Glossary
APS1: Automated Population Separator 1; bidimensional dot plot representing the first versus
Together with the microscope, flow cytometers have become invaluable instruments for cell analysis both in research and clinical laboratories, particularly for the evaluation of normal and malignant hematopoietic and lymphoid cells [1]. Since the 1980s, FC has progressively extended from basic research to clinical diagnostic laboratories. Development of the hybridoma technology and massive production of monoclonal antibodies, together with many well-characterized, high-quality reagents and a progressively broad variety of compatible fluorochromes [1], have been crucial in the translation of FC to clinical diagnostic laboratories. In the past two decades, multiparameter FC immunophenotyping of normal, reactive, and malignant cells has been augmented by: (i) a larger number of parameters that can be simultaneously assessed in individual cells [2], and (ii) greater analysis speed of digital cytometers [3] that interrogate tens of thousands of cells per second. However, the exponentially growing, complex information about individual cells (and their products) in a sample, has required changes and developments in both data analysis, representation and visualization tools [4].
For decades, conventional FC data analysis has identified unique cell populations within a sample and characterized them by defining a number of regions (e.g., ‘gates’) in single and bidimensional plots. Such gating strategies are typically set for the selection of the subpopulations of interest by an experienced operator [5]. However, due to the greater multiparameter capabilities of FCs, strategies to extract maximum information from an n-dimensional space and to provide clear, simple and intuitive graphical representation of data, have become a major challenge [4]. In fact, the real challenge is to build FC data mining tools that extract relevant information in an objective, precise, reproducible, and comprehensive way. Moreover, it should be available to users through intuitive graphical representations and user-friendly interpretation-guided tools.
FC data analysis can be roughly divided into three steps: (i) generating FC data (Box 1); (ii) data storage; and (iii) mining stored information. In this paper we review the most relevant developments achieved in FC data analysis, and provide a perspective on current needs and potential future developments in the field.
Early flow cytometers and cell counters were developed before computers were integrated into analytical instruments 4, 6, 7, 8, 9. At that time, screens capable of representing analog signals in real time were used for fast analysis of results associated with individual cell measurements [10]. Progressive development of computer technology rapidly led to its incorporation in FC. Computers were used to control the instrument and to store data derived from the cellular measurements performed with
Hematological and lymphoid tissues such as peripheral blood (PB), bone marrow, lymph node, thymus, and other specimens, are composed of different cell populations defined by multiple cell lineages, maturation stages, functional compartments, and activation states, mixed at variable percentages 1, 17, 18, 19, 20, 21, 22, 23, 24. Individual cell populations can be identified by binding high affinity antibodies to protein structures or by unique cell characteristics 24, 25, 26, 27. Although
A common step in FC data analysis is the establishment of an association between each individual cellular event and specific populations of events with biological meaning. For pattern classification, single events, modeled as elements of an n-dimensional space, are linked to a class corresponding to a population of cells of interest. This is based on the principle that cells that belong to the same population will most probably show highly overlapping (e.g., similar) immunophenotypic features.
Adequate visualization of FC data strongly contributes to optimal data analysis and interpretation. Common tools currently used for FC data visualization (Figure 2) include single-parameter histograms, bivariate plots, and 3D representations of primary FC parameters – for example, light scatter or fluorescence emissions – (Figure 2A–C, respectively). Such plots (except for 3D views) are used to select (e.g., gate) groups of events and identify cell populations of interest, in an expert-based
FC has significantly advanced with the generation of increasingly complex n-dimensional data sets. New tools and procedures are devoted to data manipulation, data analysis, visualization, and interpretation. Nevertheless, further developments are still needed (Box 5), including: (i) increased reproducibility of FC measurements across different reagent and instrument platforms; (ii) optimized computational processing times; (iii) freely available reference data files and profiles for extended
Cytognos S.L. is part of the UE-supported EuroFlow Research Consortium, and has implemented some of the algorithms described in the present study, in its proprietary software INFINICYT; Cytognos S.L. has a contract license of several patents owned by the University of Salamanca, of which A. Orfao, C.E. Pedreira, and E.S. Costa are inventors. Other authors declare no competing financial interests. Glossary APS1: Automated Population Separator 1; bidimensional dot plot representing the first versus
This work has been partially supported by the following grants: Spanish Network of Cancer Research Centers (ISCIII RTICC- RD12/0036/0048-FEDER), FIS 08/90881-FEDER from the ‘Fondo de Investigación Sanitaria’, Ministerio de Ciencia e Innovación (Madrid, Spain); Bilateral International Cooperation (Spain-Brazil), Ministerio de Ciencia e Innovación (Madrid, Spain) and CNPq-Brazilian National Research Council (Brasília, Brazil), Refs. PIB2010BZ-00565 and 560757/2010-7; the Euroflow Consortium