Samuel, these are excellent questions. We'll include as many as possible and follow up here with the answers.
Original Message:
Sent: 04-07-2026 16:09
From: Samuel Miles
Subject: Call for Questions: PAGE 2026 AI/ML Panel Discussion – Your Voice Matters!
Hi Elba,
I unfortunately will not be able to attend the PAGE event this year but I have a few questions that have been rattling around in my head for a while that I think might be interesting for this group. Feel free to use any of them you think might make sense for the session, I would love to be able to see a recording or have a follow up post with how the panelists answer if any of the questions are used!
Regulatory Theme:
1. What are the current expectations around traceability when it comes to using AI/ML on regulatory submissions? Say for example an LLM is used to transform a raw dataset into an analysis-ready dataset, is there ay guidance on a chain of traceability that would be expected to be maintained such that those transformations could be replicated (through a runnable script for example)?
2. How much "black-box" predictive modeling is acceptable in a NDA today? Is there specific guidance on how much we can rely on these models for predictive capabilities in regulatory submissions? Where do we draw the line on requiring additional methods for modeling & simulation which are more explainable versus when it might be ok to skip certain pieces of the discovery/development process?
3. Is there a future reality where "black-box" models are used as quick preliminary predictors that help make decisions about whether to take a drug forward in the discovery/development process? I am also curious to know where in this process do they think it makes sense to inject this predictive capability. Before FHD? Before moving to Phase 2/3? Somewhere else entirely or not at all?
4. Additionally, is there a reality where we have enough trust in "black-box" models that we can use them to skip certain aspects of drug discovery/development process, say skipping Phase 1/2 for example if we feel confident enough in the models predictive behavior to move straight into a larger population? If yes, what would it take for us to be confident enough in these "black-box" approaches? Do they need to be fully explainable or is there a threshold by which enough data has been used or it has been proven successful enough times that it is considered correct?
Industry Theme:
5. Are there real-world examples you can point to out there today of AI/ML models being implemented in a re-usable framework with an MLOps pipeline in place to prevent model drift when applying these techniques to the discovery/development pipeline? What technologies/platforms did they use to achieve this? What were some of the major challenges that were faced when scaling these capabilities from the testing environment into a production workflow for an entire team/organization?
6. Additionally, how are organizations thinking about the reusability of trained ML models across programs? For example, a PK prediction model trained on one therapeutic area being applied to a new molecule in the same class. What validation framework, if any, should exist before trusting cross-program generalization, and who owns that validation: the modeler, data science, or a centralized function?
7. How are your teams using AI/ML today? Are there consistent tools/technologies (Web interface, Terminal integration, App & coding space integrations, etc...) across the entire team or is it more free-for-all? Is there centralized internal guidance for your team or just open exploration and shared learning? Has one strategy worked better for your team to adopt and embrace these techniques quickly?
Use-Case Theme:
7. What are your opinions on using AL/ML and more specifically LLM's for results evaluation and report writing? I am curious to know the panels thoughts and concerns around having a LLM interpret results to write a first-draft report that could be reviewed and updated internally based on whether the accuracy of the first-draft.
8. Where do you see the most immediate, practical value of AI/ML in pharmacometrics today? Is it in the early data science/exploratory phase (dataset QC, covariate exploration, outlier detection), the modeling phase (structural model selection, parameter estimation), or the interpretation/communication phase (simulation summarization, report generation)? Which of these is most mature and where is the biggest gap still in your experiences?
------------------------------
Samuel Miles
Automation Lead
Eli Lilly and Company
Indianapolis IN
Original Message:
Sent: 03-26-2026 06:38
From: Elba Raimundez Alvarez
Subject: Call for Questions: PAGE 2026 AI/ML Panel Discussion – Your Voice Matters!
Hi AI/ML SIG Community,
We're excited to invite you to participate in our PAGE 2026 AI/ML Satellite Event panel discussion on "Use Cases and Regulatory Considerations of AI in Pharmacometrics and Clinical Pharmacology" – and we would love to hear YOUR questions to shape the conversation!
📅 Event Details: https://aiml-sig.github.io/2026-page-workshop/
🌟 Our Expert Panelists:
- Dr. James Lu – Senior Principal Investigator at A*STAR Bioinformatics Institute; Distinguished AI Scientist at Genentech
- Dr. Flora Musuamba – Assessor at Federal Agency for Medicines and Health Products; Professor at University of Namur; EMA Scientific Advice Working Party Member
- Dr. Hao Zhu – Director of Division of Pharmacometrics at FDA; ICH M15 Regulatory Chair
The session will be recorded! We're exploring ways to share it with the broader community, so even if you can't attend the workshop, your questions can still be part of the discussion.
💡 What We're Looking For:
What burning questions do you have about AI in pharmacometrics and clinical pharmacology? Whether you're curious about:
- Real-world use cases and applications
- Regulatory expectations and challenges
- Best practices and lessons learned
- Future directions and opportunities
📝 How to Participate:
Simply reply to this post with your questions by April 26, 2026. All questions will be reviewed and the most relevant ones will be incorporated into our panel discussion.
Let's make this a truly community-driven conversation! 🚀
Looking forward to your questions!
------------------------------
Elba Raimundez Alvarez
Sanofi
------------------------------