The speaker for our next department tea will be Farah Fouladi '15, who will talk about the internship she did this past summer. Lunch will be available after her talk.
By mathematically modeling biological systems we can predict the effect of small changes in the system that are difficult to measure experimentally. This analysis requires multiple simulations of the model for every parameter variation. While this type of analysis is feasible for isolated cells, running many simulations of large systems of cells is extremely time consuming. To decrease the execution time of these model simulations, I have developed a parallelized environment, which utilizes the architecture of a graphics processing unit.
The GPU has many processor cores and the ability for thousands of threads to run concurrently on those cores. For my research, a mathematical model of the human ventricular myocyte (ten Tusscher & Panfilov, Am J Physiol 291: H1088–H1100, 2006) was programmed in CUDA. Taking advantage of the parallel hardware, cell computations along with different simulations of the model are completed on the GPU by many threads running simultaneously. To validate this computational design, cell membrane voltage values were compared with an already existing model implementation in MATLAB. Results showed that the added parallelization has no effect on the computational aspect of the model and that the execution time of the CUDA program decreases by orders of magnitude compared to the previously used MATLAB program.
Using this new computational environment, I analyzed one cause of reentry in cardiac myocytes. Reentry occurs when an electric propagation loops back on itself, abnormally re-exciting cells. There is a short window of time during which a stimulus can excite cardiac tissue and cause a reentry effect due to refractory tissue blocking action potential propagation in only one direction. My program is able to efficiently identify the stimulus-timing interval when reentry occurs in a loop of 1,000 cells.
This parallelized simulation environment minimizes computational execution time and provides a framework for further analysis of more complex and physiologically relevant systems of cells