ctDNA and AI are the way forward in early cancer detection and metastatic treatment
Posted 24th June 2020 by Joshua Sewell
We spoke with Dr David Guttery about his pioneering work using circulating tumour DNA to enable early detection, monitoring and therapeutic decision making for better patient outcomes in gynaecological cancers.
As someone who was ahead of the curve in utilising liquid biopsy technologies in cancer research, what lead you to this field of research?
I was previously Senior Research Associate in Professor Jacqui Shaw’s lab looking at the utility of the liquid biopsy for the early detection of breast cancer. We made a number of landmark discoveries with liquid biopsy technology, and showed that acquired mutations in the oestrogen receptor gene can infer resistance to aromatase inhibitors.
This lead me to take these findings further in collaboration with Dr Esther Moss, with whom I was able to show that the discoveries had translational value for prognostic and therapeutic decisions in endometrial cancer.
I discovered that acquired microsatellite instability in patients with endometrial cancer can be detected using circulating tumour DNA (ctDNA), with huge implications. The discovery means that patients can be switched to more effective immuno-therapies based on analysis of blood alone.
What role have liquid biopsies and new sequencing technologies played in your research?
Over the last five years liquid biopsy methods have really exploded. This is mainly because taking a tumour biopsy is often quite difficult for patients. It’s also not representative of tumour heterogeneity.
With patients with metastatic breast cancer particularly, they are just too sick to undergo multiple biopsies. It’s very difficult to monitor how they are responding to treatment until it’s overt that they are progressing and usually by then it’s too late. So, the liquid biopsy can be used to detect disease progression earlier, more easily because it’s a simple blood test and because it’s far more amenable to patients.
Our research has used targeted next generation sequencing panels to look at key cancer drivers and determine, based on the patient’s mutational profile including their levels of circulating tumour DNA and mutations, whether they are responding to treatment; and whether we can stratify them to more effective targeted therapies.
Are there any ways you would want to see liquid biopsy technology and methods developed?
It’s well known that in the metastatic setting ctDNA has real utility and you can use it for monitoring patient response and determining patient therapies.
There are several ongoing clinical studies that are looking at intervening in a patient’s treatment based on their ctDNA profiling. The real challenge is in detecting cancers at the earliest stages, stage one breast and womb cancers are particularly difficult to detect.
In breast and endometrial cancer particularly, it is very difficult to spot or detect ctDNA because of the limitations of the technology available. The sensitivity of the technologies in particular needs to be improved.
One question is whether early stage cancers are just so small that they don’t shed enough DNA into the circulature to be able to pick it up and therefore alternative biomarkers and other means of detection is an area of development. Perhaps something that could be explored is the use of different fluids. For example, in cerebrospinal cancers or neuroblastomas, cerebrospinal fluid is far more effective than blood for spotting the cancers early.
What were the key hurdles to overcome in the translating work on breast cancer into work on endometrial cancer?
There were a couple of studies that came out which showed that ctDNA sampling wasn’t the optimal method for spotting endometrial cancer early on, so we tried to translate ctDNA into the more advanced and recurrent settings.
With endometrial cancer the cure rate is very high, so people haven’t really focused on it. But when endometrial cancer relapses, the prognosis is very poor. When it relapses, it metastasises to distant sites and the prognosis is about a year.
So, we wanted to take everything we had learnt from breast cancer in both the metastatic and relapse setting and translate that over to endometrial cancer. We have well established protocols and validated methods in breast cancer, so we wanted to see if this was translatable and the early findings are that it is.
We are seeing that ctDNA has real utility in the relapse setting. We can predict whether a patient will relapse; spot the relapse coming earlier than current methods of gynaecological examination or scans can and potentially determine what treatment a patient should be put on, based on their molecular profile.
What role has machine learning played in analysis and prediction in your group’s research?
In breast cancer Artificial Intelligence (AI) has been implemented to improve the accuracy of mammograms. At present usually only one out of every ten patients recalled due to an irregular mammogram will have a cancer. AI is being used to improve that. Machine learning algorithms are being employed to enable programs to take the scan, look at the scan image on a pixel level and look for patterns that are specific for tumours. AI is also being used to analyse histopathological slides and identify cancer types and tumour sub-types.
The research we’re doing is taking the histopathological slides and trying to develop AI programmes which predict whether the patient’s risk of relapse and therefore stratify their care, ensuring the most effective use of vital resources. If the patient isn’t at high risk, they don’t need to be followed up so often. But if they are at high risk, we can know to monitor those patients more often, take blood samples more often and use ctDNA to track them and spot relapse sooner.
When it comes to clinical applications, what are your key hopes for the future and next steps?
For me, the key hopes focus on ctDNA application in the metastatic setting and AI application in early stage cancer diagnosis and treatment.
The next steps are for large prospective clinical trials being set up for ctDNA; and a key hope is to see ctDNA sampling implemented in the clinic. It’s widely recognised that in the metastatic setting ctDNA is the way forward, so within the next five years I would like to see ctDNA techniques implemented regularly in the clinic to monitor patients who have metastatic cancer.
In early stage disease settings, I would like to see AI implemented more to improve diagnostic accuracy.
ctDNA will become mainstream in the next two to three years in the metastatic setting. AI has a way to go, I think, but probably within the next ten years we’ll see AI tools being used more commonly in the clinic resulting in better patient outcomes.
Those are my two big hopes at the minute anyway!
David Guttery is Lecturer in Cancer Early Detection in the Leicester Cancer Research Centre at the University of Leicester.
Leave a Reply