Artificial Intelligence is making cancer vaccines closer to reality

Imagine if you could get vaccinated for cancer. You would just get a shot or a couple of shots and your immune system would learn how to recognize and kill cancer. This sci-fi-sounding idea is not that far from reality. Especially now that scientists at the University of Waterloo have employed machine learning to identify tumour-specific antigens, which could lead to new cancer treatments and vaccines.

If we could learn the neoantigens of each tumour, we could create personalized custom therapies for each patient. Image credit: Wikimedia

Cancer starts with a mutation in cell’s DNA. Basically, a DNA error occurs in dividing cells and these mutated cells grow bigger in number forming a tumour. However, this actually happens a lot – just that your immune system can stop cancer in its tracks even before it is able to cause trouble.

When a mutation occurs in a cell’s DNA it gets marked as an invader and the immune system can find it. The substitution of a mutated DNA is called a neoantigen – it is a mutated peptide that appears on the surface of cancer cells. If we could teach immune systems to recognize neoantigens and kill them, we could have an effective cancer vaccine.

The problem, however, is that tumours have different neoantigens and finding them is difficult – scientists compare it to looking for a needle in a large haystack. And that’s where artificial intelligence comes in. Scientists applied methods similar to natural language processing and created a machine-learning model which is able to  determine the amino acid sequence of neoantigens based on the one-letter amino acid code. This opens the door to creating an AI-based tool to create personalized cancer vaccines. Basically, scientists can train the algorithm using normal peptides to predict the mutated peptides. Machine learning model called DeepNovo is predicting  amino acid sequences of neoantigens.

To achieve this scientists  downloaded the immunopeptidome datasets of five patients with melanoma. This information helped them to train the model and validate its results. Result – a very capable machine learning algorithm that can identify specific neoantigens for each individual patient to provide personalized treatment and care. Ming Li, one of the authors of the study, said: “Cancer immunotherapy is quickly becoming a fourth modality of cancer treatment, alongside surgery, chemotherapy and radiotherapy. Every patient is different and every cancer is different, so cancer treatment shouldn’t be the same for all. Treatment should be tailored to the patient and that’s what our personalized machine learning model allows us to do.”

Personalized treatment one day will significantly improve survivability of cancer. And we have to believe that this kind of future is really close already, because results of current research are amazing.


Source: University of Waterloo