Cryopreservation of patient bone marrow specimens allows for future purification of tumor cells for research or diagnostic purposes. It has not been determined whether this practice causes significant changes in gene expression. In order to evaluate this, we performed microarray analysis (Affymetrix U133A and U133A 2.0 gene chips) on mononuclear cells from 5 acute myelogenous leukemia (AML) patient samples (>70% blast count) that had been either a) cryopreserved, b) snap frozen, or c) resuspended in TRIzol and stored at −80°C. Samples remained frozen for a period of 1–4 weeks. While gene expression changes between groups were minor, there were some differences. Thirteen probe sets between TRIzol and cryopreserved samples, and 60 probe sets between TRIzol and snap frozen samples, varied significantly (>2-fold difference and p<0.01; t-test). While both non-TRIzol samples lost transcripts associated with mature, contaminating granulocytes, the snap frozen samples also lost transcripts associated with cell growth and metabolism. To determine whether the observed changes were enough to cause a misclassification of AML subtypes, we performed an unsupervised cluster analysis on the 15 original samples plus 15 new AML samples (13 unique cases), all of varying subtypes. RNA from all of these samples was undegraded. All samples passed strict quality control parameters, not limited to but including tumor content, % rRNA, cDNA and cRNA synthesis product size, and GAPDH 3′-5′ ratios. Differentially preserved samples originating from the same patient segregated together during the clustering process regardless of preservation method. Three main clusters emerged; cluster #1 (FAB-M0 and -M1, n=3), cluster #2 (FAB-M3, n=5), and cluster #3 (FAB-M2, -M4 and -M5b, n=22). In cluster #3, there were three sub-groupings. One group contained only M4 and M5b subtypes, another contained only the M2 subtype, and the third contained both M2 and M4 subtypes. RNA isolated from snap frozen samples was sometimes moderately to severely degraded. When we examined three snap frozen samples, not part of the above data set, exhibiting moderate RNA degradation, they all clustered incorrectly according to the above established groups. Given these results, we conclude that cryopreservation is an acceptable method of cell preservation for gene expression analysis, but snap frozen samples should be closely evaluated for RNA degradation before using in microarray analyses.