long term. “I’d like to see more research on clinical problems after injuries. In recent conflicts, for instance, we have been able to save soldiers who might have previously died on the battlefield. But the soldiers are often left with poor quality of life due to traumatic brain injury and spinal cord injury. That motivates me quite a bit.” 

He cautions that science requires a certain fortitude. “Simple things like cleaning up code can sometimes be difficult and it can make the process feel like a real slog. I think a lot of people who aren’t in science think I’m discovering things all the time—but you have to be patient. I’ve been pleasantly surprised with how well machine learning is working, and in particular with the preliminary results of my work here using machine learning to predict protein-small molecule binding.” 

ORNL is managed by UT-Battelle for the Department of Energy's Office of Science, the single largest supporter of basic research in the physical sciences in the United States. DOE’s Office of Science is working to address some of the most pressing challenges of our time. For more information, please visit