Neo-epitopes/antigens discovery for cancer immunotherapy based treatment with precision oncology using advanced Artificial Intelligence


Year:

Session type:

Theme:

Jyothi Deva datta reddy1,Deva Reddy2,Dr.ram Reddy2,Samyukta Jyothi2
1Cancer moonshot @ www.cancermoonshot.in,2Cancermoonshot

Abstract

Background

Cancer has historically been treated with surgery, radiation therapy, chemotherapy and hormone therapy, which use nonspecific mechanisms to attempt to remove or to kill cancer cells. More recently, a new class of immunotherapy known as checkpoint inhibitors is also being applied. However, the majority of patients do not respond to checkpoint inhibitors. We believe that neoantigen-targeted therapies may precisely direct the immune system to improve patient outcomes across both checkpoint-responsive and unresponsive disease. Early evidence shows that neoantigen-based vaccination can elicit T-cell responses and that neo-antigen targeted cell-therapy can cause tumor regression.

To make this happen we need accurate, fast, efficient, affordable neoantigens and corresponding T-cells with high specificity.

Method

Identifying true tumor-specific neoantigens which are rare and unique that on the cell surface is like finding a needle in the haystack. To address this great challenge, CancerMoonshot developed a proprietary artificial intelligence platform called NeoAintigen leveraging huge extensive data from real human tumor samples. NeoAintigen AI platform was trained on a huge dataset and enables us to use sequence data from a cancer patient tumor biopsy to predict/identify exact mutations that will generate tumor-specific neoantigens supposed to on the tumor cell surface.

  • Data-500+ patient samples, 1Million+ peptides, Various tumor types,HLA-Class1,CD4+ ,8-15 mer peptides
  • Technology – AI, Deep learning, Machine learning, Inference engine, In-silico

 

Results

  • Identify at least  one T-cell recognized neoepitope
  • Average neo epitopes = 2
  • Better predictive performance : 10-15 %
  • Accuracy > 5x or 5 folds compared with publicly available approaches.
  • Better targetable/correlation with T-cells
  • Near real-time

Conclusion

·         Here we show that AI can be used to predict neoantigens better than current techniques and in vitro models.

·         Need to test for more data and cancer types ,and run clinical trials successfully.

·         Need to extend for other class types like HLS-II,Cd8+ and so on in the future.