PIM: Parallelization of Ising Model for Genomics Data

Ising model was originally designed to address the interactions among the atoms inside magnetic field. As it can fit into many biological problems where adjacent entities can interact with each other, Ising model is geting more and more popular. with its help, people may have a deeper and better understanding of associations between two related entities like genes and their products. However, it may require a lot of time to solve Ising model problem when the size of the dataset is large. To tackle the speed problem, we introduce two different parallelized implementations of Ising model problem on two mainstream processors in high performance computing field, multi-core CPUs and many-core GPUs. The test results show that they can increase the performance with different speedups on datasets with varying sizes.