Skip to main content
ISOP ACoP 2025 Annual Conference Main banner
ViewAttendees 8
Symposium

Stats/ AIML/ Big Data

C1B - Harnessing Physics-Informed Machine Learning for Predictive Pharmacology and Oncology Drug Development

Sunday, October 19, 2025
11:00 AM - 12:30 PM MDT
Location: Willow Lake 3

Disclosure(s):
Nazanin Ahmadi Daryakenari, PhD: No financial relationships to disclose
Mohammad Kohandel: No financial relationships to disclose
Hongxiang Hu, PhD: No financial relationships to disclose
Allison Claas, PhD: No relevant disclosure to display

Description of session (include background & scientific importance): Pharmacokinetic-pharmacodynamic (PK-PD) models play a crucial role in understanding the relationships between drug administration, systemic exposure, and therapeutic response. However, traditional PK-PD models often struggle to make accurate predictions in real-world settings, particularly for limited and noisy clinical data, inter-patient variability, and complex drug-tumor interactions. These limitations highlight the need for advanced scientific machine learning (SciML) approaches that combine mechanistic knowledge with data-driven insights to improve predictive accuracy and optimize treatment strategies. Recent advancements in physics-informed neural networks (PINNs) and deep sequence models offer a novel approach to embedding physics-based constraints into machine learning models, ensuring better generalization even when data is scarce. By incorporating physics-informed learning into PK-PD modeling, these approaches can improve early pharmacodynamic outcome predictions, provide uncertainty quantification to enhance confidence in treatment response predictions, and optimize treatment strategies, particularly in the context of combination therapies and dose adjustments. This symposium will explore cutting-edge SciML techniques for PK-PD modeling, presenting case studies that highlight the efficacy of these methods in oncology drug development. By bridging the gap between mechanistic modeling and machine learning, these advanced techniques offer a pathway toward improving predictive accuracy, patient-specific treatment optimization, and clinical decision-making.

Inspired by the conference theme "Ascending New Heights: Envisioning The Future Landscape of Quantitative Pharmacology," this session unifies diverse approaches - from PINNs to transformer-based models - to chart a transformative future in precision oncology. The presentations form a cohesive narrative that bridges traditional PK-PD modeling with innovative machine learning strategies, exemplifying the next evolution in predictive pharmacology.

Learning Objectives:

  • Upon completion, participants will be able to understand the role of advanced machine learning models in improving PK-PD predictions, particularly in oncology drug development, and compare the performance of transformer-based models in longitudinal PK-PD analysis to traditional approaches such as LSTMs and neural ODEs. They will also explore opportunities for integrating mechanistic models and machine learning, either as hybrid models or complementary approaches, in early-phase drug development. Additionally, participants will gain valuable insights into industry and regulatory perspectives on adopting advanced analytics in precision oncology, equipping them to contribute to the next generation of quantitative pharmacology.

Presentations:

  • 11:00 AM - 11:22 AM MDT
    State-Space Model for Predicting Antitumor Drug Efficacy

    Speaker: Nazanin Ahmadi Daryakenari, PhD (she/her/hers) – Biomedical Engineering, Brown University

  • 11:22 AM - 11:45 AM MDT
    Integrating Quantitative Systems Pharmacology with Machine Learning to Enhance Predictive Accuracy and Personalization in Cancer Immunotherapy

    Speaker: Mohammad Kohandel – University of Waterloo

  • 11:45 AM - 12:07 PM MDT
    Exploring the Transformer Model in Longitudinal Pharmacokinetic/Pharmacodynamic Analysis

    Speaker: Hongxiang Hu, PhD (he/him/his) – Bristol Myers Squibb

  • 12:07 PM - 12:30 PM MDT
    In silico first – by harnessing the strengths of AI/ML + PK/PD to drive the discovery of targeted protein degraders

    Speaker: Allison Claas, PhD (she/her/hers) – Novartis