The enormous number of archived formalin-fixed paraffin-embedded (FFPE) tissues available are a valuable resource of material for research. However, the use of such tissues poses many challenges, among which is the difficulty of isolating different cell populations within the tissue. In this study, we used tissue from two types of non-Hodgkin lymphoma as a model to demonstrate a method we have established and optimized to separate FFPE samples into distinct tumor and nonmalignant populations. Using FFPE reactive tonsil sections, various approaches for antigen retrieval and labeling, and the effectiveness of flow cytometric sorting were tested. We found that, among the 11 cell surface or intracellular antigen markers investigated, CD3ɛ, CD79A, LAT, PD-1, and PAX5 could be successfully labeled after antigen retrieval in Tris-EDTA buffer (pH 8.0) at 65 °C for 60 min, and 1.8-2.7 μg DNA per million cells could be extracted after sorting with DNA quality similar to that of tissue without staining or sorting. To test whether we could perform next-generation sequencing using a custom capture platform on sorted cells, we used three lymphoma cases with FFPE tissues which had been stored for 1 to 4 years. We demonstrated that the DNA from sorted cells was adequate for exon capture sequencing. By comparing the sequencing results between neoplastic and normal populations, somatic mutations could be clearly identified in the tumor population with variant frequencies as low as 11.7%.The corresponding normal fraction clearly helps in the analysis of somatic mutations and the exclusion of artifacts. This study provides an approach using flow cytometric sorting to separate different cellular populations in paraffin-embedded tissues and to unambiguously distinguish somatic mutations from germline variants or artifacts. This approach is also useful in enriching the tumor component in samples with heterogeneous components and low tumor content.