Protein design: how scientists are creating tailor-made molecules to tackle major challenges

Using computational models and AI, scientists are developing proteins to address scientific and technological problems

Three-dimensional digital illustration of a protein, featuring organic shapes and translucent colors, representing molecular structures used in protein design research.

In 2024, the Nobel Prize in Chemistry recognized a field that has been growing in prominence in biotechnology: protein design. The prize was awarded to David Baker, Demis Hassabis, and John Jumper. While Hassabis and Jumper were honored for advances in predicting protein structures, Baker received the Nobel for developing computational methods capable of creating proteins not found in nature.

The international recognition has piqued interest in a field that is not actually new. According to Helder Veras, a researcher at the Brazilian Center for Energy and Materials Research (CNPEM), the first scientific studies on the artificial development of proteins date back to the late 1970s.

Nevertheless, recent advances in computing power and AI have repositioned protein design as one of the most promising frontiers of contemporary science.

Beyond applications in human health, researchers also see the potential of protein design to mitigate environmental impacts caused by human activity, develop new biological materials, and create more sustainable industrial solutions.

Since completing his PhD, Veras has dedicated his career to studying the relationship between the amino acid sequence, three-dimensional structure, and biological function of proteins. Protein modeling is based on this very principle: predicting a protein’s three-dimensional shape from the order of its amino acids.

This structure is not merely a geometric detail: it determines how a protein interacts with other molecules and thus what role it plays in an organism. Ultimately the relationship between sequence, structure, and function guides both protein modeling and design.

From sequence to function: how proteins are modeled

Protein design follows the opposite logic of traditional modeling. Instead of starting from the sequence, scientists begin with the desired function and the structure needed to perform it. From there, they try to predict which amino acid sequence could produce that specific structure.

The process involves repeated cycles of defining structure, function, and sequence—a task that relies heavily on advanced computational tools.

Computer modeling and AI are therefore indispensable. According to André Teixeira, senior director of the antibody platform at the Institute for Protein Innovation (IPI) in Boston, it is theoretically possible to create a protein at random—but that would be highly inefficient.

“It is possible to design a protein from scratch in a completely random way, but that would not be very smart,” explains Teixeira.

At IPI, the focus is on developing antibodies, each of which is a new protein. The team led by Teixeira produces approximately 300 antibodies per year, a volume that would be impossible without the use of AI. “To put that into perspective, a single person might create one new antibody over the course of an entire PhD research project.”

Without these tools, scientists would have to manually evaluate millions—or even billions—of variants before finding a candidate with functional potential. Artificial intelligence reduces this vast range of possibilities. Instead of analyzing millions or billions of variants, researchers now work with between 1,000 and 10,000 versions per project, making the experimental phase more manageable.

“AI is not yet advanced enough to provide a definitive solution to a problem, but it is very good at finding alternative ones,” says Teixeira.

Applications in health and beyond

Creative freedom is one of the defining features of protein design. For Veras, the main limit is imagination.

“In protein design, your imagination is the limit, because the possibilities are so vast,” he says.

Scientists at CNPEM are exploring applications in health, including the development of new therapies and diagnostic tools based on novel proteins. One current project is focused on immune recognition, with the goal of creating proteins capable of directing the immune system toward specific targets, such as tumor cells.

The project is currently in the computational stage, where the structure, function, and sequence of candidate proteins are determined.

Despite the enthusiasm surrounding the field, Teixeira has reservations. AI still has significant limitations, especially in its ability to predict which molecules will perform well under real-world conditions. Many proteins designed computationally fail when tested in the laboratory. This happens because after the computational phase, proteins need to be produced and evaluated in real biological systems, where unforeseen interactions can compromise their performance.

Furthermore, not all problems require proteins designed from scratch: “One of the major objectives of protein design is to develop antibodies, but this year marks the 50th anniversary of monoclonal antibodies, which means that problem has already been solved,” he points out.

The Brazilian landscape: scientific advances and funding bottlenecks

Roberto Lins, a senior researcher at the Aggeu Magalhães Institute of the Oswaldo Cruz Foundation (Fiocruz), spent part of his career in Switzerland and the US before returning to Brazil in 2009 to work at the Federal University of Pernambuco (UFPE). “At that time, protein design was largely restricted to experimental work; the computational side was still in its infancy,” he recalls.

Lins was responsible for several pioneering initiatives in Brazil: he published one of the country’s first articles on the subject in 2013 and organized international courses on protein design that started in 2016. These courses brought together scientists from across Brazil, Latin America, and Europe, helping to form an international network of researchers in the field.

One of the participants was Carlos Cruz, now a researcher at University College London, who is developing protein-based vaccines for birds. According to Cruz, Brazil started early, but faces a persistent struggle for funding.

Lins agrees. He points out that Brazil accounts for less than 1% of scientific output in protein design, while the US is responsible for 30% to 40% and Europe produces between 20% and 30%.

Despite scientific advances and well-trained specialists in the field, protein design will only succeed in Brazil if it receives stable funding and infrastructure. Without continuous investment in R&D, the country risks being left on the sidelines of a strategic field for the future of science, health, and technological innovation.

* This article may be republished online under the CC-BY-NC-ND Creative Commons license.
The text must not be edited and the author(s) and source (Science Arena) must be credited.