AML Registry to transform blood cancer research with the help of AI

Acute Myeloid Leukaemia (AML) is an aggressive blood cancer that spreads quickly and as a result requires immediate treatment.

Although a rare condition, it is still one of the most common acute blood cancers. But outcomes for the disease have traditionally been poor, despite some success in randomized trials. Therefore, says Professor Charles Craddock, Professor of Haemato-Oncology at Warwick University and Chair of the AML Research Network:

The UK AML community realized a few years ago that by collecting high-quality prospective registry data, we could get really quite profound insights that would teach us how to improve treatment outcomes. So, at the same time as working hard to deliver a platform of prospective randomized trials, we recognized as a group that we needed to identify funding to set up an AML Registry.

Such registries, or databases, store both patient and disease information. Doctors and researchers can then analyze this data to improve medical care for those affected. 

In the case of AML, an additional hope is that by making such information more widely available, it could help reduce the patchiness of current care in the UK. A further aim is to address concerns that members of ethnic minorities have “differential access to either targeted therapies or transplants”. 

Nationwide clinical trials

As for the significance of the data in trials being prospective, Craddock explains:

So, we say ‘right, from now on, we’re going to collect high-quality data on these patients going forward’. And that allows us to set the fields for the questions we’re asking. It’s very important because it makes it rather more representative. The danger with retrospective data is that there’s lots of selection bias as you may only collect data on people who did well or people who did badly. 

The Registry’s aim is to collect anonymized data, such as age, gender, presenting clinical features and outcomes, from patients attending one of 30 planned centers in large metropolitan hospitals across the UK. 

These centers will be run by Accelerating Clinical Trials (ACT), a social enterprise set up by Craddock specifically to focus on blood cancer. ACT will fund a national network of research nurses who will gather information from patients and enter it into the Registry. They will have a virtual meeting every two months to share information, which includes any problems, patient concerns and how to allay them.

By the end of this year, the goal is to recruit around 500 participating patients annually. This means it should be possible to “make some really important observations” after as little as two years, Craddock believes.

Finding hidden patterns

Another Registry aim, meanwhile, is to bank patients’ blood or bone marrow when they are first diagnosed. The idea here is to use Corpora.ai’s research engine to correlate the frequency and patterns of particular gene mutations with clinical outcomes. This is valuable, says Craddock, because:

What it allows you to do is explore associations between both clinical and molecular characteristics of AML in big data sets in a much more efficient way than we can do at the moment or we’ve previously been able to do. So, it allows you to look for associations that otherwise haven’t been made.

One of those associations will be in determining how well an individual is likely to respond to treatment:

Let’s say it’s a young woman who’s interested in having children. Does she have a particular gene signature that suggests she can be cured just with chemotherapy? Then she can avoid a transplant. And that really matters because when you have a transplant, it affects your fertility. Let’s say we have an older patient. Should he get intensive treatment? Should he get more gentle treatment? Or maybe we find a gene signature that says he’s not going to benefit from treatment at all. So, we wouldn’t give intensive treatment, and we’d save him a lot of suffering.

Other data the Corpora.ai system will analyze include clinical literature, biomarkers, and treatment pathways to help develop earlier diagnosis models and optimize clinical trial design. Corpora.ai is owned by serial technology entrepreneur and investor Mel Morris, who also founded the MSTART Medical and Sporting Technology Research Trust. 

Working with different registries

MSTART has donated some £250,000 of the total £2 million necessary to set up the AML Registry. The funding will also go towards meeting the cost of ACT running the data collection centers and the creation of a new three-year data science scholarship in the School of Data Science at the University of Warwick Medical School. 

The aim here is to train a PhD student to work in database design and machine learning so they can explore how AI might best be used in a Registry context. This includes developing models to interrogate data sets in new ways. Craddock adds:

The opportunity is to work with data from a range of registries, including ours. For example, at Warwick, there’s a very important registry that looks at patients who die out of hospital, that is, sudden death out of hospital. So, in these different disease settings, it’s about can we think about how to use AI to maximally generate patient-informing, treatment-informing data?

The AML Registry will also be linked with the Research Registry of Genomics England, a government-owned company that analyzes sequenced genomes for the NHS. This relationship will enable it to work with the complete set of human DNA data. 

But Registries can also provide other benefits beyond simply data analysis. According to Craddock, there is a lot of evidence to demonstrate that:

By putting patient data into a registry, you actually lift up the quality of care. This is because people start doing things in a consistent manner and you embed within your data set some guidelines about how you treat particular complications, or particular sorts of leukemia. So, it’s increasingly being recognized as a way of improving patient outcomes.

Making the Registry sustainable

The next step in making the Registry sustainable into the future will be to enable third parties, such as pharmaceutical or cell therapy companies, to pay for access to AML Registry patient data. Craddock explains:

The anonymized data we collect really gives quite valuable insights in how patients with particular sorts of AML behave. Do they respond to therapy? How equally are they getting treatment? This is increasingly important for drug companies when they put in a submission to Health Authorities to get licensing. So, we’re planning to ethically monetize these data sets because we’re generating data that gives insights into how they can best develop drugs. But also because the drug companies need those data sets if they’re going to get their drug approved.

One possible income route might be for pharma companies to offer grants in exchange for access. A second could be for the Registry to use its research to apply for grants from major institutions. A third would be to charge a “reasonable” fee to unrelated donor stem cell registries keen to understand if they are reaching underserved communities.