Molecular profiling of human cancer is complicated by both stromal contamination and cellular heterogeneity within samples from tumor biopsies. In this study, we developed a tissue-processing protocol using mechanical dissociation and flow cytometric sorting that resulted in the respective enrichment of stromal and tumor fractions from frozen pancreatic adenocarcinoma samples. Molecular profiling of DNA from the sorted populations using high-density single nucleotide polymorphism arrays revealed widespread chromosomal loss of heterozygosity in tumor fractions but not in either the stromal fraction or unsorted tissue specimens from the same sample. Similarly, a combination of KRAS mutations and chromosomal copy number changes at key pancreatic cancer loci, such as CDK2NA and TP53, was detected in a substantial proportion of the tumor fractions but not in matched stromal fractions from the same sample. This approach to tissue processing could greatly expand the amount of archived tissue that is available for molecular profiling of human cancer and enable a more accurate diagnosis of genetic alterations in patient samples.