About this Event
Overview
This hands-on workshop covers the integration of domain knowledge with data-driven methods in pharmacometric modeling using DeepPumas. We’ll bridge mechanistic NLME modeling with modern scientific machine learning (SciML), including Neural ODEs, Universal Differential Equations, and DeepNLME, to discover unknown dynamics, leverage rich auxiliary data through embedding models, and build individualized models while maintaining interpretability.
You’ll learn how techniques from generative AI connect to random effects modeling, how SciML methods enable data-driven discovery of biological mechanisms, and how pre-trained embedding models extract prognostic information from complex data modalities. The workshop shows how these seemingly disparate techniques (Neural ODEs, DeepNLME, embeddings, and generative models) relate to each other and integrate into a coherent framework for modern pharmacometric modeling.
Beyond teaching specific tools, the workshop offers perspective on the evolving landscape of pharmacometric modeling as machine learning transforms the field. It helps individual modelers upskill and strategists understand what will be important going forward.
Instructors: Niklas Korsbo and Lucas Pereira
What you will Learn:
The workshop covers material from classical NLME to DeepNLME, with a focus on practical implementation and real-world applications.
Day 1: From NLME to Scientific Machine Learning - Pumas Essentials: Hands-on introduction to population pharmacometric modeling with Pumas - Machine Learning Foundations: Neural networks, overfitting, regularization, and how they apply to pharmacometric problems - Scientific Machine Learning (SciML): Neural ODEs and Universal Differential Equations (UDEs) for discovering unknown dynamics in disease progression and treatment response - Random Effects Fundamentals: Understanding random effects, individual parameters, and population distributions in hierarchical models, building the theoretical foundation for DeepNLME
Day 2: DeepNLME and Intelligent Covariate Integration - Random Effects as Latent Spaces: The deep connection between NLME random effects and generative AI (GenAI), and why this matters for longitudinal modeling - DeepNLME - Conditional Generative Modeling: How DeepNLME extends SciML to hierarchical data, enabling individualization while learning from the population - Post-Hoc Covariate Integration via Augmentation: Using DeepPumas.augment to incorporate rich covariate information into existing NLME models without refitting the base model, learning how covariates predict individual parameters - Embeddings and Pre-Trained Models: Leveraging state-of-the-art pre-trained embedding models (from computer vision, NLP, and other domains) to extract prognostic information from complex data modalities (clinical text, imaging, omics) and seamlessly integrate them into NLME frameworks - Connecting the Pieces: How Neural ODEs, DeepNLME, embeddings, and augmented models work together in practice
Who Should Attend
This workshop is intended for pharmacometricians and quantitative scientists navigating the integration of machine learning into pharmacometric practice. Participants will benefit most if they:
- Have experience with population PK/PD or other hierarchical models
- Want to understand how modern machine learning can enhance mechanistic modeling
- Are interested in leveraging rich covariate data (imaging, omics, text) in their models
- Seek mathematically rigorous yet practically applicable methods
- Are helping their organizations understand and adopt these emerging methods
Key Takeaways
By the end of this workshop, you will:
- Understand the mathematical connections between NLME, generative AI, and scientific machine learning
- Implement neural-embedded dynamical systems for discovering unknown biological mechanisms
- Apply DeepNLME to create highly individualized models for longitudinal data
- Integrate complex covariates post-hoc into existing models using conditioning approaches
- Leverage pre-trained machine learning models (embeddings) to extract prognostic information from diverse data modalities
- Discover prognostic factors from high-dimensional data using machine learning within the NLME framework
- Recognize when and how to leverage these methods in your own work, understanding both their power and limitations
Why DeepPumas?
Machine learning is transforming pharmacometrics, but the path forward requires more than adopting black-box algorithms. The choice isn’t between mechanistic models specified a priori or purely data-driven approaches. Instead, we need methods that genuinely integrate both.
DeepPumas enables this integration through:
- Modularity: Separate development of mechanistic structure and data-driven components, then seamless combination
- Post-hoc flexibility: Addition of new covariates or data sources to existing models without refitting
- Principled hierarchical modeling: Leveraging population structure while respecting individual heterogeneity
- Maintained interpretability: Preserving mechanistic understanding where available while still leveraging machine learning’s predictive power
These aren’t just technical conveniences. They represent a different approach to scientific modeling in an era of increasing data richness and complexity.
Practical Information
- Format: Intensive hands-on workshop with mixture of lectures, live coding demonstrations, and guided exercises
- Prerequisites: Familiarity with population modeling concepts; basic programming experience helpful but not required
- Software: All materials use Pumas/DeepPumas (Julia-based); no prior Julia experience needed
- Materials: All code, data, and documentation provided
What You’ll Take Away
This workshop goes beyond teaching specific tools. It provides a framework for thinking about the evolving role of machine learning in pharmacometrics. You’ll gain practical skills for implementing these methods, theoretical understanding of why they work, and perspective on where the field is heading.
Whether you’re looking to enhance your own modeling capabilities or help guide your organization through the ML transformation in pharma, this workshop will equip you with both the knowledge and the practical experience to move forward confidently.
Pricing: $500 for Industry, $100 for Academia
Bring Your Team, Save More! Groups of 3+ qualify for a group discount. Email [email protected] to inquire.
Event Venue & Nearby Stays
Valamar Lacroma Hotel, 34 Ulica Iva Dulčića, Dubrovnik, Croatia
USD 108.55 to USD 535.38
