Predicting Disease-Causing Protein Interactions with PIONEER

Researchers at Cornell University and Cleveland Clinic developed an open-access computational platform and online database that helps identify targetable interactions between proteins for therapeutic purposes.

Named PIONEER (Protein-protein InteractiOn iNtErfacE pRediction), this digital tool has proven effective in identifying potential therapeutic targets for various cancers and disorders, as detailed in Nature Biotechnology.

Entering the Era of the Interactome in Genetic Research

Dr. Feixiong Cheng, Genome Center director at Cleveland Clinic and co-leader of the research, emphasizes that while genetic studies are crucial in developing new medications, they alone may not be sufficient. Translating genetic findings into testable drug candidates typically requires 10-15 years before clinical trials can begin.

“In theory, creating medicines based on genetic data is simple: mutated genes produce mutated proteins. We aim to design molecules that block these faulty proteins from interfering with essential biological processes. However, this is far more challenging in practice,” explains Dr. Cheng.

In the human body, proteins form connections with numerous other proteins, creating a complex “interactome” network where each protein's links contribute to an intricate web of interactions. This complexity increases when genetic mutations linked to diseases come into play.

A single disease can emerge from various interactome patterns, as a single protein with different modifications can lead to the same medical condition. Drug developers must evaluate thousands of possible interactions by mapping the structural traits of affected proteins, making this process challenging.

Modeling Protein-Protein Interactions for Future Drug Discovery

Collaborating with Dr. Haiyuan Yu, director of the Cornell University Center for Innovative Proteomics, Dr. Cheng developed an AI tool to identify promising protein-protein interactions, assisting genetic/genomic researchers and drug developers. The tool integrates data from various sources, including:

• Genome sequences from over 100,000 individuals with disease-causing mutations, whether inherited or acquired (such as in cancer).
• Data on how DNA mutations impact the 3D structures of more than 16,000 human proteins.
• Information on around 300,000 known protein-protein interactions.

This extensive dataset allows researchers to investigate protein interactions across over 10,500 conditions, from hair loss to blood clotting disorders.

When researchers identify a genetic variant linked to a disease, PIONEER can generate a prioritized list of protein interactions influencing the condition, identifying possible drug targets. Researchers can also search specific diseases to find relevant protein interactions that may drive the condition. This tool aids researchers in studying various diseases, including immune disorders, cancers, cardiovascular diseases, metabolic and brain disorders, and lung conditions.

To validate the platform, the researchers conducted lab experiments on roughly 3,000 genetic mutations affecting over 1,000 proteins, assessing their impact on approximately 7,000 protein interaction pairs.

Initial studies based on these findings are now exploring treatments for lung and uterine cancers. The team also demonstrated that mutations in protein-protein interactions within their model could predict:

• Cancer prognosis and survival rates, including in rare but deadly cancers like sarcoma.
• Responses to anticancer drugs in large pharmacogenomics databases.

Notably, mutations in the NRF2 and KEAP1 protein interaction were experimentally shown to forecast tumor growth in lung cancer, highlighting a new potential target for cancer therapies.

“Conducting interactome studies requires significant resources, posing a barrier for many genetic researchers. We hope PIONEER can help overcome these barriers computationally, making it easier for scientists to advance new therapies,” says Dr. Cheng.