Powerful AI tool can boost precision medicine, treatments

Human proteins and their interactions with other biomolecules, especially other proteins, are the cogs that enable biological processes. Each protein has a unique structure, which then fits with certain other proteins and biomolecules to create interactions that facilitate cellular function.

A new tool harnesses the power of AI and deep machine learning models to solve and predict how human proteins might interface and interact with other proteins. The tool can greatly accelerate fundamental research, clinical precision medicine and the development of therapies or the application of existing drugs to treat all types of disorders.

The tool, called PIONEER (Protein-protein InteractiOn iNtErfacE pRediction), was described in an Oct. 24 study published in Nature Biotechnology, where it was applied to 33 cancer types and their disease-associated mutations. Deep learning AI models then predicted protein interactions with those mutations to inform how well specific drugs might respond to a cancer and predict patient survival rates based on how each cancer patient's unique biology might react to specific treatments.

PIONEER, which is available as a web platformand a software package, has also been applied to seven other common model organisms used in research, such as mice.

"What we did here is solve structures for protein machineries that actually carry out cellular functions within cells," said co-corresponding author Haiyuan Yu(College of Agriculture and Life Sciences), Tisch University Professor in the Department of Computational Biology and a member of the Weill Institute for Cell and Molecular Biology. Feixiong Cheng, director of the Cleveland Clinic Genome Center in the Lerner Research Institute, is the paper's other corresponding author.

"Not only did we build this AI model, but we used it to solve all known protein interactions in human cells already," Yu said.

In the study, which is an application of the tool, the researchers mapped cancer mutations sequenced from over 11,000 tumor samples across 33 cancer types. Based on specific mutations for each cancer type, the tool identified protein-to-protein interactions whose interfaces were enriched with cancer mutations. These protein-to-protein interactions, called onco-PPIs, significantly relate to patient survival rates as well as treatment responses.

"We show in the paper there are cases where, if there are mutations on specific onco-PPIs, at least in mouse and cell models, the tumor either becomes super susceptible to a treatment or, more often, they become much more resistant to treatment," Yu said.

These results represent important information for so-called precision oncology or precision medicine, to create better individualized treatments.

The application can be generally applied to any disorder. Yu is also using a similar framework to better understand neurodevelopmental conditions such as autism and to potentially treat Alzheimer's disease.

Yu's work builds on AlphaFold, an AI protein structure prediction tool that was recognized in 2023 with a Nobel Prize in Chemistry for its developers. Yu used AlphaFold to develop PIONEER, which takes this line of research one step further to solve complicated protein-to-protein interactions.

First authors Dapeng Xiong, a Cornell postdoctoral researcher; Dongjin Lee, a former postdoctoral researcher in Yu's lab; Yunguang Qiu and Yadi Zhou of the Cleveland Clinic Genome Center and Junfei Zhao of Columbia University contributed equally to the research.

The work was supported by the National Institute of General Medical Sciences; the National Institute of Diabetes and Digestive and Kidney Diseases; the National Institute on Aging; the National Institute of Neurological Disorders and Stroke; the National Human Genome Research Institute; the National Heart, Lung, and Blood Institute; the American Heart Association; and the European Union Horizon Health program.

Public link

Please use the above public link if you want to share this noodl on another website.