University of Missouri researchers have released the world’s largest collection of protein models with quality assessment — a groundbreaking new resource that could accelerate drug development for diseases such as Alzheimer’s and cancer.
The database, called PSBench, includes 1.4 million annotated protein structure models, all verified by independent experts. It gives scientists the reliable information they need to build more accurate artificial intelligence (AI) systems for assessing the quality of protein structure models, which is critical for developing future medical treatments.
Proteins are often called the building blocks of life because they drive every biological process in the human body. Their three-dimensional (3D) shapes determine how they function — and even small structural changes can lead to disease.
Recent advances in AI, including tools such as Google’s AlphaFold, have dramatically improved protein structure prediction.
But even AlphaFold has limitations. No single AI tool is consistently accurate for every type of protein, making it difficult for researchers to know when a prediction can be trusted.
PSBench provides that benchmark.
“With PSBench, scientists can develop AI methods to assess the quality of predicted protein models and decide if they can be trusted,” Jianlin “Jack” Cheng, a Curators’ Distinguished Professor and Paul K. and Diane Shumaker Professor in Bioinformatics, said. “Our work represents a significant step toward applying protein models to understanding diseases and developing new treatments.”
Cheng and his team in Mizzou’s College of Engineering built PSBench, leveraging both in-house and community-wide resources generated in the Critical Assessment of protein Structure Prediction (CASP), widely recognized as the international gold standard for assessing computational methods for protein prediction. The biennial competition was created to independently test computer models that predict how protein chains fold into the 3D shapes they need to function.
For more than 50 years, researchers struggled to understand how proteins fold into their complex 3D shapes. At the 2012 CASP competition, the Cheng group was the first to demonstrate that deep learning could help solve the problem. Cheng’s work catalyzed the more-than-a-decade-long deep learning revolution in the field, including the development of AlphaFold, now considered one of the most accurate protein prediction tools in the world.
Cheng and co-authors Jian Liu, a Mizzou postdoctoral fellow, and graduate student Pawan Neupane recently presented their PSBench study at NeurIPS 2025 in San Diego, one of the world’s most prestigious AI conferences and the same venue where transformative technologies underlying ChatGPT were first introduced.
“With PSBench, Mizzou is not only contributing a powerful new tool to the global scientific community,” Cheng said. “We’re helping lead the next chapter in AI-driven biomedical discovery.”