A collaborative research team from KAIST led by Professor Jung Kyoon Choi from the Department of Bio and Brain Engineering and PentaMedix, a company striving to develop precision medical solutions for cancer therapy, introduced an innovative AI model named DeepNeo. This model is designed to use deep learning in order to predict neoantigens, a key component in formulating personalized cancer vaccines.

Neoantigens are antigens found unique to cancer cells that are derived from mutated protein fragments. They trigger an immune response which the immune system uses to target and fight the cancer cells specifically, reducing potential harm to healthy cells and improving the effectiveness of the treatment. As such, neoantigens have become increasingly recognized as the ideal targets for individualized cancer vaccine development.

Existing methods for neoantigen discovery have been limited to forecasting mutations in the neoantigen that can bind with major histocompatibility complex (MHC) proteins. MHC proteins play a crucial role in triggering immune responses by presenting antigens derived from pathogens or cancer cells to our immune cells. Specifically, mutations within the tumor-specific antigenic sequences, or neoantigens, were the subject of this predicted process. However, for a cancer vaccine to be truly potent, the mutations must not only bind with the MHC, but the resulting complex must also activate a T cell immune response — a challenge unmet by existing technology. Consequently, ongoing clinical trials for cancer vaccines proceed without certainty that the formed complexes can genuinely stimulate immune responses. 

The DeepNeo model goes beyond merely predicting the neoantigens that can bind with MHC proteins. It can also identify those with the potential to trigger a T cell immune response, which is a crucial aspect in the making of personalized cancer vaccines that has not been fully addressed in previous methodologies. The issue of not knowing whether the neoantigen-MHC complex would actually stimulate the immune system has been a significant roadblock in past efforts. However, DeepNeo’s unique feature allows researchers to now pinpoint neoantigens that can prompt a T cell response, enabling the creation of more effective cancer vaccines. To facilitate this groundbreaking approach, the research team launched a web service that shares the Al model’s name. This new web-based approach is anticipated to bring about a transformation in cancer vaccine development, with the ultimate goal of efficiently provoking T cell responses and providing renewed hope for patients affected by cancer. 

Professor Jung Kyoon Choi commented, “With the mRNA platform’s proven success in COVID-19 vaccines, we hope that the AI technology we developed will also aid in the commercialization of cancer vaccines.” Dae-yeon Cho, the representative of PentaMedix, announced plans to accelerate the commercialization of personalized cancer vaccines utilizing the developed platform.