Abstract
Pharmacometrics (PMX) models form the foundation of model-informed precision dosing (MIPD), underpinning and supporting clinical decision-making. However, PMX models are often complex, requiring significant computational effort for numerical solutions, with covariate selection typically being the most time-intensive step. Integrating PMX with advanced data science tools, such as artificial intelligence (AI) and machine learning (ML), could provide efficient methods for processing large, heterogeneous datasets, enhancing PMX models by supplementing small sample sizes and expanding parameter inference. These tools would have strong predictive and learning capabilities to aid model development and improvement, though they provide less biological interpretability than PMX. AI and ML have gained increasing popularity across various disciplines in the last decade, but their application in MIPD remains in its early stages, holding promise for the future.
Precision Dosing and Reinforcement Learning
Precision dosing, which customizes drug doses to optimize therapeutic outcomes and minimize risks, is crucial for medications with narrow therapeutic windows and severe side effects. Adaptive dosing strategies build on this by adjusting doses over time based on changing patient conditions. Reinforcement learning (RL) aligns well with this approach, as it mirrors the sequential decision-making process used by clinicians to adjust treatments according to patient responses. This webinar explores the potential of integrating RL with population pharmacokinetic/pharmacodynamic (PK/PD) models to create precision dosing algorithms. It reviews key studies where PK/PD models were embedded within RL systems to predict the impact of dosing decisions. Additionally, It discusses how to frame precision dosing problems within the RL framework, including system states, actions, and rewards, and how PK/PD models can enhance RL techniques.
AI-Driven De Novo Antibody Design
The rapid evolution of artificial intelligence has opened new frontiers in biologics discovery, particularly in the design of therapeutic antibodies targeting previously intractable molecules. Antibodies against GPCRs and ion channels—critical yet historically difficult-to-drug targets—are increasingly recognized as pivotal in addressing a wide range of diseases. Despite significant advancements, challenges remain in designing highly specific and functional antibodies with favorable developability profiles.
This seminar provides an overview of AI-driven approaches to de novo antibody discovery, emphasizing the integration of computational tools and experimental workflows. By employing generative models and
advanced clustering techniques, it is now possible to predict high-affinity binders with exceptional precision and efficiency. Examples will be drawn from recent successes in rapidly generating therapeutic candidates with nanomolar affinities and high functional diversity. In addition, the seminar will explore the potential of these methodologies to address broader therapeutic challenges, such as inflammation, highlighting their implications for advancing immunological research and clinical practice. The session concludes with a discussion on the future trajectory of AI applications in biologics, including the potential for accelerating innovation in antibody therapeutics.
Dr. Nikola Stefanović graduated from the Faculty of Medicine, University of Niš, Department of Pharmacy, in 2009 and completed his PhD at the same institution in 2015. Since 2011, he has been employed at the Faculty of Medicine in Niš, advancing through positions including research trainee, teaching assistant, and assistant professor. In 2023, he was promoted to Associate Professor in the field of Pharmacokinetics and Clinical Pharmacy. Since 2020, he has been a specialist in pharmacotherapy, and in 2021, he became the coordinator of the elective course Basics of Pharmacogenetics and Personalized Therapy. His research focuses on pharmacokinetics, the optimization of drug treatment, and investigating sources of inter-individual variability in drug response, with a particular emphasis on pharmacogenomics. His work is primarily centered on immunosuppressive drugs used in kidney transplantation.
Dr. Mehm
et Itik achieved his Ph.D. degree in Automatic Control and Systems Engineering from the University of Sheffield, UK. He currently serves as a Full Professor in the Department of Aerospace Engineering at Dokuz Eylül University, Turkey. He has held prestigious fellowships, including the Endeavour Fellowship at the University of Melbourne, Australia, where he explored advanced control methods, and the IAS Fellowship at the Institute for Advanced Study, University of Amsterdam, where he conducted research on mathematical models of antibiotic resistance.
Dr. Itik’s research expertise spans control systems, robotics, and the mathematical modeling of complex dynamical systems. His work focuses on the mathematical modeling and control of systems, with an emphasis on reinforcement learning and optimal control, applying these methods to real-world problems like improving robotic systems and cancer treatment optimization. He uses approaches such as nonlinear systems, chaos theory, control theory, and computational modeling to develop predictive models for clinical applications. He has also worked extensively on the modeling and control of cable-driven robots and smart materials used in automation and industrial systems.
Dr. Serbülent Ünsal is a scientist affiliated with Antiverse ltd. and Karadeniz Technical University. With over 15 years of experience at the intersection of life sciences and AI, he leverages advanced technology to drive translational research and therapeutic innovation. His career is defined by a commitment to enhancing human health through AI-driven solutions and interdisciplinary collaboration. As a Senior Machine Learning Engineer at Antiverse, Serbülent leads efforts in therapeutic antibody design. Utilizing cutting-edge AI models such as Large Language Models (LLMs), Graph Neural Networks (GNNs), and Diffusion models, he has significantly reduced therapeutic discovery timelines from two years to six months, with the ambitious goal of achieving one month. In academia, Serbülent has spearheaded research projects in cancer biology, immunotherapy, and bioinformatics, supported by a Ph.D. and M.Sc. in Medical Informatics. He has served as a coordinator and researcher on numerous projects, transforming academic insights into practical applications with measurable impact. Passionate about fostering the startup ecosystem, Serbülent recognizes startups as key drivers of technological advancement. He actively supports biotech innovation and values their crucial role in shaping the future of life sciences.