Immune profiling ‘will be a revolution in medicine’ says Prof Adrian Liston at Babraham Institute
A revolution in medicine is coming.
It could aid the diagnosis of diseases, guide the way patients are treated and inform the discovery of new therapies.
Immune profiling seeks to explain how our body’s own defences are affected by and are responding to disease.
At the Babraham Institute, Professor Adrian Liston is working on the translation of this technique from the laboratory to the clinic.
“The immune profile is much more powerful than genomic data, but it’s much easier to get genomic data,” he tells the Cambridge Independent. “You can take blood, send it overnight and get it sequenced off-site. We are not at that stage with immune system data.
“But the more we know about different diseases, the more we realise there are inflammatory, or immune-mediated, components.
“It can be a revolution in medicine. Once the infrastructure is set up and hospitals are doing the analysis routinely, we will see an explosion in utility. Right now, it’s a research tool only.”
Prof Liston has been leading work at Belgian research organisations VIB, KU Leuven and UZ Leuven to apply machine learning algorithms to immune system data that could enable the diagnosis of juvenile arthritis with a simple blood test.
The findings, published in Annals of the Rheumatic Diseases, could also be used to predict which patients would respond best to different treatment options.
“The aim of the study is to see whether the immunological signature – something you can pick up in the blood – can be used to show someone has arthritis. This is the first step towards being able to use blood-based biomarkers rather than clinical symptoms,” explains Prof Liston.
Juvenile idiopathic arthritis is the most common rheumatic disease in children, but a young patient can face a long route to proper diagnosis, as the condition presents in different forms and severities.
Once a GP has recognised that the case involves something more than ‘growing pains’, the patient will be sent to a hospital’s paediatric department to rule out other causes before seeing a specialist rheumatologist. It might take another six months to understand which sub-type of arthritis the patient has, and then further time to try out different treatment paths.
“If you could diagnose juvenile arthritis with a blood test at the GP test, you could get patients to a paediatric rheumatologist faster,” says Prof Liston.
“The other advantage of knowing the data blood signature in the patient is that it can tell you about the mechanisms of how the immune system is attacking the joints.
“It is not a very well understood disease and the first part of understanding it is to see what is going wrong in the immune system.”
Blood samples from more than 100 children – two-thirds of whom had juvenile arthritis – were collected by the researchers in Belgium. Continuing the analysis at Babraham, the team studied in unprecedented detail the levels of each type of white blood cell, which are critical to our immune response.
Acting as our standing army, white blood cells – or leucocytes – defend us against bacteria, viruses and fungi, breaking them down, binding to them or destroying them. They produce antibodies, destroy cancer cells and other parasites and form part of our allergic response.
While often grouped into five broad categories, there are in fact about 100 types of white blood cells and each was analysed in this study.
“A clinic might do a white blood cell count but it tells you nothing of which of those different types are more or less common,” says Prof Liston. “We measured the numbers of each type and tried to see which ones were up or down, and which responded to treatment.
“We took the information and ran it through a machine learning algorithm. The program you build starts with no knowledge. It’s a brute force approach. It might focus on four parameters, then the next four, and it does it millions of times. It trains itself.
“You take the algorithm and reverse engineer it, working out what factors it was using. It could be using all 100, but here it found four or five are important and the other 90 or so are not so important. This is important because this level of analysis is beyond what most hospitals could do. They can’t assay 100 cell types. But they can assay three or four.”
The test proved 90 per cent accurate at diagnosing juvenile arthritis.
“This result is a proof-of-principle demonstration that immune phenotyping combined with machine learning holds huge potential to diagnose different forms of juvenile arthritis early in the disease course,” says Prof Liston.
A replication study would be needed before a company could develop and commercialise the system, but Prof Liston believes the biggest benefit of this approach could be in providing a prognosis for patients, rather than a diagnosis, to improve the selection of patients for specific treatments or clinical trials.
“If a patient is known to have a very aggressive form earlier, you can start with an aggressive treatment earlier, and you are going to save them some pain,” he explains.
“Our study was cross-sectional, taking patients at one time point. If we did the same longitudinally on patients coming in before and after treatment, we could start to understand how each patient should be treated in a specific way according to their immune signature.”
This pioneering approach could also be used to provide biomarkers of other diseases or predict how patients will respond to treatments, aiding the drive towards personalised medicine.
“This is the first time machine learning has been used on the immunology of the blood,” says Prof Liston. “We are actively pursuing this in other immunological diseases, in cancers and in adverse response to immunological treatments.
“I think we will start to see a lot of immune profiling and machine learning in the future.”
A complete analysis of the many billions of cells in our immune system is some way off still, although scientists are trying various approaches. Cataloguing all the parts is predicted to produce 100 billion times more data than the Human Genome Project.
Currently, outside of specialist research clinics, there is little analysis of patients’ immune systems.
“One of the key reasons is how the data is analysed,” says Prof Liston. “The sample treatment analysis hasn’t been simplified to a level where it becomes a cheap, robust diagnostic. This is something we are actively working on: how can we make this simple, fast and reproducible to the point where clinics will start adopting it?”
One area where immunology has been having a major impact is in cancer treatment.
Checkpoint blockade therapies are designed to remove the brakes on the immune system that are applied by tumours, unleashing the full force of the body’s own defences on the cancer.
But the therapies are very expensive, do not work for all patients and can cause adverse reaction.
Prof Liston believes immune profiling could greatly improve our use of such treatments.
“Immunoncology treatments are based on immune discoveries from 10 to 15 years ago. In a way, it’s old technology,” he points out. “This is a key area where we need to understand who should get these immune blockade therapies.
“At the moment we are giving these treatments fairly blindly because we assay the immune system for patients very crudely. It could be much more sophisticated, which means we would be targeting the right patients and we would lower the risk of adverse effects. It would be an enormous advantage.”
There has also been excitement over the prospect of cancer vaccines.
In the case of cervical cancer, the childhood vaccination now given to girls in this country should eradicate the type caused by the human papillomavirus (HPV).
But there are relatively few cancers caused by viruses or bacteria, so most cancer ‘vaccines’ are actually immune system ‘boosters’ given after diagnosis. Success, however, has been limited.
“The problem is the cancer has these checkpoints that stop an immune response. That is why the checkpoint therapy works so well – you are taking away that checkpoint.
“In future, I think we might see a combination of cancer vaccines and checkpoint blockade therapies,” predicts Prof Liston. “That will be a boost to the immune system and you are taking the brakes off at the same time.”
While retaining links with his lab and students in Belgium, Prof Liston is now based at Babraham, where improving the prediction of adverse responses to immune checkpoint blockade therapy is one of a number of aims.
“In my patient work, the focus is on Alzheimer’s. Can we predict the onset of Alzheimer’s in the same way we can diagnose juvenile arthritis?
“It looks like there may be an immune role in some of these neurodegenerative diseases, which means we might be able to pick them up with a blood test,” he said.
“And more generally we would like to work on getting immune analysis into the clinic, making a routine tool not just for research.”
Much of his time will also be spent helping us understand the workings of our immune system better.
“I’m interested in what cells do when they leave the blood and go into tissues. It’s a hidden part of immunology,” he says. “We work on the blood because it’s available, but very few infections are actually blood-borne.
“The vast majority of infections, auto-immune diseases and responses or inflammatory diseases are in the tissues. And yet we don’t really know what is happening there.
“Here you have to go into a mouse-based system. In humans it’s very hard to get access to tissues, of course.
“We are trying to reinvent immunology from the ground up at tissue level, rather than at the blood level.”