An Efficient Biomechanical Cell Model to Simulate Large Multi-cellular Tissue Morphogenesis: Application to Cell Sorting Simulation on GPU

Abstract

In the context of tissue morphogenesis study, in silico simulations can be seen as experiments in a virtual lab bench. Such simulations can facilitate the comprehension of a system, the test of hypotheses or the incremental refining of a model and its parameters. In silico simulations must be efficient and provide the possibility to simulate large tissues, containing thousands of cells. We propose to study tissue morphogenesis at the cellular level using our virtual biomechanical cell model. This model is based on a mass/spring system and coupled to a multi-agent system. We validated the relevance of our model through a case study: a cell sorting. Moreover, we took advantage of the large parallelism offered by graphics processing units (GPU), which contain up to thousands of cores: we implemented our model with the OpenCL framework. We ran large scale simulations, with up to 106 of our virtual cells. We studied the performance of our system on a CPU Intel Core i7 860, and two GPUs: a NVidia GeForce GT440 and a Nvidia GeForce GTX 690. The absence of synchronization in our implementation allowed the full benefits of the parallelism of these hardwares.