Neil Thompson, chief scientific officer at Healx explores how digital innovation has emerged throughout the Covid-19 pandemic and beyond.
Digital Innovation
The battle against Covid-19 has dominated the world’s attention over the last year, impacting economics, societies, individuals and industries across the globe. In response to the pandemic, the UK in particular has seen a surge in digital innovations exploring the use of artificial intelligence (AI) and machine learning (ML) technology across the healthcare spectrum. As we reflect on the last year I see a number of key learnings that can enable us to maximise the value that can be derived from these innovations in the broader healthcare sector.
Data and enthusiasm are critical for innovation
The first is that there are a vast number of innovative ideas out there - some of which may not have been realised without the urgency of a pandemic. This creativity goes well beyond some of the generally accepted uses of AI and ML in imaging, diagnosis and modelling. Indeed, from tackling misinformation to assessing people’s risk of transmission, such advancements have helped monitor the spread of the virus, keep people connected, and supercharge research and development for treatments. This has been possible due to both a high quantity of data being available to make these innovations viable, and an enthusiasm from the community to actually evaluate and incorporate their use. In the UK, for example, hospitals at the Bolton NHS Foundation Trust began using the chest X-ray AI tool from Qure.ai to quickly detect signs of the virus as well as automatically monitor its progression. This, in turn, helped staff make critical decisions such as when a patient should be moved to the ICU or be intubated. Furthermore, researchers from King’s College London worked with an international consortium called icovid to develop a cloud-based AI software that could rapidly quantify the degree of lung involvement in Covid-19 patients. With this information, doctors could then triage admitted patients, potentially alleviating the burdens on intensive care units. This combination of data and enthusiasm is critical to unlock the power, and potential, of AI tools in the future as more innovative use cases evolve.
Always build with the end-user in mind
The second learning is that not all of these ideas have delivered the intended benefit - and there are some interesting lessons we can take away from that. Some of the better informed assessments have highlighted approaches that can, and should, be taken in the future to ensure we get the most from these innovations. One key thing is to ensure that the technologies are developed hand in hand with the likely end-users or consumers in order to avoid a mismatch in expectations. A second is that we should expect all innovations within the healthcare space to undergo a period of testing and evaluation before wider adoption, as with many other areas of patient care such as drug discovery and development. The closer an idea is to impacting patient experience, the more heavily regulated that evaluation is likely to be. The recently published modifications of guidance on clinical trial protocols (SPIRIT-AI) and on trial reporting (CONSORT-AI) have set out specific considerations for trials involving AI approaches which have been very helpful in setting the standard of reporting such research.
Applying AI to drug discovery could unlock huge patient benefit
The third is that, when it comes to drug discovery, AI and ML could play a critical role in revolutionising the process and opening up access to life-saving treatments to many new patient populations. For many, the urgent need for treatments and vaccines to combat Covid-19 revealed the pitfalls of legacy drug discovery methods. Indeed, to bring a new drug to market takes roughly 10-14 years, costs between $2-3 billion, and has a 95% failure rate. But with the use of AI technology, both time and risk can be reduced. In August 2020, for example, DeepMind utilised their AlphaFold algorithm to predict a variety of protein structures associated with Covid-19 which provided a critical insight into the structure of proteins crucial for virus entry and replication. With this knowledge, scientists had greater clarity on viral pathways and even potential antivirals. Others, like BenevolentAI and Insilico Medicine used machine learning to quickly identify possible treatments for the virus. These successful examples of AI impacting drug discovery have, in their own way, heeded the lessons outlined previously: starting from an informed understanding of the unmet need and working with the end users to shape the output.
But this is something we, at Healx, have believed for a long time. We use our AI platform, Healnet, to identify and progress novel treatments for the 95% of rare diseases currently without an approved treatment, working in close collaboration with patient groups, their physicians and their carers to understand the precise nature of the unmet need. Healnet integrates data from multiple sources (everything from disease and clinical trial data to scientific literature and patient group information) to form the world’s most detailed knowledge graph of rare disease information. This graph is then analysed by state-of-the-art AI and natural language processing (NLP) models, to uncover previously unknown relationships between existing drugs and rare diseases. Our expert team of drug discovery experts then interrogate those predictions and progress the most promising candidates to clinical trials generating the necessary data to support the regulatory process required before we initiate patient trials. By bringing together innovative technology and drug discovery experts we believe we can maximise the chances of the immense potential of AI in drug discovery.
What’s next?
In response to Covid-19, AI-powered technology has been utilised on multiple fronts - from monitoring the spread of the virus to developing vaccines, reducing the burden on staff and predicting infection. Indeed, it has been inspiring to see how far we have come in terms of digital innovation in just a few months. Beyond the pandemic, I hope to continue seeing AI being used in innovative and useful ways across the sector, but always guided by the end user and always subject to the rigorous testing we expect to ensure patient benefit.