Main content

How AI + genomics could equal a future of predictive, personal healthcare – with caveats

Chris Middleton Profile picture for user cmiddleton May 30, 2024
Summary:
Where is the sweet spot between AI and genomic data? The UK’s leading experts explain the key issues

Silhouettes of people mental health teamwork and communication concept and connected lines with dots © GrAl - Shutterstock
(© GrAl - Shutterstock)

Few would deny that we are living in an AI Spring, yet many of the most popular advances have been in areas that are of marginal importance. While Large Language Models, chatbots, and generative systems can unlock the value of trusted data sets, their dominant use to date - in the public imagination, at least - has been in solving non-existent problems, such as avoiding the need to pay talented, skilled, creative humans for their work.

However, one sector that will be of critical importance in the years ahead – as populations age and we strive to find cures for all manner of medical conditions – is healthcare. Particularly in areas such early diagnosis, predictive care, and personalized medicine. 

At the center of that opportunity is genomic data - where the UK is a world leader, thanks to its integrated public health system, free at the point of need - plus AI and machine learning.

But one of the challenges for the NHS is the prevalence of market forces. So believes Professor Rameen Shakur, Director of the Brighton Integrative Genomics Unit, and Director of the Centre of Research Excellence in Precision Health and Translational Medicine at the University of Brighton.

Chairing a recent Westminster Forum debate on advancing genomics research, he said:

The landscape is evolving, there are societal pressures and there are information digitization pressures. But one of the aspects which is not often discussed – because we have a reductionist mentality, of sorts, in service delivery – is market forces. Commercial forces play a much bigger role than you might think.

Could he expand on this point? He could:

We currently have 75,000 different genetic tests available globally, with 10 genetic tests coming out every day - mainly in the Far East and the United States. But private companies are driving this. So, we can't look at that and say, ‘Patience! We'll just look at the NHS as the only availability’, because the commercial element of that will drive a lot of those processes and demand.

He added that rolling out genomic programmes to local communities - rather than merely in centralized, specialist facilities - is another challenge, though the falling costs of both testing and technology make that more feasible.

However, the prevalence of market forces is not necessarily a bad thing for public healthcare, according to the event’s keynote speaker.

Dr Rich Scott is CEO of Genomics England, a company set up by Britain’s Department of Health and Social Care to run the 100,000 Genomes Project – now completed – which sequenced the genomes of 100,000 NHS patients, with a focus on cancer and rare diseases. The government’s aim now is for five million genomes to be sequenced, and Genomics England has and set up a range of partnerships to develop new treatments as it continues its work.

Before taking up his post, Scott worked as a specialist genetics consultant at renowned children’s hospital Great Ormond Street in London, helping to diagnose and treat rare conditions. Describing the NHS as being at the center of a rich, diverse ecosystem - as opposed to the fragmented, private insurance-based US healthcare market - he said:

We should celebrate the journey that we've been on. But I think we should be really dissatisfied with how far we have got, because there is still so much to do. That restlessness of not accepting that today is good enough is really important. 

We need to work through how, together, we can design what health systems and the use of genomics should look like in the future, because this isn't something where you disappear into a dark room and come up with a single, simple answer.

But there is enormous promise. If we develop the evidence in the right way and build the right infrastructure, and if we develop the expectations together - and that means together with the public and all of those who could benefit from the use of their genomic data - there is so much more that we can do.

Our vision at Genomics England is a world where everyone benefits from genomic healthcare. And that's really speaking to genomics becoming part of the fabric of day-to-day healthcare provision.

A bold vision, with other players in the ‘together’ he referred to including NHS England, the UK Biobank, universities, and seven out of the top 10 pharma companies - an “infinity loop” of data, innovation, and iterative improvements, he said. 

So, what of the role of new technologies such as AI in that process? Scott said:

Where we need to place our effort is particularly in the management of data at large and at national scale, which at the moment is predominantly focused on whole genome data. And making that available in systems and platforms that others can access to do their great work.

This has repercussions throughout the public healthcare system, he explained:

Each time we design a research programme, we think - again - carefully about the right models. Before the launch of the NHS Whole Genome sequencing service, for example, we spent time thinking and exploring via public dialogue the right model for consent. 

There aren't single right answers: there are complexities and practical delivery questions in any model where it is critical is that it is widely accepted by people receiving the care, because we need that that breadth of adoption. 

For it to have the impact we want, this isn't something that should be the preserve of a small vanguard of people. It needs to be something that people are generally comfortable with in the context of their routine care

On this point, there is good news, he said. 

From the people we see in clinic having whole genome tests for rare conditions, or cancer in the NHS Whole Genome service, more than 90% of people who are asked, say yes to the research.

He added that this is one field where a partnership between a public healthcare service and private providers can truly work for public benefit - despite the political sensitivities involved, perhaps.

This is an example of how we work with the NHS, who provide us with the bulk of the work and the people-power - in terms of the Whole Genome sequencing service, the clinical side, and the laboratory side for DNA extraction and expert interpretation of the findings. Then we partner with [San Diego, California-based biotech company] Illumina for the whole genome sequencing itself. 

And we provide the bioinformatics and the data analytics platforms that can process the data - in a coherent way that can be held centrally and re-looked at when discoveries arise after the initial analysis. And, where people consent, that can also be deposited into a research infrastructure in a coherent way.

Powerful, rich datasets

So, how does this partnership between the public and private sectors protect patient data, and prevent the NHS from being exploited for commercial gain without permission? He explained:

With people signing up to join our research cohort, the National Genomic Research Library, we do expect researchers to run their own projects. They can discover things that might end up in research papers, for example. But they can't export individual patient datasets. 

However, they can export – through a carefully curated airlock process – summary data and the sort of data that would be expected to go into a research paper, which protects privacy in accordance with participants’ expectations.

Critically, they can also recontact patients, participants, or their clinicians where that would be useful to move a discovery forward, or pass stuff back that is important to the participants’ care. 

We also partner with others, particularly with the NHS and the Office for National Statistics, to bring in other data alongside the genomic data to sit in that trusted research infrastructure. So we can see the longitudinal healthcare experience of those participants and carry out research that would not be possible without more than just the generic data. 

And I think it's that richness of data that makes the UK research datasets particularly powerful.

To date, more than 150,000 whole genomes have been sequenced – more than 60 petabytes of data - and more than 500 researchers a month access that information across academia and industry, with more than 600 projects ongoing at any time.

So, what of the technology of the day: artificial intelligence, which - while offering enormous promise-– has often been associated with more frivolous applications? 

Scott said:

In the last 11 years, since Genomics England was founded, the range of potential avenues for genomics has been expanding and improving, and have really changed. 

There are a large number of organizations involved. For example, Google DeepMind has provided AI-assisted algorithms that can simplify many currently manual steps, particularly in the interpretation of genetic variations. 

We see the potential and have read all the research papers about how the advances in Large Language Models, and AI more generally, will help. So, we are in a period where we can see that AI and machine learning will - or could - bring some really substantial changes in truth at the moment, for the current uses of genomics.

But he added:

We're in the foothills, and there are lots of questions for us to work through. This isn't just about tech. It's also about what safeguards and expectations we want to put in place.

For example, if you're developing one of these models, has it performed for different communities with different ancestries? Is it equitable, or is it not if it's just been trained on particular datasets? 

And second, if you're developing an algorithm, how do you make sure you protect the privacy of those whose data has been used in its development, especially when that algorithm is going to be used elsewhere? 

The devil is in that detail, and there is important stuff that, together, we need to navigate.

Addressing the needs of patients

In spite of this, Scott shared some recent discoveries that are extremely promising:

Two groups simultaneously have identified a new, non-coding gene that we think relates to previously undiagnosed children's and young adults' rare developmental conditions. It's non-coding, something that you wouldn't have been able to look for using other technologies. And it's come through discovery across the dataset at large scale.

Because of this, we know that hundreds of families have been diagnosed [with this condition] internationally, simply because people have been able to look. So that's an example of being able to look back at existing data and discover new findings!

And we are looking, in cancer, at the use of imaging data alongside clinical genomic data to see what more discoveries we can make - for example, around the ability to use images to predict genomic and other biological features, something that could perhaps better stratify treatment very quickly. 

We have a diverse data programme, to make sure that our programmes are all engaged with - and are inclusive of - the full range of communities in the UK, and that the National Genomic Research Library is representative of UK communities more generally.

He added:

We are also looking, with our participants, to develop evidence around whether or not whole genome sequencing should be offered to all new-born babies, driven most acutely by its potential use in screening for 200-plus conditions. That programme is just launching now. 

Whole genome data is part of a spectrum of genomic data. And increasingly, I think Genomics England is about the value of that data, holding it together in a coherent way to bring together a picture that can either drive better healthcare immediately, or better research and discovery in the longer term.

At the moment, we are focused particularly on diagnostics, in the use of genomics in routine care. But, as I said, there are questions around the potential use of whole genome data in new-borns, which is more about the prediction of severe, treatable conditions 

There's also lots of work in cancer in early detection and diagnosis, and in tailoring treatments and monitoring. And enormous potential in pharmacogenomics. When you look across that suite, we believe that, if the evidence plays out, genomics could be very much part of the fabric of routine healthcare.

Excellent news. 

Professor Emma Baple is Professor of Genomic Medicine and Medical Director of the South West Genomic Laboratory Hub at the UK’s University of Exeter. She explained:

My belief is that successful collaborative partnerships between academia, industry and the NHS are essential for us to ensure that the research we're doing is actually addressing the needs of patients and families.

In a single test, we can already look at 6,000 rare genetic conditions and return a diagnosis in less than 10 days. But we can't do that for every patient with rare conditions, unfortunately, because we haven't got the capacity and capability. Even so, this test is life-changing for children.

However, she warned that the absence of data, capacity, and funding is a “valley of death” for some, adding:

I'm passionate about the NHS and what it allows us to do for patients. It really does enable us to involve patients at every point in the research journey. And it's unique, there isn't anything else like it in the world. We should be very proud of that. 

And this also allows the international research community to access that data for patients, and for their family members to consent within an ethically approved framework for that data to be used in the research environment.

She went on to explain:

None of us knew that there was a disease gene out there that would account 5.5 percent of children and adults with undiagnosed neurodevelopmental disorders. And the two groups that led that work are in the UK. That is a testament to the fact that we have linked genomic data with healthcare data and can look at the outcomes.

But there are gaps in that data, she warned: 

In genomic healthcare there are underrepresented populations. For example, we have far less available data from individuals who are not of European origin. […] So, we need to break down the barriers in terms of encouraging people from a more diverse group to participate in research in the UK. 

Plus, we need to recognize that partnerships with our international and the international community will absolutely benefit patients in the UK too.

My take

An exciting area of research – one that crosses boundaries and offers a vision of a more collaborative, tech-infused future from which we all might benefit.

Loading
A grey colored placeholder image