SUMMARY

The goal of precision medicine is to provide optimal treatment to patients based on their individual characteristics or disease state. Current genomic-based approaches are limited by the amount of patient sample needed, high turnaround time, and reliant on reported mechanisms of drug action and patient response. Microfluidic devices provide a way to directly and efficiently test small quantities of patient samples for functional outcomes; these devices can incorporate features such as the cellular environment to better model physiological variables. Turning microfluidics-derived data into actionable precision medicine insights requires collaboration between the fields of microsystems engineering, computational biology, and clinical medicine. On the computational side, novel data analysis pipelines and interpretable statistical modeling methods are needed to extract the maximal amount of information from microfluidics data and to generate clinically actionable insights. The objective of this work was to leverage computational and mathematical approaches to develop robust predictive models of patient sample response assayed in microfluidic devices. This approach was validated using microfluidics-generated datasets from two hematologic applications: combination drug screening in leukemia, and rheological biomarker correlation in sickle cell disease. Specific outcomes were: 1) development of an analytical pipeline that measures drug synergy and efficacy metrics for combinations used in leukemia, 2) identification of a subpopulation of patients with sickle cell disease that may benefit from a novel therapy, and 3) correlation of microfluidic-based rheological metrics to symptomatic severity in patients with sickle cell disease. Overall, this work demonstrates the ability of the combined experimental-computational frameworks to extract important patient-specific features from multi-factorial experiments, optimize discovery of synergistic drug interactions, and provide personalized recommendations for therapy.