Pan-cancer neoantigen discovery and functional immunogenicity prediction using integrative deep learning approaches

Using machine learning to unlock new targets for cancer vaccines  

As the field of cancer immunotherapy rapidly evolves, researchers are racing to uncover the next generation of tools to help the immune system better recognize and destroy cancer cells. One of the most promising tools in this fight is something surprisingly familiar: vaccines. 

“Cancer vaccines are a form of highly personalized therapy,” explains Vivian Chu, a PhD Candidate at the Princess Margaret Cancer Centre. “They aim to train the immune system to detect and attack cancer cells using molecular markers called neoantigens – unique proteins that appear only on cancer cells, not on normal ones.” 

In this context, making neoantigen discovery faster and more efficient is extremely important. That is why Chu, with new support from the 2025 MOHCCN Health Informatics and Data Science Award, is developing an advanced machine learning model to accelerate the identification of the most promising neoantigen candidates across a broad range of cancers.  

By identifying new neoantigens, we move towards developing more precise and personalized cancer treatments, such as cancer- or patient-specific vaccines. As the field of cancer immunotherapy continues to evolve, new methodologies to rapidly and accurately discover neoantigens may accelerate the translation of innovative immunotherapies into clinical practice, offering new avenues to improve patient outcomes. 

While many current strategies focus on mutations in cancer DNA, Chu is expanding the search to include other overlooked but promising sources of neoantigens, such as alternative splicing and gene fusions. 

To do this, she will apply integrative deep learning models to analyze DNA, RNA and protein data from patients from the Marathon of Hope Cancer Centres Network, identifying and prioritizing potential neoantigens. The most promising candidates will then be validated in Living Biobank models to assess their potential for use in vaccine design. 

“I am incredibly honored and thrilled to receive the 2025 MOHCCN Health Informatics & Data Science Award,” said Chu. “Our research aims to advance precision oncology using deep learning and multi-omic integration to discover and screen novel neoantigens. I am excited by the potential to translate these discoveries into more effective patient-specific cancer vaccines and am deeply thankful for the opportunity to do so with the support of the Marathon of Hope and the Terry Fox Research Institute.” 

According to Dr. Housheng Hansen He, Chu’s research mentor, this work addresses a major challenge in the field of cancer immunotherapy. 

“Vivian’s study addresses a critical gap in cancer immunotherapy by integrating deep learning with multi-omic data to identify novel neoantigens across tumor types. Her approach could reveal novel targets arising from mutation, splicing and gene fusions, accelerating the development of personalized mRNA vaccines. This work is poised to make a significant impact on precision oncology by expanding the repertoire of immunogenic targets and informing future strategies for patient-specific cancer treatment.” 

By pushing the boundaries of data science and immunogenomics, Chu’s work has the potential to significantly expand the scope of personalized cancer treatment—and bring us one step closer to truly individualized care.