‘It’s an opportunity on a galactic scale’: Genomics England’s CEO, Chris Wigley, thinks AI can help change healthcare forever
Chris Wigley likes to talk in metaphors. It is easy to understand why: his day job focuses on datasets so huge and problems of such minute origin that they are almost impossible to imagine.
As CEO of Genomics England, a government company originally set up to sequence 100,000 genomes from NHS patients, Wigley intends to help the NHS realise the true potential of genetic medicine.’
The former diplomat and partner at consulting firm McKinsey joined Genomics England in October 2019 from AI and machine learning specialists QuantumBlack. In the run-up to Tortoise’s Global AI Summit on 3 December, Tortoise sat down with Wigley to talk about how AI could help genomics change the world.
A basic tutorial in genomics seemed a good place to start.
Genomics for dummies
Imagine a thread running from London to New York and decorated with 20,000 beads, Wigley suggested. The thread is a genome, made from DNA. The beads are genes. Genes encode ribonucleic acid (RNA), which in turn encodes proteins, which perform all kinds of functions in your body.
Genomics — the study of the structure, function, evolution and mapping of the entire thread — enables scientists to understand much more about why people become ill, and how to treat their conditions. Genes (the beads) form only 1.2% of our overall genome; the rest (the thread) is called “non-coding” DNA.
A mutation in a single gene can cause conditions like Sickle Cell Anaemia or Huntington’s disease. Meanwhile, Cancer is often a product of gene mutation, but can also be caused or affected by changes to whole sections of the thread, not just the beads.
This is all pretty mind-boggling, especially as a little maths suggests if the beads were arranged one after the other on our transatlantic thread, they wouldn’t get past the M25…!
What role for AI?
Artificial intelligence already plays a big role in genomics.
The first application is in data wrangling. Medical data are “often super messy, despite all the great work to clean them up” says Wigley. “There are dozens of diagnostic systems, from GPs to secondary care and they are all encoded according to different taxonomies.
Natural Language Processing and Machine Vision can help Genomics England and others wrangle data out of PDF medical records into forms they can analyse. “It can do a lot of heavy lifting for us,” Wigley said.
The second is in diagnostics. “Researchers can already use machine learning and artificial intelligence to understand what your risk is of a disease like heart disease, or diabetes is across thousands or hundreds of thousands of locations on the genome,” Wigley explains.
Machine learning is one of the ways to generate a ‘normal’ genome that an individual’s variations can be compared against. There could be hundreds of thousands of variations across each of the genome’s three billion base pairs. Genomics England sorts the variations into tiers, according to how serious they think the variation is.
Each genomic analysis requires processing vast quantities of data at high speed. “It’s a combinatorial problem on a galactic scale,” says Wigley. “Without complex algorithms to help, it would be impractical,” he said.
The third application of AI is in research, both from academics and from biotech and pharma companies: figuring out how variations in the genome eventually lead to changes in the body and supporting pharmaceutical companies to find targets for treatment.
In the years to come, quantum computing could multiply the percentage of a patient’s genome scientists can analyse, and the speed at which they can do it. “I’m not a quantum computing expert, but Genomics is the kind of problem that seems amenable to these new techniques” said Wigley.
Genomics England is also tackling the most pressing problem today: the coronavirus pandemic. Genomics England is working alongside the University of Edinburgh, Public Health England and others in the Coronavirus Genomics consortium to sequence the genome of both the virus itself, and the humans who get COVID-19. Across both sequencing programmes, the government has invested over £50m to turbocharge the insights genomics can bring to the struggle against the virus.
In sequencing humans, the aim is to understand why some patients with coronavirus experience a mild infection, others require intensive care and why some patients die from the disease.
It is too soon to say what the results will be but Wigley says rare variations in the immune system appear to affect how severely someone is affected by Covid. Specific susceptibilities caused by variations in the genome could be treated by repurposing existing drugs hitherto used for other purposes. “The most famous example of this kind of approach is probably Viagra, which was originally for treating high blood pressure, but then turned out to have unexpected positive side effects!” said Wigley.
Existing drugs, already used for other conditions, are now being used in Covid Recovery trials.
Does genomics have limits?
However, some of the genomics sectors’ biggest challenges are ethical, not numerical.
Wigley recalls a run-in with someone he describes as a ‘campaigning lawyer’ at an event in Ireland, describing how the man told him Genomics England’s work was both illegal and immoral, that Wigley himself was evil and that he was exploiting people’s data and selling it to Big Pharma.
Wigley clearly took some satisfaction in telling him he was ‘unequivocally wrong’.
Participants in the 100,000 Genomes Project elect regional representatives who in turn elect 30 national representatives. They are represented in the Genomics England “Access Review Committee”, which decides which researchers with which projects can go ahead. Researchers pitch their projects to the panel, and the participant representatives can veto any idea. Only the Secretary of State for Health can overturn their objection.
Wigley believes this approach is both novel and important. “These people [involved in the genome project] are really switched on,” he said. “If you or your kid have a rare disease, you become the world expert on that. They’re making fully informed choices. They are the people who approve the research projects, not me. And they don’t want less research into their or their kid’s condition, they want more.”
And if they change their mind? Wigley explains participants can pull their data out of the project if they become uncomfortable for any reason. “They can raise that with us directly, they can discuss it with me directly. If they want to change things, they can put themselves forward on the basis of disagreeing with what the current representatives are doing.”
Neo-natal or pre-natal genomic sequencing is another potential area fraught with ethical questions. The NHS already offers parents-to-be screening tests (using ultrasound scanning) for Down’s syndrome, Edwards’ syndrome and Patau’s syndrome. Technically, scientists could test a lot more by looking at foetal DNA, which is easy and painless.
“On the basis of a genomic sequence, we could in theory tell parents about their child’s risk of some cancers, cardiovascular disease, many things about this future human being. Unless we’re thoughtful in this area we risk getting into dangerous territory. So unsurprisingly it’s an area we’re approaching with great caution.
“But we are exploring it because the benefits to health could be so profound. It would be really good to tell parents that we have tested for, say, 300 treatable rare conditions, and their kid either doesn’t have any of them, or if they do we can start treatment from birth, which often massively improves outcomes”
“So as we think through these benefits and these risks, no one person knows the right answer,” he said. One of Wigley’s goals with Genomics England is to push for a “national conversation”, similar to the 1980s Warnock Review into embryology, to find practical ways forward to such complex questions.
In the more immediate future, Covid-19 remains on his ‘to-do’ list.
The National Institute of Health Research (NIHR), and Genomics England were awarded £1.3 billion in November’s government Spending Review for R&D to improve outcomes for patients of illnesses including Covid-19.
“NIHR get and pass on most of that, and it feeds the whole ecosystem — we just get a small slice,” he laughed. “But thankfully it’s enough both to keep the lights on, keep supporting the services we provide, and keep innovating and pushing forward.”