Computational Methods to Assist the Development of Liver Cryopreservation Protocols

<p dir="ltr">Liver transplantation is a breakthrough treatment option that has saved innumerable lives worldwide. However, there is a significant unmet need with extensive transplant waiting lists and a general lack of availability to this life-saving procedure. Advanced preservation methods have shown the ability to significantly extend the ex vivo life of the organ, allowing for improved availability, better donor matching, assessment of marginal grafts which would typically be discarded, and even restoration of function. These advanced preservation strategies are highly complex processes with an incredible number of free design variables. Furthermore, there is scarce availability of human livers to develop these experimental protocols on. Researchers often use animal models instead, but successes are not directly translational due to differences in structure and scale. </p><p dir="ltr">To address these challenges, we propose the creation of computational models of the liver to conduct early phases of cryopreservation protocol development. While it is impossible to fully capture all the complexity in the human liver in a model, we see these computational models as a method to inform experimentalists to more efficiently conduct experiments with plausible protocols. These computational models will allow for broad exploration of a highly complex design space, which was not possible in the constrained experimental case. </p><p dir="ltr">We start by adapting an algorithm to generate three-dimensional vascular models of the liver. We demonstrate the ability of this method to create models of varying depth and topology, as well as the ability to generate models that closely match morphological statistics and structure. We create fully connected models of the liver, such that flow can be simulated in and out of the vasculature. With these models, we create a linear system of equations to simulate fluid flow throughout the entire vasculature. We then simulate the loading of cryoprotectants throughout the vasculature and highlight how the model can be used to interrogate the effects of varying boundary conditions on wall shear stress and cryoprotectant concentration throughout the model. We view this type of investigation as the model’s primary benefit to experimentalists and clinicians when developing new cryopreservation strategies. </p><p dir="ltr">We extend the liver vascular model to simulate heat transfer, and we couple the fluid and heat transfer models to account for heat transfer due to perfusion using the bioheat equation. The heat transfer model produces a temperature field that is used to update the viscosity of the perfusate in the fluid model. The fluid model then solves for vessel flow rates at every step, which are in turn used to determine the heat transfer due to perfusion. We step between the two models and can account for changes in the flow and heat transfer due to the freezing of specific vessels and regions in the liver. We demonstrate how the model can be used to investigate the effect of varied boundary conditions on the spatial and temporal behavior of temperature and flow throughout the model. </p><p dir="ltr">Finally, we utilize machine learning and Bayesian optimization techniques in conjunction with high-throughput screening techniques to optimize cryoprotective agent cocktails. We iterate between experiments to determine the viability of various cocktails, and machine learning methods that intelligently select the most informative and optimal prospective cocktails to test next. We optimize for multiple objectives, namely low toxicity and high concentration, by making use of state of the the art hypervolume Bayesian optimization methods. Similar to the liver models, this project leverages computational methods to accelerate experimental discovery for the benefit of cryopreservation protocol development</p>