Technical Articles

Improving a Quantitative Pharmacology Course with SimBiology at Harvard Medical School

By Jagesh Shah, Harvard Medical School, Lorette Noiret, Massachusetts General Hospital and Harvard Medical School, and Fulden Buyukozturk, MathWorks


Modern methods of evaluating drug action require not only a background in pharmacology but also a detailed understanding of cellular pathways and drug binding modes. Analytical models can be sufficient for straightforward studies of such pathways and modes, but studies involving multiple levels of drug action require more complex computational models.

In a course we teach at Harvard Medical School, graduate and postdoctoral students are learning how to construct complex pharmacological models with MATLAB® and SimBiology™. Most students who take the course are familiar with the basics of enzyme biochemistry and the use of simple compartmental models for pharmacokinetics (PK) and pharmacodynamics (PD). This course gives them hands-on experience in applying these concepts to pharmacological problems within a quantitative computational framework.

Course evaluations have been uniformly positive. Many of the students noted that the skills they acquired and concepts they learned in the course were immediately applicable to their own research projects. We also received feedback from other faculty members, who were pleased that students returning to their labs after taking the course were comfortable with applying computational pharmacology principles.

The director of the Therapeutics program at Harvard Medical School recently designated the course as fulfilling a quantitative requirement for the program.

Planning and Designing the Course

In a previous core course on drug action and pharmacology, students ran simulations using premade computational models. Course evaluations revealed that the students did not see the relevance of the models; some even reported finding the modeling assignments a waste of time. (Colleagues in pharmaceutical companies in Boston were flabbergasted to hear that the students had so little enthusiasm for a topic so central to the industry.)

To address this concern, Harvard Medical School faculty decided to offer a course that focused on computational pharmacology and emphasized hands-on assignments rather than canned examples. Local researchers would be invited to speak to the class about the value of quantitative models in understanding the principles of drug action.

Basing the course on SimBiology made it possible to offer the course without making expertise in programming a prerequisite. One lab at Harvard Medical School uses a Python® coding framework to construct models, and even postdoctoral students have difficulty applying that framework to their own experimental work. With SimBiology, students can build models graphically and run analyses, such as parameter estimation, with a few clicks. This aspect alleviates the challenges of programming and helps the students to focus on the modeling aspects.

Make the Most of Limited Class Time

The 12-week course met every week for two hours. To make the most of every minute in class, we decided to use a flipped classroom approach in which students learned the background material beforehand and spent class time working hands-on with SimBiology.

We asked the students to download and install MATLAB and SimBiology on their laptops before the course started. Harvard Medical School’s Total Academic Headcount (TAH) license made this process seamless. The TAH is part of a larger initiative at the school to encourage students to use MATLAB for data analysis, simulation of biological experiments, image processing, and other applications. The TAH meant that we did not have to worry about getting all the students running the right software for their operating system.

Incorporating Current Research and SimBiology Webinars

To run the flipped classroom most efficiently, we provided the students with material to review before class. For the first assignment, students came to class having read articles on liver injury due to acetaminophen overdose. In class, they integrated a dynamic model of acetaminophen metabolism and toxicity, which we provided, with a one-compartment PK model in SimBiology (Figure 1). The model is adopted from a paper published in Hepatology1. Running simulations using the model, students compared different initial oral doses of acetaminophen and answered questions such as what happens to acetaminophen plasma concentration if a patient has liver disease and which dose amount is fatal.

Figure 1. SimBiology model describing oral absorption, metabolism, and toxicity of acetaminophen.

Figure 1. SimBiology model describing oral absorption, metabolism, and toxicity of acetaminophen.

Many published research papers have associated SimBiology models. For some of the subsequent classes, we asked the students to study some of these papers and then work with the supplementary models. Whenever possible, we used existing papers with SimBiology models and online videos to ensure that the students were learning from relevant sources. Together, these resources reduced the amount of material that the teaching faculty needed to produce and provided models for the students to build on as part of the classwork.

For example, on one assignment we asked the students to work with a PK/PD model of drug distribution for tumor therapy. The underlying model is based on a paper published in Journal of Pharmacokinetics and Pharmacodynamics2. MathWorks offers a webinar on using SimBiology to create a model that uses the paper as a reference. The students read the paper and watched the webinar before class so that they were prepared to begin working in SimBiology when they arrived. The students inserted a tumor submodel into a physiologically based pharmacokinetic (PBPK) model with multiple compartments that we provided (Figure 2). They ran simulations to see how different parameter values affected the amount of the drug getting to the tumor.

Figure 2. SimBiology implementation of a physiologically based pharmacokinetic (PBPK) model coupled with a PD model of tumor growth inhibition.

Figure 2. SimBiology implementation of a physiologically based pharmacokinetic (PBPK) model coupled with a PD model of tumor growth inhibition.

Building on this model, the students ran other scenarios, including changing the PD readout, relocating the tumor, and modifying drug dosing schedules to find an effective therapy (Figure 3).

Figure 3. Simulation results showing of the effect of different dose amounts on drug concentration in tumor mass over time.

Figure 3. Simulation results showing of the effect of different dose amounts on drug concentration in tumor mass over time.

Lastly, we assigned an extended semester project based on the students’ own research. Students employed the modeling and analysis methods they learned throughout the course in their own research problems. Toward the end of the semester, they provided one-page description of the model they’d worked on using a rubric covering the background information, why the model was needed, what modeling assumptions were made, and the basic structure of the model itself. In the last two sessions of the course, the students presented their models to the rest of the class, providing the SimBiology file to their peers, and also created a one-page exercise for the class to explore their model.

Expanding the Course with Guest Lectures from Industry and Academia

To motivate students and help them appreciate the relevance of what they would be learning, we invited scientists from industry to speak about the use of SimBiology, PK/PD, and quantitative systems pharmacology (QSP) models in real-world research. Scientists from Genentech, Pfizer, and Rosa &Co guest-lectured in the course. They spoke about the use of SimBiology and quantitative models in real-world research, presented case studies, and provided their SimBiology models for students to explore certain aspects and test hypotheses. One of the case studies took advantage of the integration of SimBiology and MATLAB to enable the students to automate QSP analyses by using SimBiology programmatically via MATLAB scripts. For example, in one exercise students were asked to simulate and then compare virtual patient scenarios using a QSP model, where each virtual patient represents an alternative set of model parameter values representing a certain hypothesis about the model. To automate this analysis, they simulated the SimBiology model repeatedly for three virtual patient scenarios and for 12 different dose schedules using a MATLAB script.

We also included a lesson highlighting the use of modeling in pathophysiologic discovery, illustrated with a SimBiology model that describes HIV transport and latency in a variety of compartments and is also implemented by first-year students in the Harvard-MIT Health Sciences and Technology (HST) MD program. The lecture was presented by John Higgins, associate professor of systems biology at Harvard Medical School and an associate pathologist at Massachusetts General Hospital.

The importance of learning computational approaches to pharmacology will become clearer to these students as they progress in their careers. It is already clear to our faculty, who consistently rank computational approaches among the most important topics for students to learn.

References

1 Remine, CH, Adler FR, Waddoups L, Box TD, and Sussman NL. “Mathematical modeling of liver injury and dysfunction after acetaminophen overdose: early discrimination between survival and death.” Hepatology. 2012 Aug; 56(2):727-34.

2 Koch, G., Walz A., Lahu, G., and Schropp, J. (2009), “Modeling of tumor growth and anticancer effects of combination therapy.” Journal of Pharmacokinetics and Pharmacodynamics. 36:179-197.

About the Author

Jagesh V. Shah, Ph.D. is an associate professor of systems biology at Harvard Medical School. His research interests include investigating how cells make intracellular and extracellular measurements. His lab uses a mixture of quantitative biological perturbation and measurements and computational modeling.

Lorette Noiret, Ph.D. is an assistant professor at Pierre and Marie Curie University, and a former postdoctoral fellow at Massachusetts General Hospital and Harvard Medical School. Her research focuses on developing quantitative approaches to describe pathophysiological processes in order to find new biomarkers and help treatment decisions.

Fulden Buyukozturk, Ph.D. is an industry marketing manager at MathWorks. She focuses on computational biology market segment and its applications.

Published 2017 - 93177v00

View Articles for Related Capabilities

View Articles for Related Industries