A parallel computing paradigm for transcription network construction from microarray data

Deciphering transcriptional networks that convey information on members of gene clusters and cluster interactions is a crucial analysis task for gene expression data. To eliminate the ambiguity and subjectivity of human decisions based on Random Matrix Theory demands tens of thousands of computing cycles to calculate all eigenvalues and eigenvectors of the huge correlation matrices of gene expression data. Thus, high performance computing resources such as supercomputers and Linux clusters as well as parallel programming techniques are being utilized to address this problem. We propose and develop a parallel framework of utilizing MPI C Program and ScaLAPACK routine PDSYEVX/PDSYEV in this paper so as to solve the computing bottleneck of processing real yeast gene expression data.; This paper will be a complete instruction manual for how to use ScaLAPACK routine PDSYEVX/PDSYEV to get selected eigenvalues and eigenvectors. At the mean time, it is aimed to construct a parallel paradigm using ScaLAPACK to solve computing problems.