and Optimization for Biomedical Research. Peng Zhang1, Chao Gao1, Na Zhang2, Li Zhang2, Seetha Pothapragada2, Yuefan Deng2,3, Danny Bluestein1.
Cloud Supercomputing: Multiscale Simulation Techniques and Optimization for Biomedical Research Peng Zhang1, Chao Gao1, Na Zhang2, Li Zhang2, Seetha Pothapragada2, Yuefan Deng2,3, Danny Bluestein1 Departments of 1 Biomedical Engineering and 2 Applied Mathematics, 3Institute for Advanced Computational Science, Stony Brook University, New York, United States Multiscale Simulation: The Coagulation cascade of blood may be initiated by flow-induced platelet activation, which prompts clot formation in prosthetic cardiovascular devices and in arterial disease processes. We employ a multiscale simulation which interface nanoscale microstructures of human platelets and microscale transport of blood flows, for providing a more accurate flow-induced dynamic stress mapping on platelets and predict their activation [1-3].
Multiscale Optimization and Acceleration We exploit the Triple-Punch acceleration strategy: (1) MTS (multiple time-stepping) Algorithm is developed for handling the disparity in integration stepping sizes. The results showed 3000x reduction in computing time over standard methods for solving the model. (2) Supercomputing: for a system of 10.89 million particles (16 platelets), simulating 1-ms phenomena can be reduced from 973 days to 77 days (512 cores). (3) GPU Acceleration: the GPU system is 2~3 times faster than the CPU-only system. [3] Platelet Flipping: The results indicate that neglecting the platelet deformability may overestimate the stress on the platelet membrane, which in turn may lead to erroneous predictions of the platelet activation under viscous shear flow conditions.
The mechanotransduction process of flow-induced platelet activation in pathological blood flows plays a key role in platelet-mediated hemostasis and thrombosis.
Multiscale Techniques: Two spatiotemporal scale methods [1, 2]: (1) Top/microscale using dissipative particle dynamics (DPD) to describe viscous blood fluid flows (viscosity, compressibility); (2) Bottom/nanoscale using coarse-grained molecular dynamics (CGMD) to describe the platelet membrane, cytoplasm and the cytoskeleton.
Figure: Effect of deformability on surface stress accumulation [1]
Cloud Supercomputing for Biomedical Research
Conclusions: The Algorithmic efficiency and the infrastructural scalability are two crucial factors for efficient use of large-scale computer systems including Stampede (US) and Tianhe-2 (CN). With the advent of novel algorithms and the advance of computer engineering, the era of solving biological systems with nanoscale details to a time scale of seconds is starting. Bioengineering It is not difficult to project a future that more complicated biological engineering problems involving hundreds of platelets can be simulated using millions of computer cores and advanced accelerators, in near future.
SuperComputers
References: [1] Zhang, P., Gao, C., Zhang, N., et al., "Multiscale Particle-Based Modeling of Flowing Platelets in Blood Plasma Using Dissipative Particle Dynamics and Coarse Grained Molecular Dynamics", Cellular and Molecular Bioengineering, vol. 7, no. 4, pp. 552-574, 2014. [2] Zhang, P., Zhang, N., Deng, Y., Bluestein, D., "A Multiple Time Stepping Algorithm for Efficient Multiscale Modeling of Platelets Flowing in Blood Plasma", Journal of Computational Physics, vol. 284, pp. 668-686, 1 March, 2015. [3] Zhang, N., Zhang, P., Zhang, L., Zhu, X., Huang, L., Deng, Y., "Performance Examinations of Multiple Time-Stepping Algorithms on Stampede Supercomputer", XSEDE15, July 26-30, 2015.