From the acceleration of regulatory submissions – by identifying data gaps that have led to delays or rejections in the past – to the transformation of the conduct of clinical trials and patient safety monitoring, artificial intelligence (AI) has substantial potential to change the way life sciences organisations operate.
AI and machine learning have risen rapidly up the business agenda in a wide range of industries – during the last year in particular. On the basis that computers can analyse and interpret data far more quickly and holistically than humans can, market innovators are staking their reputations on the breakthroughs that those analyses and interpretations will enable – ranging from improved customer self-service to advanced problem-solving in such areas as health diagnosis and predictive maintenance. It isn’t just that machines return results at higher speeds or that they can work around the clock; machines also learn extremely efficiently so that their performance improves exponentially over very short periods of time.
Those are some of the reasons that science documentaries and news reports have begun focusing on the potential for AI and machine learning to facilitate, for instance, earlier medical diagnoses – particularly in complex or baffling cases. That’s because advanced algorithms powered by huge computing resources can combine and mine vast sources of global data to identify significant clues compared with even the most-experienced clinicians, who can draw on only their own learning and their own experience.
In the field of life sciences, the scope for exploitation of AI for operational – not to mention commercial – advantage is expected to be very wide and to extend right through the product life cycle. Researchers here too are increasingly using the power of AI to mine data, making it conceivable that, in the near future, they will be bringing AI-discovered drugs to market. Increasingly intelligent robotics are likely to remove routine tasks from the lab and – elsewhere in product development – free scientists to focus on more-bespoke, higher-level tasks.
Getting closer to patients
One of the more exciting options is that AI can take pharmaceutical companies deeper into the realms of wellness and the proactive prevention of illness – especially self-inflicted health problems – as technology learns how to recognise warning signs and then prompt better decisions or timely interventions. At a time that the future of life sciences is being hotly debated, the opportunity for the industry to refine and expand its role is an important one to take advantage of. Back-end technology already exists to facilitate more intelligent and proactive health monitoring by taking things forward as drug companies rely on finding the optimum ways for patients to interact with and use the tools.
Social media and web forums together with personal apps offer obvious and already well-accepted means whereby patients can submit and monitor information about themselves for purposes of analysis and reporting. Plus, life science companies are looking more and more at proactive ways in which they can harness those means within all of the legal and industry rules covering what’s acceptable.
If approached responsibly and within regulatory guidelines, web and social listening offers companies a way to determine early on how their drugs are being experienced and the impacts those drugs are having. There is also important safety monitoring potential and drug feedback potential, as long as intelligent tools based on AI and machine learning are in the background offering companies what to look for and ways of deciphering what it all means. GSK is one company that has decided to take a progressive approach to social monitoring, and indeed to AI in general. It has established a pharmacovigilance centre of excellence to collect web and social data for the aforementioned purposes.
Proactive intervention
Already AI is being used to identify women whose Twitter posts indicate they may have increased risk of developing two relatively rare diseases: ovarian cancer and cervical cancer. And that AI application has produced accurate alternative diagnostic insights. Elsewhere, companies are attempting to aggregate data from patient health records by way of AI-derived ontological approaches that try to extract useful data from handwritten doctors’ notes.
AI could have a strong bearing on personalised medicine too, because clear insights into data make it easier to determine which sets of patients a drug might be best suited to. AI could also support a range of applications in product safety – for example, the use of algorithms to predict toxicity in clinical trial studies, which would give companies information for developing better molecules.
Because – with access to the same data – machine learning can pick up on subtleties that are less obvious to a human, the potential for the use of AI in clinical trials is considerable. Wearable devices relaying data feeds from patients could mean the slightest effects would get picked up – and early trends noted, without the potential for human bias or misinterpretation – earlier than previously possible, all of which could have a bearing on drug development and improved outcomes.
In some of the areas, regulators may have to take time to catch up to the potential when it comes to their willingness to accept AI-enabled insights as a valid part of reporting.
Increasing regulatory submission success
On a more traditional, operational basis, AI offers a path through the data complexity that has typically held back life science organisations from becoming more agile, innovative and responsive.
An issue that will become exponentially more significant as the International Organization for Standardization’s new Identification of Medicinal Products standards come into play is that data about pharmaceutical products has historically been created, gathered and stored by multiple, different functions in multiple systems. Because of that, one of the challenges companies typically face is that the usefulness of those multiple systems relies on whether that created, gathered and stored data can be combined and counted on as a definitive record of product truth.
In a 2016 survey by Gens and Associates called Pursuing World Class Regulatory Information Management, around half of the companies surveyed said they were already investigating AI to streamline regulatory information management, and a further third said they were monitoring what other companies are doing in this area.
By uncovering gaps in the data, machine learning can detect regulatory dossier anomalies that might be holding up product approval. With what could be tens of thousands of pages making up a submission, it can be hard for a human to manually spot where data or metadata may not be stacking up. An AI system that is taught to assess previous, successful submissions can quickly learn patterns and alert regulatory teams if it detects anomalies with the latest submission and, based on past experience, can analyse whether a drug is likely to become authorised.
Watch this space
Without question, AI has much to offer across the life sciences life cycle. Experts in the industry are increasingly looking at how machine learning can be deployed across a broad spectrum of potential applications – from exploiting the full depth and breadth of regulatory data to finding new routes to innovation and improvement of the patient experience.
As more and more companies identify opportunities to turn AI-enabled insights into timely and beneficial outcomes – whether by accelerating market entry, successfully mining social media for potential adverse events and other patient feedback, discovering new indications, or improving the manufacturing and supply chain process – advanced automation through increased machine intelligence looks set to be the way forward.
The foregoing insights were distilled from a recent debate on artificial intelligence hosted by ProductLife Group with participation from a cross section of life science industry senior representatives and advanced life science technology companies. Taking part in the discussion were Mark Davies, vice president of biomedical informatics at BenevolentAI; Jackie Hunter, CEO of the BenevolentBio division of BenevolentAI; Martin Gouldstone, partner at Results Healthcare; Steve Gens, managing partner at Gens and Associates; Christopher Rudolf, founder and CEO of Volv Partners; and Marco Anelli, head of the pharmacovigilance advisory and medical affairs practice at ProductLife Group.