An Elsevier Feature piece published this week evaluates the risks and rewards of artificial intelligence in driving “digital R&D” within the pharmaceutical industry. The combined quantitative and qualitative method used in their analysis judged the industry as being in an ‘early mature’ phase with regards to using AI to add value to the R&D process.
The article is written by an assorted mix of five, coming from academia: Reutlingen University and the University of St. Gallen; industry: Sony and PricewaterhouseCoopers; and pharma: Novartis Institute of Biomedical Research.
For the report, the team analysed annual company reports, investor relations information, patent applications, and scientific publications from the top 20 pharma companies by the 2018 total revenues, plus Merck KGaA. The analysis showed that AI does not yet have market relevance in terms of the core strategy of most of these companies, as only two: Johnson&Johnson and Novartis, have commercialised any of their AI products. Compared to the thousands of patent applications submitted by leading tech companies, it’s clear that pharma is only just beginning to apply AI.
As to why some companies better integrate AI than others, the authors cite three macro factors: variations in the R&D strategy, variation in therapeutic areas, and the range of risk factors involved in moving towards digital. The risk factors are sub-divided into preventable, strategic and external.
The most obvious examples of external risks include variable regulatory environments or high market costs for proprietary AI technologies.
Internal (preventable) risks might include the costs of digitalisation, inflexible existing processes, social dynamics within the company for digital transformation, and poor data quality and governance. Preventable risks can be further divided and categorised into technical, performance, control, ethical, and economic risks.
The biggest strategic risk to these companies is investing in costly AI systems and more broadly undergoing a digital transformation, without fully knowing whether it’s entirely or will return value.
The authors conclude that whilst there are many risks involved and that most companies have not yet seen a significant improvement in R&D efficiency with AI, there are a plethora of potential high-value rewards along the AI journey. They actually went as far as to list the top twenty applications of AI in drug development, including everything from helping perform better 3D protein structure simulations to predicting blood-brain barrier permeability of drugs.
For those companies who have yet to fully transition into ‘digital players’ the authors have a number of recommendations: set up comprehensive digital strategies, develop related business cases, and designate AI-specific budgets. What’s more, they need to direct human and financial resources to new data science R&D functions and accept the high costs of digital upskilling.
