The Adoption of AI: Critical Concerns in the Life Sciences

Artificial intelligence (AI) is continuing to expand and develop across many industries, including the life sciences. A recent survey conducted by the Pistoia Alliance at its annual conference uncovered that most life sciences experts acknowledge AI’s potential, but concerns remain about its trustworthiness and implementation at scale.

Technology Networks spoke with Dr. Becky Upton, president of the Pistoia Alliance, to learn more about the key findings of the survey and their significance. In this interview, Dr. Upton also discusses some of the main hurdles preventing the wider adoption of AI in pharma and how these challenges may be overcome.

Anna MacDonald (AM): Can you give us an overview of the aims of the survey conducted at the Pistoia Alliance conference and the key findings?

Becky Upton (BU): The Pistoia Alliance conducted its recent survey to gauge the extent of AI adoption in the life sciences sector. Our conference was the perfect place to conduct this research, gathering nearly 300 experts from across the pharmaceutical, technology and regulatory fields.

Key findings revealed that while 70% of life sciences experts acknowledge AI's potential, many struggle with its implementation at scale, largely due to issues like data integrity and interoperability. The survey also uncovered concerns about the trustworthiness of AI. A notable number (63%) of respondents expressed worries that poor data quality could lead to incorrect conclusions and even potentially harmful clinical decisions.

These insights underscore how urgent it is for more industry-wide collaboration to address these challenges, including by standardizing data practices and supporting the ethical use of AI. The survey results are instrumental in shaping our ongoing and future initiatives aimed at fostering innovation through collaboration.

AM: Did any of the survey findings surprise you?

BU: The survey findings also echo trends uncovered by the Alliance’s Lab of the Future (LoF) research from September 2023. The LoF report equally cited issues related to data management, including once again that low-quality and poorly curated datasets are the number one barrier (58%) to implementing AI. Privacy and security concerns around data were also raised as a challenge by 34% of respondents in the report.

It’s no surprise to see these issues still persist more than half a year later. While AI technology is advancing rapidly – with iterations to consumer models like ChatGPT dropping monthly – it will take longer for the highly regulated life sciences industry to adopt AI at scale. Working habits must change, and the industry must ensure AI is being adopted ethically and safely, with relevant regulatory guidance in place.

Ultimately, we are dealing with patients’ health, so decisions about the technology’s use cannot be taken lightly. This is why bodies like the Pistoia Alliance are so important – to create a shared space to tackle challenges related to emerging technologies together and to ensure that collaboration delivers.

AM: What are the main hurdles that are preventing wider adoption of AI in pharma?

BU: The adoption of AI in pharma is hindered by a combination of technical and cultural challenges. Underpinning the successful application of AI is the need to first establish a robust foundational data backbone. Yet, data quality and management clearly remain a significant barrier in the life sciences industry. Many organizations struggle with data silos, unstructured data and a lack of metadata standardization, all of which prevent data from being easily accessible and interoperable. This fractured data environment slows down workflows and diminishes the potential benefits of AI.

Cultural resistance to new approaches and to sharing expertise is another hurdle. Many companies have been using the same processes and toolsets for years, so there is often pushback to shaking workflows up by introducing AI, even if overall it would speed up R&D. This is why it’s so important to build user-friendly tools and for technology champions to share best practice across their organization and the industry as a whole.

One final barrier is the lack of proven business cases for AI adoption. Senior stakeholders often need tangible evidence of AI’s benefits to justify further investment, such as time saved or the number of new drug targets identified. Without clear success stories, it’s challenging to secure the necessary buy-in for further AI programs from decision-makers and budget-holders.

AM: Can you tell us about some of the projects that the Pistoia Alliance has launched and how they will help to address some of the challenges highlighted?

BU: Since its inception, the Pistoia Alliance has been fueling the successful adoption of emerging technologies and acting as champions of science through its portfolio of projects and communities. The Alliance is committed to ensuring collaboration delivers tangible outcomes – from producing best practice guides to developing new data models and standards. Here are just a few of our initiatives.

AI Community of Experts: This community provides a safe space for organizations to share ideas and best practices using our pre-competitive legal framework. The community is looking to develop dedicated projects in several important areas, including AI in regulatory compliance, ethics in AI, information security for AI and benchmarking for AI models and recently established new large language models in biological R&D projects.

Identification of Medicinal Products (IDMP) Ontology: The Alliance created a common data model in collaboration with Bayer, Novartis, GSK, Roche, Merck KGaA, Boehringer Ingelheim, Johnson & Johnson, AstraZeneca, Amgen, AbbVie and Pfizer. This ontology sets data standards to improve substance identification, cross-border prescriptions, regulatory process integration, supply-chain analytics and pharmacovigilance.

Clinical Trial Environmental Impact Community: In partnership with more than a dozen Pistoia Alliance members from pharma, service and technology providers and the Sustainable Health Coalition, Pistoia Alliance is developing a standard methodology for measuring the carbon footprint of clinical trials. This will allow companies to quantify the environmental impact of their clinical development programs against an industry-wide standard, including an evaluation of the impact of the use of decentralized components.

AM: How does the Alliance plan to strengthen ties with global regulatory agencies?

BU: The Alliance plans to deepen our involvement with regulatory agencies in our projects and communities, to decrease the amount of friction in regulatory submissions. The partnership is mutually beneficial; via the Alliance, regulators can speak to multiple companies under one roof about new standards and processes for companies to adopt. At the same time, pharma companies can share their existing standards and processes, allowing regulators to identify potential issues and requirements accordingly. This collaborative approach saves time, money and, crucially, speeds up the delivery of new treatments to patients.

We have already seen success with this strategy, notably in our work with the FDA on the In Vitro Pharmacology Group and Global Substance Registration System (GSRS) Consortia.

AM: 70% of the life science experts surveyed recognize AI's potential. Did the survey uncover reasons why the remaining 30% do not share these views? Is the Alliance planning to investigate this further?

BU: Of the remaining respondents, 13% said that practical AI applications in the life sciences industry are limited, saying the technology is too theoretical at the moment. This further reiterates stakeholders needing tangible evidence of AI’s benefits to justify further investment.

On the other hand, 10% said that AI has already led to a lot of breakthroughs in R&D. It’s true that some areas have seen AI success, such as accelerating existing workflows in small molecule discovery and lead optimization for new drug candidates – shaving months off time-intensive processes.

The final 7% said that only cutting-edge startups and tech-led companies can use AI effectively, but we don’t believe this is the case at the Alliance. Established pharma companies possess substantial data science expertise and a wealth of data that have the potential to build robust AI models. The key lies in implementing the right data standards, developing a clear plan for AI use cases and actively collaborating with and learning from the successes of other companies. Facilitating this level of collaboration is exactly the reason why the Pistoia Alliance exists.

Dr. Becky Upton was speaking to Anna MacDonald, Senior Science Editor for Technology Networks.

About the interviewee:

Dr. Becky Upton was appointed as the Pistoia Alliance’s first female President in June 2022. She is a long-time supporter of pre-competitive collaboration in R&D and the critical role it plays in advancing science. Becky currently leads the Pistoia Alliance’s strategy, defining its future within areas of increasing importance to the industry, such as data standards, emerging technologies, diversity and inclusion, sustainability, and precision medicine. Becky has worked across business development and scientific services for companies including VWR (now part Avantor) and Pion. She has a PhD in Biochemistry from Imperial College and an MBA from Cranfield University.