Ratio analysis of cumulatives for labeled cell quantification from immunofluorescence histograms derived from cells...

The problem of the quantitative analysis of immunofluorescence histograms "with weak labeling" is considered. In these distributions, the fluorescence intensity mean of labeled cells is slightly above that of unlabeled cells, and therefore there is a considerable overlap between the two populations. The procedure of selecting "by eye" an intensity level that separates labeled from unlabeled cells is evidently inadequate, leading to subjective and not reproducible results. Threshold and subtraction methods, although objective, quick, and easy to use, are also inaccurate. Till now, correct enough estimates of the percentage of labeled cells have been obtained only by applying mathematical modeling methods, that are, however, not very easy to handle. A new method for estimating the labeled cell proportion from immunofluorescence histograms is presented. It relies on the analysis of the ratio of the cumulative distributions associated with the routinely available test and negative control histograms. Since this ratio results in a well-defined "sigma-shaped" distribution from a certain channel onwards, a suitable fitting function is used to obtain directly a preliminary estimate of the proportion of labeled cells. Then, an iterative procedure is designed for updating the estimate until it is judged acceptable, according to a suitable criterion. Moreover, to take into account the overall variability in the control data, the estimation process is repeated using the cumulatives associated with replicate negative control histograms. The accuracy of the method was tested with the same set of experimental data analyzed by means of a mathematical modeling method described in a previous article (Lampariello and Aiello, Cytometry 1998;32:241-254). There was a very good agreement between the known and estimated proportions of labeled cells, and the latter were, in many cases, even closer to the true values than those previously obtained. On the basis of the results reported, the method, which is objective, very simple, and much easier to implement than our previously proposed one, appears to be effective in the entire range of the percentage values, and to provide results independent of the acquisition precision.