Deterministic simulations of large-scale models of cellular processes accelerated on Graphics Processing Units

The computational investigation of mathematical models of cellular processes represents an essential methodology in Systems Biology, complementary to conventional experimental biology, to understand the emerging behavior of biological systems. The simulation of the dynamics of these models, usually required for computational studies such as parameter estimation or sensitivity analysis, can become burdensome if the corresponding biochemical reaction network is characterized by hundreds or thousands of different molecular species and reactions. In this work, we introduce a novel GPU-powered fine-grain deterministic simulator of large-scale models of biochemical reaction networks, and test its computational performances on a set of randomly generated synthetic models of increasing size. We show that our parallel simulator, running on a GPU Nvidia GeForce GTX Titan Z, outperforms the sequential version, running on a CPU Intel i7-4790K 4.00GHz, achieving up to 7.8× speed-up