A comparison of mathematical methods for the analysis of DNA histograms obtained by flow cytometry

Abstract. Twelve methods for analysing FCM‐histograms were compared using the same set of data. Some of the histograms that were analysed were simulated by computer and some were taken from experiments. Simulated data were generated assuming asynchronously growing cell populations and (i) measurement coefficients of variation (CV) from 2 to 16%; (ii) constant measurement CV or CV's increasing from G₁ to G₂ phase, and (iii) varying fractions of cells in each phase. Simulated data were also generated assuming synchronous cell populations in which a block in early S phase was applied and released. DNA histograms were measured for L‐929 cells at various times after mitotic selection. Labelling indices were also measured for these cells at the same time. The fractions of cells in the G₁, S, and (G₂+ M) phases were calculated by each analytical method and compared with the actual fractions used for simulation, or in case of experimental data, with autoradiographic results. Generally, all methods yielded reasonably accurate fractions of cells in each phase with relative errors in the range of 10–20%. However, most methods tended to overestimate G₁ fractions and underestimate S fractions. In addition, variations in the shape of the S phase distribution often caused considerable errors. Phase fractions were also calculated for histograms of kinetically perturbed populations, simulated as well as experimental The errors were only slightly larger than for histograms from asynchronously growing cell populations.