Increasing the number of measurable markers in flow cytometry (FACS) experiments can reveal complexity of cell populations that cannot be detected with fewer markers. Available software for analyzing FACS data, however, operates almost exclusively on 1- or 2-dimensional views, often leading to subjective and sub-optimal definitions of cell populations and even failure to find all populations. This is a critical problem in what we call Hi-D FACS where more that the usual 3-5 colors are used. Appropriate instrumentation and reagents are becoming more available, but the lack of objective procedures for direct multi-dimensional analysis is a factor limiting realization of the potential power of multi-color FACS measurements. We propose to develop techniques for automated and objective identification and delineation of cell populations in Hi-D FACS data (and also in more conventional 2-3 color work). The method is based on newly developed ideas in applying a scale-space approach with kernel density estimation and will operate directly in the multi-dimensional context. We have successfully carried out an exploratory project showing the feasibility of using scale-space methods for automated data analysis and population identification in multi-color FACS data. The method was implemented in a two-dimensional version, which was designed to be extended to higher dimensions. Tests on various data sets showed that the software consistently identified cell populations correctly and selected appropriate boundaries for them. These results make us confident that embarking on a serious development project is justified and likely to yield valuable new techniques for improved analysis of multi-color FACS data. These techniques should also provide automated and more consistent analysis of large-volume FACS data. We plan to implement the extension to higher dimensions, develop visualization techniques for multidimensional results and develop ways to sort the defined populations. Cooperation with a very active project on B cell development in murine bone marrow will provide extensive data (up to 11-colors and 13-parameters) and expertise for testing and evaluating the new techniques. When the techniques are working well, we will transform them into a form that can be made available to FACS users generally.
| Herzenberg, Leonard A; Herzenberg, Leonore A (2004) Genetics, FACS, immunology, and redox: a tale of two lives intertwined. Annu Rev Immunol 22:1-31 |
| Lu, Li-Sheng; Tung, James; Baumgarth, Nicole et al. (2002) Identification of a germ-line pro-B cell subset that distinguishes the fetal/neonatal from the adult B cell development pathway. Proc Natl Acad Sci U S A 99:3007-12 |
