Competing computational approaches to reaction-diffusion equations in clusters of cells

2nd International Conference on Mathematical Modeling in Physical Sciences 2013

物理学;数学

Stella, Sabrina^1 ; Chignola, Roberto^2 ; Milotti, Edoardo^1,3

Department of Physics, University of Trieste, Italy^1

Department of Biotechnology, University of Verona, Italy^2

Istituto Nazionale di Fisica Nucleare, Sezione di Trieste, Italy^3

关键词: Abstract modeling;    Biochemical model;    Computational approach;    Molecular pathways;    Molecular species;    Reaction diffusion;    Reaction diffusion equations;    Reaction mechanism;   

Others  :  https://iopscience.iop.org/article/10.1088/1742-6596/490/1/012129/pdf
DOI  :  10.1088/1742-6596/490/1/012129

来源: IOP

【 摘 要 】

We have developed a numerical model that simulates the growth of small avascular solid tumors. At its core lies a set of partial differential equations that describe diffusion processes as well as transport and reaction mechanisms of a selected number of nutrients. Although the model relies on a restricted subset of molecular pathways, it compares well with experiments, and its emergent properties have recently led us to uncover a metabolic scaling law that stresses the common mechanisms that drive tumor growth. Now we plan to expand the biochemical model at the basis of the simulator to extend its reach. However, the introduction of additional molecular pathways requires an extensive revision of the reaction-diffusion part of the C++ code to make it more modular and to boost performance. To this end, we developed a novel computational abstract model where the individual molecular species represent the basic computational building blocks. Using a simple two-dimensional toy model to benchmark the new code, we find that the new implementation produces a more modular code without affecting performance. Preliminary results also show that a factor 2 speedup can be achieved with OpenMP multithreading, and other very preliminary results indicate that at least an order-of-magnitude speedup can be obtained using an NVidia Fermi GPU with CUDA code.