|
Abstract:
|
Nonlinearity in high-dimensional biology data and their integration are challenging problems in the post-genomic era. Inspired by the correlation integral analysis used in physics to measure fractal dimensions, we propose a nonparametric metric, maximum local correlation (MLC), to detect nonlinear association. A delta statistic based on MLC is proposed to detect critical genes and pathways by measuring association change between the control and experimental animals. Data integration is performed through extracting principal components (PCs) of genes using expression data for each pathway, and then detecting association changes among the PCs between the control and experiment. We applied our method on a microarray dataset from the control mouse and the SWV mouse, which is susceptible to neural tube defects, induced by hyperthermia. The results include promising candidate pathways.
|
