Week 3 - Alex

Week 3

Being at the end of the program, I feel comfortable saying the last two weeks may turn out to be one of the more consequential experiences of my Ph.D. Not because of what I learned, but because of my exposure to modern cardiac EP modeling tools. The first two days focused on machine learning and data mining. These lectures were less informative for me, as I took a machine learning course last year. I was still far from understanding all of the math in the data mining lecture. However, I was aware of most of the tools, and knew where I could look to learn more about them.

Wednesday through Friday focused on applying FEniCS to solve multi-cell cardiac EP problems, then applying those tools to a summer project. This was the portion of the course that I think will have the biggest impact on my future research. I am working with two students to develop and probe a multi-cell microtissue and whole heart EP model for arrhythmic behavior after drug application. 

This project requires the use of FEniCS, Docker, and a cell model repository called CellML. Both FEniCS and CellML are new tools to me. FEniCS, as I mentioned in my week 2 post, is a fast C++-based PDE solver with an intuitive Python wrapper. Additionally, FEniCS solves single-cell ODEs in a fraction of the time of my current implementation. It allows for easy implementation of new single cell models into a micro-tissue, and hopefully a whole heart model. While scaling up from the single cell model wasn’t on my radar before this program, it may become one of the Specific Aims for my thesis after this project. 

The CellML repository includes hundreds of cardiomyocyte ODE systems, implemented in a format called CellML. A research scientist at Simula shared a module with me that can convert any CellML cell model to a FEniCS-readable Python file. This tool will enable me to make comparisons between multiple models, and run fitting algorithms to compare the conductance differences between different cell models. For example, this may be important as many people are interested in understanding the differences between iPSC-derived CM models and mature human CM models. 

Finally, Docker is a tool that is used to ensure complex programs with many software dependencies can be run on any computer. It works by running a virtual environment with all dependencies for a program. This is particularly helpful for FEniCS, as it’s a large and difficult-to-install software package. Also, it is possible to create a Docker environment that can run, and plot all of the figures for a given simulation. This can be used to address the issue of reproducibility. For example, I could create a Docker environment (called an Image) that anyone could download and run on their own computer – the image would run any simulations and plot all of the figures from a published paper.

I expect that my experience working on a summer project with FEniCS, CellML, and Docker will set me up to write a strong thesis proposal by next Spring.