Tumorigenesis is increasingly recognized as a multifactorial process driven not only by genetic and epigenetic alterations, but also by profound metabolic reprogramming that sustains uncontrolled proliferation, enables adaptation to nutrient stress, and promotes survival in hypoxic microenvironments. Capturing these metabolic shifts requires analytical and computational methods that monitor dynamic changes in key metabolites with high temporal precision. Here, we present an integrated experimental–computational framework that couples a multiplexed microfluidic bead-based aptasensor (GluLac-Capchip) with physics-informed neural networks (PINNs) to quantitatively monitor glucose and lactate metabolism in cancer. GluLac-Capchip integrates bead-based aptasensing mechanism with a high-efficiency micromixer and a deterministic lateral displacement washing module to automate incubation and purification, enabling sensitive and specific analysis in microscale samples with substantially reduced assay time relative to conventional enzymatic methods. The platform achieves limits of detection of 0.005 M for glucose and 0.007 M for lactate, while simultaneously quantifying both metabolites within 11 minutes. Leveraging these measurements, the PINN models infer key kinetic parameters with agreement to literature values and experimental trends, supporting predictive, mechanism-consistent reconstruction of metabolic dynamics. This combined sensing-and-modeling strategy provides a scalable route to connect rapid multiplexed metabolite profiling with data-driven mechanistic analysis of tumor metabolism.
Vandvajdi et al. (Thu,) studied this question.