Study of a cancer patient’s tumour DNA, and the mutations within it, gives us a window into that specific tumour
A human’s DNA offers valuable insight into their biological characteristics. In addition to coding external characteristics such as height and eye colour, genes are also responsible for telling our cells what to do: some genes control hair growth (e.g. telling our follicles to divide), some tell a damaged cell to die (e.g. when infected). Mutations in these genes cause cancers, as cells no longer receive the correct instructions on how to behave and start to grow out of control, forming tumours. Study of a cancer patient’s tumour DNA, and the mutations within it, gives us a window into that specific tumour; allowing scientists to understand the specific biological processes driving it. From this, we can tailor a treatment strategy unique to the cancer and patient: this is called personalised medicine.
DNA is found in cells in the human body. Traditionally, these cells are sampled and the DNA extracted and ‘read’ (sequenced), in order to decipher the DNA code and get a picture of a person’s genetic makeup. In the study of cancer, cells are usually taken from the tumour tissue itself and compared to non-cancerous cells in the patient, allowing immediate insight into which genes are different (i.e. are mutated) in the cancer.
Understanding the genetic mutations in a cancer patient is highly valuable. If we are able to know the exact mutations that code a cancer, we can find out really useful things: is the mutation becoming more prevalent over time? is the cancer becoming more/less complex (more/fewer mutations) throughout treatment? are there therapies that specifically target ‘driver’ mutations?
Taking a (solid) tumour tissue sample is an invasive surgical procedure. This means that, in order to preserve the safety and comfort of the patient, samples are usually only taken before and (sometimes) after treatment. Relying on solid samples therefore means that we lose out on the incredible information available during a cancer patient’s treatment. Finding a way to reliably sample the tumour, without physically having to cut some out, is a true game-changer.
Interestingly, when cells die, they burst and the contained DNA is released into the blood, where we can detect it. In a cancer patient, both normal and tumour cells will release DNA in this way, meaning that we may no longer have to rely on tumour tissue samples in order to investigate cancerous DNA. Through a routine blood-test (a liquid biopsy), scientists can extract DNA ‘floating’ in the blood (called cell-free DNA) which can then be compared to the patient’s normal and tumour solid samples in order to distinguish between the different types of DNA, and identify which originates from the tumour. Blood tests are minimally invasive and a cancer patient will have them periodically over the course of their treatment (circa every 2 weeks). The clinical insights gleaned from regular biopsies are highly valuable, decreasing the window of time between tumour ‘observations’ from months to weeks. Over the course of a typical (6 month) cycle of chemotherapy, that’s up to 10 more time-points at which to analyse a patient’s response to treatment. However, without useful interpretation, a clinician can’t act upon these insights: equipping a doctor with the technology to easily understand and act upon changes in a tumour’s DNA, could be life-changing.
Analysing tumour DNA from a liquid biopsy carries with it a unique set of challenges. Indeed, it is often a challenge to even identify the tumour DNA itself. If you compare the size of a tumour to a human body, that is the ratio of cancer to non-cancer DNA in the blood: a cancerous buoy floating on a rough sea. This is, at its heart, a statistical problem and one that we are working hard to solve. Advanced analytics are often needed in order to be sure that we are tracking cancer DNA mutations over time, rather than other random mutations or even statistical errors. This process (deemed ‘mutation calling’) is usually accomplished by comparing an initial blood sample to a known (solid tumour) sample, to work which of the mutations present originate from the tumour itself. Statistical analysis takes many forms: from using top-of-the range machine learning algorithms that trawl through the data, to deciding what is statistically significant in order to warrant further study.
This is a big challenge, but one we, at Cambridge Cancer Genomics, feel well equipped to solve. Combining less invasive liquid biopsy sampling techniques with intuitive artificial intelligence and machine learning led software will improve DNA sampling and analysis, resulting in better, more considerate, personalised cancer care: for everyone.