Cleveland Clinic and IBM researchers are using quantum computing to tackle one of the most challenging problems in immuno-oncology: predicting which tumor mutations will trigger an immune response.  

Their framework, Quantum Convolutional HLA Immunogenic Peptide Prediction (Q-CHIPP), combined predictions of antigen presentation and immunotherapy response into a unified framework. It outperformed classical computing methods by improving accuracy under data-limited conditions, according to results published in Science Advances. 

Neoantigens are abnormal proteins produced on a cancer cell’s surface when the cell acquires mutations in their DNA. Neoantigens signal the immune system that a cell is foreign or harmful to the body. An essential part of immunotherapy is teaching the body to recognize and attack neoantigens.  

The challenge is that there can be thousands of potential neoantigens in one tumor, but very few will be recognized by the immune system.  It’s also now well understood that these well recognized (immunogenic) neoantigens are a primary driver of response to immunotherapy.  

To develop cancer vaccines and other treatments, researchers need to understand which neoantigens will activate the immune system to attack cancer cells. Cleveland Clinic researchers recently published an atlas of 196 candidate neoantigens likely to trigger an immune response with the goal of furthering the power of neoantigen predicting machine learning models.  

However, machine learning models require large data sets to learn complex patterns and exhibit high predicting accuracy, and the 196-candidate atlas is still considered a small dataset, which might result in model overfitting, poor performance, and limited ability to generalize effectively.  

To overcome current computing methods’ limitations, the team developed a quantum convolutional neural network approach that learns meaningful biological patterns from small datasets. By combining IBM’s computational expertise and Cleveland Clinic’s biomedical knowledge, the team created a first-of-its-kind model that can learn from training sets with as few as 150 samples.  

“The creation of this model exemplifies team science,” says senior author Tyler Alban, PhD. “This was a multidisciplinary team where we had experts in cancer, computational sciences and quantum computing, all at the table teaching each other and learning from one another. We could not have achieved this without a team approach.” 

The model they developed was able to outperform traditional computing methods when evaluated under similar constraints and parameter limits. The team also achieved a significant technical milestone by scaling their approach to full-length peptide modeling on quantum hardware using 46 qubits.  

The team plans to continue enhancing this model to enable more accurate identification of therapeutic targets, potentially accelerating the development of personalized cancer immunotherapies and next-generation vaccines. 

“This collaborative approach exemplifies how Cleveland Clinic and IBM are pioneering quantum computing in immunology and broadly speaking in healthcare,” says Sara Capponi, corresponding author and Senior Research Scientist, IBM Research. “Only a team combining this breadth of expertise could translate the immune system’s features into quantum circuits, ensure the underlying math faithfully captured the biology, and validate the results against real-world patient data.” 

Journal

Science Advances

DOI

10.1126/sciadv.aec3824