Artificial intelligence will be a key technology in ensuring that the new pharmaceuticals of the future will be effective for as many people as possible, explains Professor Jackie Hunter
In the last decade alone, over a trillion US dollars were spent around the world researching and developing medicines. Yet the top 10 selling drugs on the market today only work for around 30-50 per cent of the patients for whom they are prescribed, and over 300 million people currently suffer from rare diseases that, unless we dramatically disrupt current economic and development models, won’t see any successful medical treatments for the foreseeable future.
It’s a problem of both time and cost; developing and commercialising a drug can easily take 10-15 years at a cost of more than US $2.5 billion. One of the main reasons for this is because of high failure rates; even if a molecule reaches clinical development, the chances of it then reaching the market are still less than 10%. Long term, this isn’t sustainable. We need to increase our chances of success and cut the cost of failure.
- AI lab-bot enrolled in drug discovery programme
- Magnetic sorting and genomic technique for drug discovery
- Digital filter aids data driven drug discovery
One of the main reasons I started working at Benevolent AI was that the pharmaceutical industry wasn’t innovating in a way that society needed. Currently, most drugs fail in phase II and III clinical trials because we aren’t actually modulating the right target in the disease. Sifting through clinical drug reports is often like looking for a needle in a haystack; trials can include several hundred patients and generate reports 26,000 pages long. You can’t effectively analyse that data with any degree of granularity by hand.
But AI flips this problem on its head; with machine learning, vast sums of data can help to improve the quality of the result. We can train systems to recognise and select several high-quality ‘needles’ out of that haystack and focus our resources on following those leads, rather than scrabbling around manually to find just one. AI gives us rapid access to more relevant data for the disease and helps us to understand its biology – and ours – more comprehensively. Increasingly, AI is pivotal to maximizing the benefit of the available genetic or genomic information we accrue, as well as for analysing existing electronic medical records.
This could have a transformative effect on both time and cost. We can move a drug from a chemical starting point to a molecule ready for clinical testing three times faster, and we can do it using just 10% of the molecules that we would have used before.
There are benefits to using AI at the individual level as well. It allows for more personalised treatments and lifestyle management strategies for patients; it frees healthcare workers of the more routine aspects of their work and enables them to focus their experience and insight on abstract problems; and it democratises healthcare information for clinicians by allowing them to get answers to questions they’ve never been able to ask before.
Artificial intelligence permeates nearly every industry today and its application in healthcare is the only way that the vast amount of medical data being generated around the world will be fully utilised. But the health benefits will only materialise if we continue to invest in AI, and there are legitimate concerns about doing so.
We’ve seen what’s happened in other areas such as with Facebook, Google, and Amazon, where concerns over privacy haven’t been acknowledged and there has not been transparency in the system of how they hold the data, what they use it for, and so on. It’s especially important to remain transparent about how and why we use people’s private medical data, a requirement we highlighted last month at the Global Grand Challenges Summit in London, and one that we need to keep at the centre of future discourse as well. Engineers must understand the potential problems they could create if they develop medical solutions in silos.
An ecosystem is only as strong as its weakest link. By creating solutions in a bubble, or by focusing too closely on one area – be it the industry, the healthcare provider, or the patient – we risk overlooking key data and either creating new problems for people, or excluding entire swathes of the population from effective treatments. It has to be an integrated discussion about that whole healthcare data ecosystem.
But if we do it right, then we all stand to benefit. The transformative effects of engineering innovation in healthcare are already evident. Machine learning brings more effective medicines to more people more quickly, at a fraction of the cost and resources. We need to push the boundaries of AI and machine learning and unlock the power of decades of research, to truly understand diseases and to develop inclusive treatments for the millions of patients who need them.
Professor Jackie Hunter CBE FMedSci, Chief Executive Clinical Programmes and Strategic Partnerships of Benevolent AI, spoke at the Global Grand Challenges Summit in London in September, hosted by the Royal Academy of Engineering, the US National Academy of Engineering and the Chinese Academy of Engineering