Understanding impact of pandemic-driven AI on drug discovery

In 18 months, vaccine research advanced AI in medicine by five to seven years. That’s cause for optimism — and investment — in the digital transformation of clinical development. Jeff Elton, CEO of ConcertAI, who has extensive experience in this sector, shares his insight with DDW. There has been a recent flurry of news at the intersection of artificial intelligence (AI) and drug development. So, it might have been a surprise when AspenTech released a survey1 revealing that only 50% of pharmaceutical manufacturers in the US and Europe believe that AI can help bring new drugs to market more rapidly and securely. A remarkable 96% are still facing challenges using AI to derive value from their data. But I think we’ll see those statistics change significantly. Yes, the complexities and challenges in overhauling legacy systems and workflows means almost all pharma companies are still working through the deployment of AI in their operations. But Covid-19 has shown how effectively AI can accelerate innovation in medicine. Aligned to specific questions and requirements of Covid-19 research, healthcare providers, biopharma, diagnostic labs and others confederated, linked and otherwise made available data at scale to each other in collaborative models. Equally important, AI solutions rapidly evaluated different scientific directions and accelerated programme design and clinical trials. My assessment is that the field advanced by five to seven commercial years over the past 18 months due to the pandemic. Covid-19 vaccine researchers understood that the effective use of AI in medicine is not a matter of dropping new tools onto existing data silos or those that could be accessed through traditional approaches. In fact, up to 70% of the work in applying AI is identifying, confederating, and preparing the data so it can be used by purpose-built AI solutions spanning from discovery through post-approval applications. In other words, pharma companies will accelerate value from their data and see exceptional benefits from deploying AI across their enterprises when they change legacy processes, operations, and systems to accommodate the requirements of AI solutions. Starting with specifics As the enormity of the Covid-19 pandemic became clear, scientists were liberated from previous notions of how to work, how to collaborate, and the art of the possible in the quest for insight on transmissions and differential vulnerability paths and approaches to a vaccine. Labs, research centres, and biopharma companies forged an intersection between traditional biomedical research and data science-centric approaches, bringing together new data sources, types, models, technologies. and operational approaches. Traditional ‘attrition’ and ‘rational design’-based approaches gave way to new ones that rapidly provided insights beyond what was available in the scientific literature and traditional cycles of scientific meetings and publications. While failure and non-productive avenues for research are to be expected, the speed of vetting alternatives and generating insights allowed these collaborators to quickly reach termination or acceleration decisions that became ever more critical as the pandemic continued. The crisis unfroze and integrated models of collaboration and data sources that were historically unavailable to each other. The impact of this new approach is massive and offers a few key learnings. First, ‘questions’ should determine the data we need versus the data defining the limits of possible insights. Second, data at scale, data that can be confederated, and data that can be linked are required to bend the curve on the time, cost, and credibility of actionable insights. Third, AI and Machine Learning approaches can bridge published and un-published insights with clear biological or disease relevance and actionability. Finally, the new operating models that are succeeding are data centric. Founded on collaborative approaches, they are encouraging the dismantling of silos, the embrace of AI across enterprises, and the enablement of entirely new digital operating models. Life sciences companies are deploying, adopting and integrating AI technologies at an unprecedented pace. Healthcare providers are also increasingly using them. The legacy business architecture of a life sciences company is now formally changing as they transform how they store their data, how they make their data more accessible, how they are increasingly partnering for broader data access, and how they affirm the commitment to bring AI and ML technologies and data science talent capabilities to bear across their enterprises. Transforming clinical trials As an example of how profound this transformation might be, let’s consider clinical trials in oncology. Clinical trials are launched around insights into disease biology and possible strategies for narrow patient cohorts who may benefit from a new therapeutic or an existing therapeutic brought to a new disease. A very operational process then begins as clinical sites are activated, patients identified, and data collected – all to get to a clean dataset ready for biostatistical analysis as part of the regulatory submission package. This increasing use of confederated first-party and real-world data and AI capabilities have far-reaching implications for new treatments and better patient outcomes – and it starts with clinical trials. In the drug discovery and translational phases, AI and ML solutions are validating targets and identifying the specific characteristics of disease biology that can be cancer or patient-cohort specific. Harnessing a combination of research assets and real-world data, data science-guided translational medicine initiatives are increasingly helping identify populations of significance and design clinical trials for the most devastating cancers where the current standards of care may be inadequate. I’d be surprised if a single study goes into a clinical setting for development in the next three years that has not gone through some form of AI and ML investigation and optimisation to complement traditional approaches. Study designs can be optimised with AI and ML solutions, too, assuring complementary goals such as coverage of traditionally underrepresented racial, ethnic, and economic subgroups and, using advanced AI software-as-a-service (SaaS) technology tools, even selecting the sites where trials are deployed. This shift will extend into the healthcare provider environment. Physicians and research staff will be presented information for their consideration on potential trial eligibilities for each patient – data that has been screened locally through AI and ML tools. These possibilities will be a gamechangers in terms of who has access to live-saving cancer treatments, assuring all those eligible for a potentially beneficial clinical trial are presented for consideration. Currently, 80% of US cancer patients are treated at community centres and 20% at academic centres. Yet the majority of clinical trials are conducted at the academic centres. Using AI and ML SaaS solutions, we can now create world class technology infrastructure with a minimum of incremental staffing requirements to build research capabilities at the community level. My own projections are that these solutions will allow us to double community-based trial participation in the next few years. A 100% increase in a short period of time, even off a small starting base, especially where 80% of patients receive their care, would deliver a profound acceleration of needed new therapeutic entities. Trials designed for and executed in the community can assure that that study results are aligned to the characteristics of the majority of US patients and more generalisable to where the majority of patients receive their care. This would be not just a democratisation of innovation, but also an acceleration of innovation and relevance. Scalable technology means change Full value realisation of potentially transformative technologies usually necessitates corresponding changes in operations, processes, talent, and organisation. This required operating model shift is one of the reasons some pharma companies have been slower to realise returns from their early AI investments. Today, punctuated by Covid-19 initiatives, we are in a moment where life science and other biomedical companies recognise the urgent need, and high value, of adopting and integrating novel AI and ML technologies. In the end, the strongest motivation for building an integrated ecosystem of these technologies will be more precise decision-making and the acceleration of biomedical innovations. This type of change isn’t easy. Two years ago, pharmaceutical manufacturers weren’t sure it was necessary. Then Covid-19 swept across the world and in that moment, the scientific community showed us what medical innovation will look like in 2020 and beyond. Now it’s time to get to work. About the author Jeff Elton is CEO of ConcertAI, Precision Oncology company. Previously, he was Managing Director, Accenture Strategy & Predictive Health Intelligence, global COO of Novartis’ Institutes for BioMedical Research and a McKinsey partner in Pharmaceuticals & Medical Products. Dr. Elton co-authored, Healthcare Disrupted: Next Generation Business Models and Strategies in 2016, creating an industry roadmap for AI, advanced analytics, real-world data and digital medical solutions. References 1. https://www.ddw-online.com/half-of-pharma-believes-ai-can-help-bring-new-drugs-to-market-faster-12275-202107/