RII Track-4:NSF: Physics-Informed Machine Learning with Organ-on-a-Chip Data for an In-Depth Understanding of Disease Progression and Drug Delivery Dynamics

Kansas State University, Manhattan KS

Investigators

Abstract

Conventional animal models used in drug development often fail to accurately mimic the human body's complexities, which poses significant challenges in creating effective medicines. Furthermore, these animal experiments raise ethical concerns and the need to reduce their use. To tackle these issues and advance drug development and personalized treatments, this NSF EPSCoR RII Track-4 Research Fellows project focuses on creating computational models of healthy and diseased tissues and organs closely resembling the human body dynamics. The project integrates cutting-edge organ-on-a-chip (OoC) experiments with the advanced computational tools and physiochemical-based multiscale models to predict how diseases progress and how drugs interact with the body. This innovative approach also improves the OoC experiments by reducing the number of experiments needed to get useful data. It makes the preclinical process more efficient and helps develop more effective drugs with the right doses and fewer side effects. The research has major benefits for society: it speeds up the discovery of effective drugs, potentially tailors treatments to individual patients, reduces side effects and treatment failures, and ultimately leads to better, quicker, and more affordable healthcare while reducing the need for animal testing. The remarkable potential of OoC technology to accelerate drug discovery and reduce the associated costs necessitates developing a state-of-the-art framework to achieve and assess it. The research focuses on developing a learning-based multiscale modeling framework to enhance the understanding of drug delivery dynamics using OoC data. This fundamental and highly challenging problem will be addressed by a hybrid modeling approach integrating machine learning (ML) with the first-principles multiscale models. The proposed hybrid model has better properties than the standard ML-based models. It can accurately interpolate and extrapolate the OoC data. It is easier to analyze, interpret, and requires significantly fewer training samples. Such advantages are rational due to leveraging the benefits of theoretical and data-driven modeling approaches. Furthermore, our integrated approach optimizes the OoC experiments by minimizing the required experiments to collect informative data, increasing preclinical process efficiency, and guiding toward developing more effective drugs with optimal dosages and fewer side effects. To accomplish this, the EPSCoR Track-4 Research Fellows program supports an Assistant Professor and a graduate student at Kansas State University to visit and collaborate with one of the leading bioengineering research institutions, the Terasaki Institute for Biomedical Innovation (TIBI). The proposed approach will be benchmarked on liver-on-a-chip systems, a well-established OoC technology at TIBI for modeling nonalcoholic fatty liver disease. In addition, the research will provide a platform for interdisciplinary student training, mentoring, and engagement with the community. The PI aims to produce chemical engineering graduates with high-level mathematical, computational, and data-science expertise. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.