PE: Today you appeared in a keynote interview titled “AI Perspectives from the FDA.” For the audience, could you briefly describe the discussion and the questions that stood out to you?
Fakhouri: One thing I found interesting was the FDA's experience with submissions that include AI and machine learning components. I explained that since 2016, we have received over 300 applications involving AI and machine learning components across various stages of drug development, the majority of which are clinical studies. Another hot topic was the FDA's role in making this technology a reality across the board. We also discussed the fact that AI is being used a lot in drug discovery, such as predicting molecules that might be targets for clinical trials and predicting protein folding. Because these treatments ultimately go through clinical trials, this is typically outside the scope of the FDA's review, which is where the FDA would consider this data.
The third topic that I found really interesting was how the FDA is considering the effectiveness of these models used in drug development and what areas they are focusing on. did. The way we look at these models is that we want to make sure they are reliable in a given usage context. Therefore, you will be asking about the data used to train or develop these models. We want to make sure the data is good for use and robust enough to develop good machine learning or AI models. Also, examine the performance of these models to ensure that they are working as intended. do.
PE: You recently led the development of a discussion paper titled “Using Artificial Intelligence and Machine Learning in Drug and Biologic Development.” Could you discuss the key points of this paper?
Fakhouri: This discussion paper was published in May 2023 and was actually accompanied by another discussion paper that focused on the use of AI in pharmaceutical manufacturing. In these discussion papers, we wanted to discuss areas where AI is used in drug development. What are the key considerations when using these tools? Again, issues related to AI governance, accountability, trust, transparency, and data quality, as well as issues related to model performance. Focus on the problem. The paper was published in May and he received over 800 comments from 65 different organizations. We're actually going to analyze all of these comments and use all of that information. We received comments from pharmaceutical companies, small biotech companies, and academia. We will use that information to inform guidance that is being developed this year and will hopefully be published by the end of this year.
PE: Artificial intelligence is a fairly new concept when it comes to pharmaceuticals. How do you think FDA policy regarding its use in drug development will change as it evolves?
Fakhouri: For the FDA, our approach has always been to approve drugs on a risk-based basis. We will also do our best to respond to new technologies. This is not unique to AI, given the use of real-world data and the use of digital health technologies in clinical research. From a clinical trial perspective, these are all new technologies that are continually evolving. We plan to grow with all these new tools as they become available. As I mentioned, we received over 300 submissions that included AI and machine learning components. Depending on the specific circumstances of the risks associated with the use of that model, to what extent will you rely on information and data from that model to make regulatory decisions? Determine the type.
