People
W.A. "Drew" Pruett, PhD
 | Instructor Office/Lab: G152; (601) 815-1316 |
Areas of expertise
- General: Use of mathematical modeling to study cardio-renal axis, machine learning for data interpretation, visualization of high dimensional data sets for pattern recognition
- Specific: Systems integration of hormones, vasculature and tubular effects to maintain water and electrolyte homeostasis, topology-based analytic methods, support vector machine and ridge regression methods for surrogate model construction, nearest neighbor regression and Markov chain methods for model calibration/parameter inversion, in silico clinical trials of pharmaceutical or device interventions targeting the cardio-renal axis, Boolean and multilevel network analysis
Research methods
The computational modeling lab uses an integrative model of human physiology, HumMod, as its prime research tool, as well as a suite of ancillary software to calibrate, validate, extend, and analyze the model. Our commonly used methodologies include:
- Topological data analysis
- Visualization of high dimensional data
- Pattern recognition algorithms
- Surrogate modeling techniques
- Parameter inversion methods including nearest neighbor regression and Bayesian melding
Research Summary
My research interests include the use of physiological modeling to generate hypotheses and to understand integrative physiological mechanisms that are not observable in either whole animal or human experiments. For almost 50 years, our department has been developing computer simulations of integrative physiology for research and educational purposes. The current model, HumMod, is comprised of 14 organ systems, and includes neural, endocrine, circulatory, and renal physiology. We have created techniques that generate and analyze >1000 unique models (a population of virtual patients) by randomly varying underlying physiological parameters and relationships. Published data from our laboratory show that our model is robust and can realistically simulate salt sensitivity, multiple types of hypertension, and renal denervation in a virtual population. Our current research shows similar virtual population responses to an ACE inhibitor and renal denervation as compared to clinical data. Approximately 10% of U.S. patients with hypertension have uncontrolled blood pressure even with full adherence to 3 or more drugs (resistant hypertension). As compared to whites, African Americans develop hypertension at an earlier age, have a greater frequency and severity of hypertension, have poorer control of blood pressure, as well as a greater prevalence of comorbid conditions. Even after adjusting for clinical and socioeconomic factors, relatively high rates of resistant hypertension persist in African Americans. Despite these known disparities, there has been little attention or advancement in the blood pressure management in African Americans. My current research focuses on using clinical data, HumMod, and predictive analytic techniques to develop a realistic virtual African American population for studying antihypertensive treatments that have well-known (diuretic or salt reduction), variable (ACE inhibition), or unconfirmed (renal denervation and baroreceptor activation) therapeutic efficacies in hypertensive African Americans.