First winner of AAAS-Chen Institute Prize builds tool to visualize biomolecular interactions

(Press-News.org) For his work to help capture and view dynamic small-scale behaviors of biomolecules that have gone unseen – and which are critical to applications like drug development – Zhuoran Qiao has been awarded the inaugural Chen Institute and Science Prize for Al Accelerated Research. The prize recognizes innovative young researchers who apply techniques in artificial intelligence to help the research community solve important problems and accelerate their work.

“I was thrilled to partner with the Chen Institute to launch this new prize initiative,” said Yury V. Suleymanov, senior editor at Science. “Our winner, Zhuoran Qiao, has shown outstanding achievement in the field. His work introduces a transformative approach to decoding and reprogramming molecular biology using artificial intelligence (AI)-driven structural foundation models. It demonstrates how AI can help to overcome the limitations of traditional methods, paving the way for new opportunities in molecular design and therapeutic development.”

Interactions among biomolecules, like proteins and smaller molecules, are key to supporting the fundamental processes of life. Identifying these interactions at ever smaller scales is useful for developing new drugs, among other applications, but doing so requires decoding these interactions’ three-dimensional structures. That requires having zoomed-in snapshots of molecular compartments.

Traditional methods for determining molecular structures, such as X-ray crystallography and cryogenic electron microscopy, are powerful – but slow. It could take months of work in the lab to generate important molecular images.

Recently, AI-driven protein structure prediction tools made powerful progress in this regard. They can predict the three-dimensional structures of proteins from their amino acid sequences. However, these new tools represent “just the beginning of the journey toward creating a fully-fledged ‘computational microscope’ for molecular biology,” writes Qiao, founding scientist at the San Francisco-based artificial intelligence startup Chai Discovery, in his prizewinning essay. Seeing things at the scale of the biomolecule, for systems with not just 100 atoms, but thousands of atoms, and in various conformations     , is also crucial, he said.

In February 2024 in Nature Machine Intelligence, Qiao and his colleagues advanced on the work of AI-driven protein structure prediction to date by developing novel generative machine learning approaches to create a clearer view into two critical activities: protein-ligand interactions, and the landscapes in which these interactions occur.

When a ligand – a molecule or ion that binds to a central atom or molecule – does its binding work, it influences the structure it binds, which in turn greatly influences chemical and biological processes key to our daily lives.

“If you want to develop newer drugs, you need to model biomolecular interactions really accurately,” Qiao said. “You need to get the structure right and understand how strongly the two proteins or small molecules interact. That is the first thing you need to know      if your drug is going to be successful.”

It’s very complicated work, he add. “Showing how molecules move in real life is like navigating a very complicated maze with thousands of dimensions     .”

The tool Qiao’s team developed to visualize these interactions is called NeuralPLexer. It takes into consideration that biomolecules are highly dynamic, requiring numerous snapshots to fully capture their behaviors. Thus, the tool starts from an initial sketch of the entire molecular complex and progressively refines the finer-grained details of the structures it generates. This process help researchers “quickly obtain the full picture of molecular interactions with atomistic details.”

Qiao and his colleagues used NeuralPLexer to do tasks like predicting the formation of “cryptic pockets,” special binding sites that are absent unless spurred by ligand binding. They showed the tool had strong capabilities for identifying new drug binding pockets, among other tasks.

“If you compare this approach to traditional approaches, we are      delivering what high-throughput methods would do in six months in one day,” he said.

Qiao was motivated in this work from early days in the field, based on his recognition that scientists understand the theoretical framework of a lot of systems they study in computational chemistry, but that not many of the related problems are actually computable.

“It is a great honor for      me to win the prize,” said Qiao. “It is a huge recognition to the research path I have chosen. It's also deeply humbling because it reminds me to continue doing impactful work, including mentoring others to be interested in computational chemistry. With new technology, we are seeing how computational studies have real translational potential to develop better drugs and healthcare.”

Qiao is delighted to be part of how molecular modeling is changing, even as there is still a lot of work to be done. He is eager to address some of the next steps at the San Francisco-based startup, Chai Discovery.

“We were excited to receive such an impressive range of applications from around the world, spanning many different scientific disciplines,” said Chrissy Luo, Chen Institute cofounder. “At a time when AI is radically accelerating global scientific discovery, we’re delighted to work with AAAS and showcase three incredible young researchers who are using these powerful new technologies to expand the frontiers of human knowledge.” 

 

 

 

END