nnNorm

nnNorm normalizes two-channel microarray data from dual-label (two-color) cDNA or long-oligonucleotide arrays to correct intensity- and spatiality-dependent systematic biases using a feed-forward neural network.

Key Features:

• Bias Correction: Corrects both intensity-dependent and spatiality-dependent biases within two-channel microarray datasets.
• Neural Network Methodology: Approximates the dependence of the log-intensity ratio (M) on average log-intensity (A) and spatial coordinates (X,Y) using a feed-forward neural network.
• Resistance to Outliers: Assigns weights to each spot based on deviation from the median M among spots with similar A and uses pseudospatial coordinates instead of direct row/column indices.

Scientific Applications:

• Gene Expression Analysis: Normalizes two-channel microarray data to enable more accurate measurement of gene expression levels from dual-label cDNA or long-oligonucleotide arrays.
• Differential Expression Studies: Reduces systematic noise to improve detection of biologically relevant differences in differential gene expression analyses.
• Cross-Slide Comparisons: Mitigates slide-to-slide biases to support reliable comparisons across multiple two-color microarray slides.

Methodology:

Models M as a function of A and spatial coordinates (X,Y) with a feed-forward neural network trained to recognize bias patterns; weights spots by deviation from the median M among spots with similar A; employs pseudospatial coordinates instead of row/column indices.