leave the theoretical, and that is the difference between spillover and compensation coefficients. As we mentioned earlier, they are very close, but not identical in a two-color problem. As the number of fluorescence parameters increases they further diverge. For example, when three f1uorochromes are used, the total signal detected in channel A results from the true signal from fluorochrome A, plus the spillover of fluorochrome B into A, plus the spillover of C into A. We wish to correct (compensate), such that the effects of spillover are eliminated and only the true signal from fluorochrome A is measured. But the proportion of the signal in B to be subtracted from A is not correctly described by the spillover 204 Handbook of Human Immunology, Seconu Editiun coefficient of B into A. because some of the signal detected in channel B i~ due to spillover of A into Band C into B. Fortunately. this type of a problem i~ addre~sed by the field of linear algebra and the true compensation coefficients can be described by a series of simultaneous equations and solved with matrix algebra. AII modern automated compensation software use this algorithm, sometimes referred to as the l/1atrix inversion method [581. Now that we have seen why color compensation is necessary, we can exam ine practical ways in which it can be implemented and assessed in simple and complex applications. The CD34 test used in our clinical laboratory (StemKit®, Beckman Coulter) is nominally a four-color assay, but one fluorescence channel is used to identify viable cells by 7-AAD exclusion, and one channel is used to detect StemCount® beads as a reference for absolute cell counting. Thus, we have a two-color compensation problem (anti-CD45 FITC versus anti-CD34 PE). Because these are single dyes for which the emission spectra are constant, the amount of color compensation required should never change, provided the PMT gains are consistent from day to day (this is ensured by validating the PMT target channels). However, accrediting bodies such as the CAP require color compensation to be validated on a regular basis. For the CD34 assay, we validate our compensation settings weekly using unstained, FITC and PE Calibrite® beads. We use the same fluorescence PMT settings to be used in our assay (but different forward and side scatter, adjusted for bead detection). These settings are adjusted daily against FlowSet® beads. In the clinical laboratory, we are still using a very mature software package (System 1I®, Beckman Coulter) that includes a semiautomated compensation algorithm, which mimics old-fashioned manual compensation by matching the fluorescence intensity of the negative population with that of the stained beads (e.g., compensation is increased until the intensity of the FITC standard in the PE channel matches that of the unstained beads). We have not found it necessary to change the compensation values over a period of years, providing an additional independent validation of the PMT settings. Real-time data analysis is performed during data acquisition using the validated compensation values; compensated data are saved to list-mode files containing the original analysis protocol, in case reanalysis is necessary. For complex multicolor analyses, real-time color compensation and data analysis may not always be feasible or desirable. Further, the classical technique of matching median fluorescence intensities described earlier is not the most efficient or the most accurate. All modern flow cytometry analysis software packages support offline compensation. Compensation is computed from measurements of spillover or slopes of single-stained samples, using the sweet range of the cytometer rather than the troublesome first decade. The advisability of performing a fully automated compensation on multicolor data cannot be over emphasized. As detailed previously, PMT setting validation, linearity, and sensitivity testing are done daily, and single-stained beads (directly labeled or Ig capture) are acquired for each run. It is critical that the single-stained control beads be acquired at the same fluorescence PMT settings as the actual samples, and for tandem dyes, that the single-stained controls be performed with the actual