Billions of years ago, straightforward organic molecules drifted across Earth’s primordial landscape – nothing more than basic chemical compounds. But as natural forces shaped the planet over hundreds of millions of years, these molecules began to interact and combine in increasingly convoluted ways. Along the way, something spectacular appeared: life.
“Life is magical to some extent,” says computational biologist Sergei Kotelnikov. Plain organic compounds combine to form polymers, which combine to form living cells and, ultimately, organisms – the whole is greater than the sum of its parts.
“You can write formulas for the behavior of a molecule,” he says, referring to the world of quantum mechanics. “And yet somehow, several orders of magnitude higher, on a larger scale, it gives rise to such a mystery.”
Kotelnikov builds models to analyze and predict the structure of these biomolecules, especially proteins, the basic building blocks of every organism. This year, he joined MIT as part of a School of Science Dean’s Postdoctoral Fellowship to work with Keating’s labin which researchers focus on the structure, function and interactions of proteins. Its goal is to develop modern methods for modeling proteins with potential applications from medicine to agriculture, using machine learning.
Hunger for problems to solve
Kotelnikov grew up in Abakan, Russia, a compact town located in the very center of Eurasia. As a child, one of his favorite pastimes was playing with Lego blocks.
“It encouraged me to build new things rather than just follow instructions,” he says. “You can do anything.”
Kotelnikov’s father, who had a background in engineering and economics, often challenged him with math.
“Your brain — you can feel a kind of expansion of your understanding of how things work, and it’s a very rewarding feeling,” Kotelnikov says.
His desire to solve problems led him to participate in Science Olympiad competitions and later to a public science boarding school located near the Russian Academy of Sciences, where he often met scientists.
“It was like being in a candy store,” he recalls, describing the period as a life-changing experience.
In 2012, Kotelnikov began his bachelor’s degree in physics and applied mathematics at the Moscow Institute of Physics and Technology – considered one of the leading STEM universities in Russia and the world – and continued there to obtain a master’s degree. That’s where biology came in.
During a statistical physics course, Kotelnikov first encountered the idea of ”complexity emergence.” He was fascinated by “the mysterious and attractive manifestation of biology… that evolution that sharpens a physical phenomenon” to create, drive and shape life as we know it today. By the time he completed his master’s degree, he realized he had only scratched the surface of computational biology.
In 2018, he began doctoral studies at Stony Brook University in Novel York and began collaborating Dima Kozakovrecognized as one of the world leaders in predicting the interactions of proteins and convoluted structures.
Studying the architecture of life
Proteins act as the building blocks of the body, supporting almost every cellular process, from tissue repair to hormone production. Like the pieces of a Lego tower, their structure and interactions determine the functions they perform in the body.
However, diseases arise when they are folded, rolled, twisted or connected in an unusual way. To develop medical interventions, scientists dismantle the tower and examine each component to find the culprit and improve their shape and fit. With confined experimental data currently available on protein structures and interactions, simulations developed by computational biologists such as Kotelnikov provide key information that impacts fundamental understanding and applications such as drug discovery.
Under the direction of Kozakov at Stony Brook’s Laufer Center for Physical and Quantitative BiologyKotelnikov transferred his knowledge from physics to create modeling methods that are more effective, competent, reliable and generalizable. Among them, he developed a modern way to predict the structures of protein complexes mediated by chimeras targeting proteolysis, or PROTACs, a modern class of molecules that can trigger the degradation of specific proteins that were previously considered impossible to treat, such as those found in cancer.
Modeling PROTACs has been tough, in part because they are composed of proteins that do not naturally interact with each other, and also because the linker that connects them is malleable. Imagine trying to guess the overall shape of a bent Lego piece attached to two other pieces of various irregular, mismatched shapes. To efficiently find all possible configurations, the Kotelnikov method conceptually divides the linker into two halves and models each of them separately, then reformulates the problem and computes it using a powerful algorithm called the speedy Fourier transform.
“It’s a bit like applied math judo, where sometimes you have to do some difficult calculations,” he says.
Kotelnikov’s cutting-edge methods have played a key role in ensuring his team has excelled in many international challenges, including Critical evaluation of protein structure prediction (CASP) – the same competition that presented the Nobel Prize-winning AlphaFold system for predicting the 3D structure of proteins.
Physics and machine learning
At MIT, Kotelnikov collaborates with Amy Keating, Jay A. Stein (1968) Professor of Biology, chair of the biology department and professor of biological engineering, to study the structure, function, and interactions of proteins.
A recognized leader in the field, Keating uses both computational and experimental methods to study proteins, how they interact, and how they may influence disease. By combining physics with machine learning, Kotelnikov’s goal is to improve modeling methods that can greatly lend a hand in applications such as cancer immunology and crop protection.
“Kotelnikov has a lot to gain from working closely with wet lab researchers who are conducting experiments that will complement and test his predictions, and my lab will benefit from his experience in developing and applying advanced computational analyses,” Keating says.
Kotelnikov also plans to cooperate with professors Tommi Jaakkola AND Tess Smith IN MIT Department of Electrical Engineering and Computer Science to explore a field called geometric deep learning. Specifically, it aims to integrate physical and geometric knowledge of biomolecules with neural network architectures and learning procedures. This approach can significantly reduce the amount of data needed for training and improve the generalizability of the resulting models.
In addition to the two departments, Kotelnikov is also excited to see how the diversity and interdisciplinary mix of the MIT community will lend a hand him generate ideas.
“By building a model, you enter this imaginary world of assumptions and simplifications, which may seem difficult due to the lack of contact with reality,” says Kotelnikov. “There is great value in being able to communicate effectively with experimenters.”