[1218] Classification of Low-Grade B-Cell Lymphomas Using Hierarchical Clustering of Raw Flow Cytometry Data.

Anna R Balog, Howard J Meyerson. University Hospitals Case Medical Center, Cleveland, OH

Background: Analysis of data by flow cytometry is currently subjective. An objective method to classify lymphomas using flow cytometric data would be advantageous. Hierarchical clustering is a useful means to achieve this goal. The use of raw data is beneficial since it would allow for automated data extraction and, theoretically, disease classification. However, variations in the performance of the instrument over time could generate varied fluorescent values interfering with ability to successfully perform an objective analysis. Therefore, we assessed the ability of unsupervised hierarchical clustering to classify a series of low-grade B-cell lymphomas using raw flow cytometric data extracted from routine analyses performed over a 1 year period in the clinical flow cytometry laboratory at University Hospitals Case Medical Center.
Design: Flow cytometry data from 34 patients diagnosed with low-grade B-cell lymphomas between 1/2008 and 12/2008 were studied. The lymphoma cases included 22 chronic lymphocytic leukemias (CLLs), 6 marginal zone lymphomas (MZLs), and 6 mantle cell lymphomas (MCLs). Flow cytometry was performed on either a peripheral blood or bone marrow sample using the same antibody panel. The raw mean flow cytometric fluorescent values for 24 antigens and forward and side angle light scatter were extracted after initial selection of the tumor cell populations. Antigen expression on tumor cells was compared to that of B cells from three normal peripheral blood specimens analyzed in the same time period, and the data was log transformed prior to clustering. Hierarchical clustering was then performed on Cluster 3.0 software using complete linkage and Pearson's correlate. Data was analyzed and visualized using Java Treeview.
Results: Hierarchical cluster analysis segregated the CLL, MCL and MZL cases successfully. The within cluster correlation coefficient of the CLL cases was 0.59. Additionally, within the CLL cluster, three distinct subgroups were identified. The within cluster correlation coefficient of the MCL cases was 0.75. Of the 6 MZLs, 3 cases co-clustered with a correlation coefficient of 0.72, but the remaining 3 cases segregated independently.
Conclusions: This study suggests an objective approach using raw flow cytometric data is a feasible method for classification of low-grade B-cell lymphomas into major categories from specimens analyzed serially over time. Although only a limited number of MZL cases were analyzed, the lack of a distinct cluster within this group suggests MZLs may be heterogeneous.
Category: Hematopathology

Tuesday, March 1, 2011 1:00 PM

Poster Session IV # 172, Tuesday Afternoon


Close Window