This year, July marked National Fragile X Syndrome Awareness Month, an annual event focused on raising awareness of the rare condition which is the leading genetic cause of learning difficulties in the world.

The event comes at an interesting time, especially in the UK where England’s Rare Diseases Action Plan 2022 was recently published, outlining how the government’s Department of Health and Social Care will implement the UK Rare Diseases Framework, which has been designed to improve the lives of those living with rare diseases.

More attention from the media and government is being paid to rare conditions than ever before. But this may not come as much of a relief to the more than 400 million people living with a rare disease worldwide – especially when at least 95% of these rare diseases lack any approved treatment.

Bringing a new treatment to market takes years, if not decades. And it relies on a good understanding of disease biology, as well as a large enough pool of patients to test a potential treatment in. These, unfortunately, are often absent from rare conditions and, as a result, it is not surprising that the dialogue around rare diseases so often focuses on diagnosis and support rather than treatment.

However, advances in technology – particularly in artificial intelligence (AI) – are providing hope for those living with rare diseases and, more than that, delivering actual treatment to those in need.

As the name suggests, rare diseases are very uncommon. The EU definition is a disease that affects less than one in 2,000 people.

Their rarity presents serious hurdles for the development of effective treatments. Little is known about some, and finding patient pools to run research studies presents challenges for others, meaning pharmaceutical companies have not traditionally been incentivised to focus on them.

But with over 7,000 known rare diseases, this means there is a huge number of people left behind. In fact, there are over 3.5 million people in the UK alone living with a rare disease, and as many as one in 15 people worldwide live with one.

It’s understandable that rare diseases have been neglected by pharmaceutical companies; as mentioned, bringing a drug to market is an expensive, time consuming endeavour that ends in failure 96% percent of the time.

‘There are more than 400 million people living with a rare disease worldwide and at least 95% of these rare diseases lack any approved treatment’

The consequence, however, is that much of the conversation around rare diseases revolves around awareness, support and simply coping with a condition.

England’s Rare Diseases Action Plan, for instance, mainly focuses on diagnosis and makes no mention in its 91 pages of the potential for the biotech sector to deliver treatments using machine learning tools. In fact, it only mentions AI once, in its use to improve blood spot screening.

But this is exactly where AI can end the deadlock and bring new treatments to patients in need.

In the last decade, we’ve seen a real renaissance in AI and how it can be applied to drug discovery.

Improvements in software and computing power now mean it is possible to take an entirely different approach to discovering new drugs or applications for them, which is crucial for rare disease treatment, given the disincentives faced until now.
There are numerous different ways AI can be leveraged in this field to overcome the challenges.

By applying systems to enormously expensive or large mathematical computations, for example, we can uncover new disease biology and use this to identify potential compounds to synthesise and then test those that we think might work.

DeepMind’s recent success with its AlphaFold AI is a great example of using AI to solve long-studied scientific challenges; the company backed by Google’s owner Alphabet has in the space of a year been able to accurately predict the 3D structure of almost every protein ever catalogued and release them to the public – all 200 million of them. It’s no surprise then that it has also spun-out a separate company, Isomorphic Labs, to apply this process specifically to drug discovery.

Another approach is to set AI to look for subtle links in large swathes of information that even the finest minds might miss. Even the most voracious scientists cannot keep up with every journal article in every language, or data set released as part of them. And humans inevitably have biases – where we choose to start with a problem and deciding upon what we think might or might not work.
That’s the approach companies like BenevolentAI have taken. Its treatment for atopic dermatitis is now in phase 2 clinical trials and was discovered using machine learning technology.

It’s also the approach we’ve taken at Healx. Our AI platform, Healnet, analyses millions of drug and disease data points from papers, trial results and more to find novel connections that could be turned into new treatment opportunities.

It might spot that one drug – or combination of drugs – has an unexpected side effect that would be advantageous as a treatment for an entirely different condition, for instance. Or it might spot an existing compound that could be slightly tweaked to improve its efficacy in another disease.

Using approved drugs – of which there are over 20,000 – as a starting point for this discovery process vastly speeds up the time taken to bring a new treatment to market, and reduces the cost of trials, because the safety profile of the compound is often already known and there might also be useful preclinical data we can use as a springboard.

This process has helped us reach the stage where the US Food and Drug Administration (FDA) has given Healx its Investigational New Drug (IND) approval for a phase 2a clinical study of the compounds HLX-0201 and HLX-0206 for Fragile X syndrome.

The other crucial opportunity AI provides is the ability to bring these approaches at scale. As computing power further increases, companies will be able to run these AI systems in parallel, discovering drugs or new applications for drugs for whole categories of conditions at once, rather than one at a time.

This is the only way we will be able to provide meaningful change for the hundreds of millions of people suffering from rare diseases within our lifetimes – any other approach will just be too slow.

Undoubtedly, there is still work to be done raising awareness of rare diseases. Patient support groups and advocates have made tremendous progress helping the wider public understand rare conditions in recent years, and it is great to see the media and governments take the challenges faced by the rare community more seriously.

But we must broaden the conversation too and start to think more critically about how we can leverage technologies like AI to find and deliver treatments faster and at scale.

It can be done. Rare diseases are life-changing, but the potential to treat many of them, and not merely support those diagnosed, now exists. From the healthcare industry to governments, we need to make good on that.