By Maria João Cruz, Research Analyst, and Dr Francesca Properzi, PhD, Research Manager, Centre for Health Solutions
February 29, a pretty rare date, was this year’s Rare Disease Day, a campaign to raise awareness among the public and policy makers on rare diseases and how they impact the lives of patients and their loved ones.1 Indeed, the advocacy, commitment and tenacity of rare disease patient organisations have played a critical role in elevating ‘rare diseases’ as an emerging global public health priority. Particularly in advocating for legislation to develop drugs and treatment programmes that will meet patients’ needs. In today’s blog we highlight the importance of rare diseases and, in particular, the opportunities that artificial intelligence (AI) tools offer to speed-up the discovery of new treatments.
What constitututes a rare disease and what is the scale of the problem?
There is no universal definition of what constitutes a rare disease. In Europe, a disease is considered rare when it affects less than one in 2,000 people, or fewer than 200,000 people in the United States.2,3 Currently, there are over 7,000 rare diseases described in the literature (with 250-280 new ones identified each year); of these, 80 per cent have a genetic basis. Collectively, however, rare diseases affect over 350 million people worldwide, representing a considerable global public health burden.4,5 Around half of those who suffer from a rare disease are children and 30 per cent of them will not live to see their fifth birthday.6
Why is the development of drugs to treat rare diseases so challenging?
The most prominent challenge in understanding, diagnosing and treating rare diseases is, of course, their ‘rarity’. Indeed, it is logistically more complicated, if not impossible at times, to reach such small and widely dispersed patient population. This means that patient recruitment to test new therapies that target specific rare diseases is extremely difficult, often resulting in cohort studies that lack statistical power. Additionally, in the majority of the cases, there is insufficient knowledge of the disease pathophysiology, especially regarding the molecular pathways involved, which could be used as therapeutic targets, and a lack of specific biomarkers for early diagnosis.7,8 This limited or no access to diagnosis, treatment and resources often prevents patients from receiving appropriate and timely care and crucially limits the possibility of enrolment and participation in clinical trials.
The pharmaceutical industry has traditionally been reluctant to invest in the development of drugs to treat rare diseases, also known as orphan drugs, due to the limited prospects of financial return from such a small market.9 Moreover, as our report ‘Measuring the return from pharmaceutical innovation 2019’ shows, big pharma companies take on average 10 to 12 years to bring a drug to market (a third of this spent on drug discovery) with average costs exceeding US$ 2 billion per drug.10 With the number and types of drugs in the pipeline tending to be focused on those with the potential for higher returns. To address this problem and galvanise the development of orphan drugs, a number of governments have introduced incentive mechanisms, including regulatory simplification of marketing approval, extended market exclusivity and financial incentives such as tax breaks or credits.11
What needs to change and is AI the answer?
Despite the formidable work by patient organisations and regulatory bodies to spur pharma companies’ interest in investing in orphan drug development, rare diseases are still a major unmet medical need with some 95 per cent still lacking an adequate treatment.12 Efficient strategies that allow for a prompt diagnosis and effective treatments are urgently needed.
Recent advances, such as genomic analysis through next generation sequencing and other ‘omics’ approaches, have led to an explosion of clinically-relevant data that can also be analysed and interpreted. The power of this ‘big data’ can be unlocked by harnessing the use of AI tools, including machine learning (ML), to design algorithms that continuously mine, analyse, ‘learn’ and make data-driven decisions.13,14 This potential is particularly high in the early phases of drug discovery as AI can speed-up the identification of new therapeutic targets and design highly efficient drugs for these targets in a very short time. This is explored in detail in our Intelligent drug discovery: Powered by AI report, which highlights the unprecedented potential of AI to revolutionise our understanding of drug compounds and targets, as well as their binding affinity. The significant reduction in the timings of early discovery and preclinical testing provides a unique opportunity for pharma companies to diversify drug pipelines and enable more competitive R&D strategies, addressing rare diseases and other areas of unmet medical need.15
Just last week, the MIT Technology Review recognised AI-discovered molecules as one of the 10 breakthrough technologies of the year, due to its ability to use techniques like deep learning and generative models to make the process of developing medicines faster, cheaper and more effective!16
Indeed, our report identifies the growth in exciting new AI for drug discovery start-ups, corporations, investors and R&D centres developing AI solutions, many of which are focused on discovering new drugs or optimising and repurposing existing drugs to target and treat rare diseases.17 Some of these compnies are now applying their skills to improve the efficiency and effectiveness of the clinical development stage.
One such case example is BenevolentAI, which creates and applies AI technologies at every step of the drug development process, from early discovery right through to late-stage clinical development. Their focus is on hard-to-treat and rare diseases in the areas of neurology, immunology, oncology and inflammation. In their drug programme for Amyotrophic Lateral Sclerosis (ALS), a rare motor neurone disease, BenevolentAI’s platform was used to produce a ranked list of potential treatments. Five of the most promising candidates were then taken to the Sheffield Institute for Translational Neuroscience (SITraN), a world reference on ALS. Remarkably, an ALS lead molecule emerged from a breast cancer drug, which showed delay of symptom onset when tested in the gold standard disease model.18 Since our report was published, BenevolentAI’s website highlights how their workflow has resulted in lead molecule that exhibit a profound rescue effect in ALS patient cells and are currently in the late stage of lead optimisation to then initiate clinical studies.19
Does the use of AI erald a brighter future for treatment of rare diseases?
The large patient community, behind ‘Rare Disease Day’ every year, has been instrumental in driving the legislations that support orphan drug development, which in turn has sparked the interest of pharma companies in this field. However, there is much still to be done. We are living through a spectacular era of cutting-edge technological innovation and the pharma industry is at a tipping point in harnessing digital technologies, particularly AI. Patients suffering from rare diseases, and their families, can and should be able to look forward to a future where, with the help of these technologies, an efficient and cost-effective drug discovery and development framework is established to deliver the much needed treatments.