Hybrid Parallel Computation for Sparse Network Component Analysis

The gene regulatory network analysis primary goal is understanding the gene interactions topological order and how the genes influence each other. Network component analysis is a vital technique for build gene regulatory network. However, the network component analysis technique is time consuming and computational intensive. Therefore, parallel techniques are required. PSparseNCA is a parallel network component algorithm. This work present an improved version of PSparseNCA, referred as hPSparseNCA (Hybrid Parallel Computation for Sparse Network Component Analysis). hPSparseNCA uses the hybrid computational model to enhance the performance of PSparseNCA. hPSparseNCA is outperformed PSparseNCA achieving speedup reached 192.77 instead of 36.03 for PSparseNCA on 40 processing nodes. Furthermore, the speedup of the proposed algorithm reached 728.48 when running on 256 processing nodes.