2020 Virtual undergraduate Research symposium
Utilizing a Constrained Ensemble Kalman Filter to Determine Optimal OGTT Duration for Estimating Insulin Sensitivity
PROJECT NUMBER: 2
AUTHOR: Griffin Hampton, Applied Mathematics and Statistics | MENTOR: Cecilia Diniz Behn, Applied Mathematics and Statistics
ABSTRACT
—
Quantification of insulin sensitivity (SI) is crucial for characterizing metabolic disease and optimizing patient care, but protocol design may affect estimates of SI. Fitting the Oral Minimal Model (OMM), a differential equations-based representation of glucose-insulin dynamics, with glucose and insulin concentration data from an oral glucose tolerance test (OGTT) enables estimation of SI under physiologic conditions. During an OGTT, plasma glucose concentrations increase as ingested glucose enters the blood and decrease as glucose is taken up by peripheral tissues. In insulin resistant (IR) individuals, disruption of typical glucose-insulin dynamics may cause estimated SI accuracy to depend on OGTT protocol duration with shorter durations producing unreliable estimates. To determine the shortest OGTT protocol duration for which OMM provides accurate SI estimates, we applied a constrained ensemble Kalman filter (CEnKF) to OMM to investigate the convergence of SI estimates over a six-hour OGTT protocol in an IR population of obese adolescent girls. We found that SI estimates generally improved with protocol duration, however, shorter protocols may be sufficient to estimate SI within a defined threshold. This approach provides insight into the relationship between data and OMM parameters across the OGTT, and it may inform OGTT protocol design for other IR populations.
VISUAL PRESENTATION
—
AUTHOR BIOGRAPHY
—
Griffin is a Boettcher scholar at the Colorado School of Mines and is a sophomore majoring in Computational and Applied Mathematics with a second major in Biochemistry. He is a McBride Honors student getting a humanities minor in public affairs. He is an active research student in Dr. Cecilia Diniz Behn’s research group and is an active member of the combined math bio research group. He is intrigued with metabolism and has a special interest in glucose and insulin dynamics problems. He has enjoyed using data assimilation techniques to study glucose and insulin dynamics and wants to continue exploring metabolic processes to better understand them.
Nice work! Seems like the 2-hour and 6-hour results are almost equally good. Might be good to discuss this. Also, you might consider checking for highly influential points in your regression and think about if they might be outliers.
Thank you Dorit. I had a good time working on this project. They do seem pretty equal and is one of the ideas we want to investigate as we move forward. That outlier in the data is an interesting case. Because there is still some variability in the system, we don’t always get convergence, i.e. we get individuals like the outlier in my regression plots. I have had other runs where this entire population has a constant of correlation at 0.99 at 2hr and 6hr. I think moving forward, it would be useful to analyze multiple runs of the whole population.