Microarray data analysis (MDA) has become an important tool in molecular biology. Modern microarray data provide vast amounts of useful biological information, but their analysis is computationally challenging. Molecular biologists need fast, reliable, and advanced MDA software, e.g., to locate clusters of genes responsible for specific biological processes. Novel mathematical algorithms for MDA, including the recently available GeneChip tiling arrays, are necessary to widen the bottlenecks of existing software, which primarily are slow performance and low accuracy. Our core expertise is in development of iterative methods and software for numerical solution of eigenvalue problems [1-4]. In this work, we develop eigenvalue solvers tailored for MDA, specifically for spectral gene clustering.
REFERENCES
[1] A. V. Knyazev. Preconditioned eigensolvers - an oxymoron? ETNA, 7, 104-123, 1998. http://etna.mcs.kent.edu/vol.7.1998/pp104-123.dir/pp104-123.pdf
[2] A. V. Knyazev. Toward the Optimal Preconditioned Eigensolver: Locally Optimal Block Preconditioned Conjugate Gradient Method. SISC, 23(2), 517-541, 2001. http://dx.doi.org/10.1137/S1064827500366124
[3] A. V. Knyazev, I. Lashuk, M. E. Argentati, and E. Ovchinnikov. Block Locally Optimal Preconditioned Eigenvalue Xolvers (BLOPEX) in hypre and PETSc. SISC 25(5), 2224-2239, 2007. http://dx.doi.org/10.1137/060661624
[4] F. Bottin, S. Leroux, A. Knyazev, and G. Zerah. “Large scale ab initio calculations based on three levels of parallelization.” Computational Material Science. In print, 2007. http://dx.doi.org/10.1016/j.commatsci.2007.07.019