Analysis of allelic loss in archival tumor specimens is constrained by quality and quantity of tissue and by technical limitations on the number of chromosomal sites that can be efficiently evaluated in conventional analyses using polymorphic microsatellite markers. Newly developed array-based assays have the potential to yield genome-wide data from small amounts of tissue but have not been validated for use with routinely processed specimens. We used the Affymetrix HuSNP assay, composed of 1494 single nucleotide polymorphism sites, to compare allelic loss results obtained from both formalin-fixed and frozen breast tissue samples. Tumor cells were separated from normal epithelia and nonepithelial cells by dissection and bivariate cytokeratin/DNA flow sorting; normal breast cells from the same patient served as constitutive normal. Allele results from the HuSNP array averaged 96% reproducibility between duplicates and were concordant between the fixed and frozen normal samples. We also analyzed DNA from the same samples after whole-genome amplification (primer extension preamplification). Although overall signal intensities were lower, the genotype data from the primer extension preamplification material was concordant with genomic DNA data from the same samples. Results from genomic normal tissue DNA averaged informative single nucleotide polymorphism at 379 (25%) loci genome-wide. Although data points were clustered and some segments of chromosomes were not informative, our data indicated that the Affymetrix HuSNP assay could provide an efficient and valid genome-wide analysis of allelic imbalance in routinely processed and whole genome-amplified pathology specimens.