Impact of pixel intensity correlations on statistical inferences of expression levels in cDNA microarray experiments

In a cDNA microarray experiment, the final measurement is intensity ratio at a spot in the microarray chip. The objective of the present study is to estimate the uncertainty associated with the final intensity ratio at each spot in cDNA microarray chips and also to explore the role of pixel intensity correlations in statistical inferences of gene expression levels. We estimate uncertainty at each spot using the theory of error propagation under two different situations: (1) when there is no correlation between pixel intensities and (2) when the pixel intensities are positively correlated. The inverses of these estimated uncertainties are used as weights in downstream analysis to test the significance of each gene. The analysis was verified on a data downloaded from the GEO database. Our study shows that the uncertainty and statistical inference of gene expression levels depend on correlation between pixel intensities within a spot.