High-density DNA arrays allow measurements of gene expression levels (messe
nger RNA abundance) for thousands of genes simultaneously. We analyze array
s with spotted cDNA used in monitoring of expression profiles. A dilution s
eries of a mouse liver probe is deployed to quantify the reproducibility of
expression measurements. Saturation effects limit the accessible signal ra
nge at high intensities. Additive noise and outshining from neighboring spo
ts dominate at low intensities. For repeated measurements on the same filte
r and filter-to-filter comparisons correlation coefficients of 0.98 are fou
nd. Next we consider the clustering of gene expression time series from sti
mulated human fibroblasts which aims at finding co-regulated genes. We anal
yze how preprocessing, the distance measure, and the clustering algorithm a
ffect the resulting clusters. Finally we discuss algorithms for the identif
ication of transcription factor binding sites from clusters of co-regulated
genes. (C) 2001 American Institute of Physics.