AI and Gen AI in Drug Development Overview from DPHARM 2024
Novo Nordisk’s Corporate VP of AI and Analytics, Dr Faisal Khan, provided the DPHARM 2024 audience with an overview of the role of AI/GenAI in medicine development and we are pleased to share a summary of the key takeaways from his presentation.
Key takeaways
• Existing uses and current efficiencies: Artificial intelligence, including generative AI, is enhancing efficiency and effectiveness in trial design, patient recruitment and drug manufacturing, and significantly improving efficiencies in operational processes, such as literature reviews, protocol generation and regulatory submissions.
• Data transparency, quality and algorithms: Successful AI applications in clinical research depend on high-quality data and appropriate algorithm selection. Poor data quality, biases and the complexity of integrating multiple data types can introduce significant risks to model accuracy and generalizability.
• Ethical and regulatory challenges: As AI becomes more integrated into clinical trials and drug development, ethical concerns surrounding patient privacy, data integrity and consent are becoming more pronounced. Regulatory frameworks, such as the EU AI Act, are emerging to ensure responsible use of AI in these sensitive areas.
Existing uses and current efficiencies
"There's virtually no area of the whole drug development life cycle where artificial intelligence doesn’t play a role,” said Dr Khan. AI is already being utilized in various capacities from early-stage drug discovery to post-market surveillance. One of the most prominent uses of AI has been in precision medicine, helping to identify which patient populations are most likely to benefit from a specific therapy by analyzing omics data.
AI-powered algorithms are now used to optimize patient recruitment by predicting which sites will enroll the most eligible participants and by identifying individuals most likely to meet the inclusion criteria. AI can also model patient recruitment curves, helping teams predict timelines and improve trial efficiency.
Generative AI tools, such as ChatGPT, are also enhancing operational efficiencies. Researchers, for example, can use generative AI to review and summarize recent literature more quickly, accelerating the process of gathering relevant data and insights and helping researchers stay current and informed.
"There's virtually no area of the whole drug development life cycle where artificial intelligence doesn’t play a role."
Data transparency, quality and algorithms
The foundation of AI's effectiveness in drug development is high-quality data. AI models are only as good as the data they are trained on, and inconsistent or biased data can lead to inaccurate predictions.
Wearable devices generate vast amounts of data, but this data can be noisy and subject to issues like varying calibration and device discrepancies. Integrating data from different sources presents additional challenges, particularly when the data comes at varying speeds and formats. These issues underline the importance of data consistency and the need for high-quality, clean datasets to build effective AI models.
The choice of algorithms is just as critical. Deep learning requires large amounts of data to function properly, and while deep learning models can detect complex patterns, they can also potentially become too reliant on irrelevant data features.
One example Dr Khan used was an algorithm that differentiated between wolves and dogs by looking at the surrounding terrain, versus the animals themselves. That example illustrated that if the data or algorithm isn’t transparent and well-understood, the results could be misleading or wrong.
Ethical and regulatory challenges
The ethical and regulatory challenges surrounding AI in drug development are becoming more pressing. AI’s reliance on patient data raises concerns about privacy, consent and data security. Dr Khan reiterated the need to ensure that patients’ rights are protected and that data is used accurately and with proper consent.
There are also concerns around introducing risk if AI generates synthetic data or manipulates existing datasets to improve model performance. The emerging regulatory frameworks, including the EU AI Act, are designed to address some of these concerns and ensure that AI is used responsibly and ethically.
"It is our responsibility to make sure we’re thinking about [artificial intelligence] appropriately."
Areas of caution for AI in drug development
Dr Khan stressed that AI is not a magical solution to all problems. One of the most significant pitfalls in AI is the "black box" nature of many models. Deep learning models are often opaque in their decision-making processes, making it difficult to understand how a model arrived at a specific conclusion. This lack of transparency is especially problematic in clinical applications where safety and accuracy are paramount.
AI might identify patterns that humans can’t see, but it also might misinterpret data or draw incorrect conclusions based on irrelevant factors. Understanding the algorithms and their limitations is essential, especially when applying AI to sensitive areas like patient recruitment, safety monitoring and clinical trial design. Generative AI tools could introduce errors if relied upon for critical decision making without human oversight.
Final thoughts on AI/GenAI
AI tools can significantly enhance the speed and accuracy of drug discovery, clinical trial design and patient recruitment, but successful implementation of AI requires collaboration across a variety of stakeholders, including data scientists, clinicians, ethicists and regulators. “It is our responsibility to make sure we’re thinking about [artificial intelligence] appropriately,” said Dr Khan, “to ensure that ethical, legal and data quality concerns are addressed.”
The promise of AI is in the ability to ingest complex datasets and provide insights that were previously unattainable. For AI to reach its full potential in clinical research, it must be used with a clear understanding of its limitations and risks. As the technology continues to evolve, clinical researchers must be vigilant about its applications to ensure that patient safety remains the top priority.
