Public Funding for Synteny Biotechnology Limited

The NHS (and every healthcare provider globally) are struggling to improve outcomes for people with cancer. People are far more likely to survive cancer if it's detected before it advances and spreads, because treatments work better. Only 55% of people in the UK currently seek treatment when their cancer is at StageI or Stage II. Introducing screening programmes for more types of cancer could improve this. However, 80% of cancer deaths are from cancer types where there is no reliable screen, or screening would identify too many people with cancers that would not shorten their life. When a person has cancer, their immune system produces characteristic types of T-cells in response. Genes in these cells have 'shuffled' to encode new types of receptors in response to the cancer. Because genetic sequencing has gotten much faster and cheaper in the last 20 years, it is now possible to sequence sampled T-cell receptor (TCR) DNA for <£200\. Researchers have already shown that new TCRs are produced before a cancer can be detected reliably using any other method. However, the TCRs that may be produced in response to cancer vary infinitely amongst different people. We could not simply look up in a database if a particular TCR gene sequence indicated that someone had cancer; the TCR sequence would almost certainly be different to everything in the database. This project aims to use machine learning to determine if particular TCR sequences indicate the presence of cancer and in which body tissue that cancer exists. We first use datasets to build basic AI tools and then use biobank tissue samples, taken at various times before someone developed cancer and from people who remained healthy, to generate novel sequences. We can then test how well our tool performs against these novel sequences. We have already used publicly-available data from breast cancer patients to validate the immune signal of our initial model. In this project, we will use lung and colorectal cancer samples and data to extend our capabilities further and validate using samples collected prior to diagnosis. Ultimately we want to develop a tool that costs below £200 and can detect 95% of early stage cancers with 98% specificity and correctly identify in which organ the cancer is 95% of the time. In combination with emerging stratified treatments (e.g. immunotherapy; tailored drug cocktails) it will prove transformative for cancer survival, particularly ovarian, pancreatic and liver cancer.