News

Model Helps Predict Protein’s Immune-Boosting Ability

A promising area of cancer treatment, called immunotherapy, uses the body’s own immune response to seek out, target, and destroy tumors. This approach leverages fragments of proteins, or neo-epitopes, that result from mutations in cancer cell genes. Neo-epitopes, because they arise from the tumor’s own cells, are particularly effective for simulating the body’s immune cells—encouraging or stimulating an anti-tumor response.

Neo-epitopes are known for their ability to spark a response, and the greater the response, the greater their immune-boosting ability (i.e., immunogenicity). However, predicting potential immunogenicity can be a significant bottleneck in bioinformatic pipelines.

Researchers funded by NCI's Informatics Technology for Cancer Research Program have a new model, called ICERFIRE (or “ICore-based Ensemble Random Forest for neo-epitope Immunogenicity pREdiction”), to help you predict neo-epitope potency.

Read the full report, “A Large-Scale Study of Peptide Features Defining Immunogenicity of Cancer Neo-Epitopes,” in NAR Cancer. You also can try the model by submitting your own data.

According to the study authors, ICERFIRE outperformed current models in predicting which epitopes were likely to be most immunogenic.

Corresponding author on the study, Yat-tsai Richie Wan, in the department of health technology at the Technical University of Denmark, notes that discovering potential neo-epitopes is a crucial step in refining and advancing precision medicine, especially in the area of cancer vaccines.

He said, “Given the heterogeneous nature of cancer mutations, finding key targets is crucial for personalized therapy. Knowing which neo-epitopes are more likely to trigger an immune response in a given patient, at a given time, could help tailor the best treatment, in both cancer immunotherapies and personalized vaccines.”

The researchers are refining their model. More high-quality data will help train ICERFIRE to help avoid bias and to make the data behind the model as inclusive as possible.

Vote below about this page’s helpfulness.