Presented by: Jens Kringelum, Director of Genomic Immuno-Oncology at Evaxion Biotech
Transcribed by: Tia Byer
The Genomic Immuno-Oncology department of Evaxion Biotech aspires to become a world leader in AI immuno-oncology, through their work in decoding the human immune system. Currently, Evaxion Biotech is further developing their dynamic and highly accurate computer models to predict the immune system’s reaction to the influence of bacterial and viral pathogens, and various cancer neo-epitopes. As part of Immuno Week: Online held in July 2021, this article will look at how artificial intelligence (AI) and machine learning (ML) can be used to generate patient-specific immunotherapies.
The Human Immune System: The Body’s Weapon Against Diseases
The immune system acts as an effective weapon against diseases. Not only is it built to fight infections, but it also surveys all healthy cells, making sure that we do not develop cancer. Throughout life, we all tend to have some cancerous cells in our bodies. The immune system can detect and eliminate these; however, cancer can still evade the natural defence mechanism. Immunotherapies, therefore, become a tool and strategy used to correct this fault of the immune system.
Patient-Specific Medicine: The Next Frontier in Immunotherapy:
Traditionally, for 100s of years, maybe even 1000s of years, doctors and scientists had a ‘one medicine fits all’ approach. In the last century, this has involved identifying a disease, finding its target, and then developing a drug that was assumed to produce a certain effect in all patients. However, modern science indicates that this is not the always the most effective approach for all patients. Some patients react better to medicinal therapy than others, whilst some do not react at all or even experience adverse effects. Now, scientific practice looks towards designing therapies and pharmaceutical products to address the individual and unique needs of a patient. Patient-specific medicine takes into account an individual’s genes, proteins, and immune system and aims to create the best possible treatment for each individual patient.
Neoepitopes: Unique Targets in Cancer
AI is driving patient-specific medicine and is increasingly applied for the intelligent design of cancer immunotherapies. One approach is to identify cancer specific epitopes, i.e. small and subtle differences that reside on the surface of cancer cells and not healthy cells. One class of these epitopes are referred to as neoepitopes and arise from mutations in cancer cells.
When a cancer cell develops a mutation in its DNA, some of these gets translated into protein products and eventually degraded by the proteasome and potentially displayed as small peptide fragments bound to MHC molecules on the cancer cell surface – a mechanism referred to as the endogenous antigen presentation pathway which are active in all cells. As the mutation are only present in the cancer cells the mutation will only be displayed on the cell surface and not normal cells. This subtle difference enables the immune system to differentiate between cancer and normal cells and, if induced, specialised cells (T-cells) of the immune system will kill cancer cells uniquely. As the immune system systematically surveys the body’s cells, it would detect these altered mutations in multiple different tissues making neo-epitope treatments ideal for cancer treatments. The strategy for neo-epitope treatments is thus to identify cancer specific neo-epitopes in a patient cancer cell by genomic sequencing and AI computer models and subsequently manufacture an immunotherapeutic drug that can induce a neo-epitope specific immune response in patients.
AI Driven Immunotherapy: The Road to Patient-Specific Immuno-Oncology Treatment
To help restore the ability of the immune system to eliminate cancer cells, Evaxion Biotech has developed its PIONEER system to identify neo-epitopes and other unique cancer epitopes. The automatic system aims to only identify potential targets that are unique for the cancer cells, and not found on normal and healthy cells. Epitopes are particularly useful targets because they are not known by the immune system. Therefore, it is possible to make a vaccine or engineer T cells to specifically recognise the neoepitopes without worrying that they will affect the normal cells.
As part of the PIONEER system, Evaxion Biotech first collects a tumour or blood sample from a patient. The sample is then processed by the system which identifies the epitopes. A precision cancer therapy is created and administered to the patient. Due to the nature of cancer being an aggressive disease, these steps must be completed within a quick timeframe to have a reasonable chance of helping the patient in question.
Challenges in AI Driven Precision Medicine
There are several important considerations to bear in mind when applying AI driven patient-specific therapy to cancer. Firstly, there is a risk of biopsy impurity. In a tumor lesion not all cells are cancer cells and in many cases, there are only 15 to 20 % cancer cells in a tumor biopsy used for designing a patient specific treatment. This needs to be taken into consideration when designing algorithms that identify cancer specific mutations.
The next challenge is to investigate the DNA sequence of the tumour and healthy cells. There must be a balance between the speed and quality of the sequence. Sequence depth and granular quality is also important to consider. Finally, the AI system must be able to identify both the cancer mutation and what exactly the immune system reacts to. A high level of quality control is important as the AI-systems used are directly influencing the design of the drug administered to patients which efficacy cannot be tested before administration.
Forward Outlook: What’s Next for AI Driven Precision Medicine?
For some time, it has been known that it is not only the peptide:MHC affinity but also the peptide:MHC stability that plays a significant role in what resides on the surface of cancer cells. Previous technologies restricted to measuring the peptide:MHC stability have proven of limited use in practice. In response, Evaxion Biotech has developed an assay using mass spectrometry to detect the stability of a peptide from naturally processed ligands. Using a novel ML model, Evaxion Biotech was able to demonstrate peptide:MHC stability as a measure of neo-epitope efficiency superior to traditional affinity and mass spectrometry detected MHC ligands. Evaxion Biotech is currently in Phase 1/2a with two programmes based on the Pioneer system. We look forward to seeing the next advancements made by this technology over the coming years.