Candidate: Shadi Vandvajdi
Date: November 13, 2026
Time: 10:00 AM
Location: Online
Supervisors: Prof. Mahla Poudineh and Prof. Mohammad Kohandel
All are welcome!
Abstract:
Metabolic reprogramming enables cancer cells to adapt to changing nutrient and oxygen availability while sustaining proliferation. Although extracellular glucose and lactate provide accessible indicators of metabolic activity, their concentrations alone do not reveal the underlying pathways or cellular behaviors. This seminar presents an integrated experimental and computational approach that combines microfluidic biosensing, mechanistic modeling, and physics-informed machine learning to investigate cancer metabolism across population and spatial scales.
First, Physics-Informed Neural Networks (PINNs) and Universal Physics-Informed Neural Networks (UPINNs) are applied to a compartmental model of glucose–lactate metabolism to estimate metabolic parameters and learn an incompletely specified glycolytic relationship. Evaluation using synthetic data and experimental measurements from glioblastoma cell lines demonstrates the potential to infer cell-specific metabolic dynamics. Next, a multiplexed microfluidic bead-based aptasensor, GluLac-Capchip, is introduced for simultaneous glucose and lactate quantification in approximately 11 minutes. Repeated measurements in prostate cancer and patient-derived glioblastoma cell cultures are combined with PINN modeling to investigate proliferation and metabolic switching.
Finally, a microscale reaction–diffusion model describes glucose, lactate, and oxygen transport around a tumor microvessel, revealing spatial differences in metabolic pathways and ATP production. Inverse analyses show that accurate reconstruction of measured concentrations can coexist with substantial errors in inferred parameters and unobserved oxygen. Adding complementary synthetic ATP-production observations improves recovery under the tested conditions, highlighting the importance of informative measurement design.
Together, these studies connect rapid metabolite sensing with mechanistic interpretation while clarifying the capabilities and limitations of model-based inference in cancer metabolism.