This page contains suggestions for improving the fit of difficult to model cytometric data.
Preprocessing
Removing doublets, dead cells, and debris will improve the condition of data. Whatever steps you typically follow to do this can be applied to cell cycle data. Some common methods for doing this are:
- Doublets can be removed by plotting FSC-A versus FSC-H and gating on the cells along the most dense diagonal. These are cells with a consistent ratio between two measures of the forward scatter pulse. Dots that are not along this diagonal have a different ratio, which is likely to be caused by agglomerates.
- Dead cells can be removed using a viability dye. There are now viability dyes available that can be used with cell cycle assays.
- Debris can be removed by making a series of 2D gates on all combinations of parameters and removing the cells along the axes. It will be especially important to do this for the parameter you use to measure DNA content.
Constraining
Constraints fix certain aspects of the model, forcing the model to be exactly the same from file to file, if you apply a model to a group of samples. So the more parameters you leave unconstrained, the more freedom you give the software to create accurate models of other data files when applied to a group of files. Alternatively, you may wish to analyze a control sample that has a good distribution, and set constraining ranges for the peak(s) based on that sample. When you copy the Cell Cycle analysis to the remaining sample, the ranges will automatically be applied. This can be especially useful in which the experimental conditions result in data in which the peaks are either difficult or impossible to identify.
Constraining tips
- Start with an unconstrained model on a control sample, or some other sample you expect to follow the classic cell cycle distribution. Apply this model to the group, and use batching in the layout editor to create images of all the files to see how well unconstrained works for your experiment, and which files are troublesome.
- Always begin by setting the CV’s to be equal. This should almost always be true, is an easy constraint to explain in the methods section of a publication, and leaves the model with plenty of flexibility to fit data.
- If you need further constraints, try constraining the mean of the G1 and G2 peaks relative to each other. Within a given experiment this should be consistent, and this particular constraint still allows the model the flexibility of shifting the peaks to fit the data.
- If you need to constrain the location of a peak, choose to constrain it to the widest possible region. This preserves flexibility, while eliminating gross mistakes such as setting G1 to be a small apoptotic peak.
Experimental design tips
- If apoptosis is an important aspect of your analysis, use a marker for apoptosis such as Annexin V. It is easy to show that many of the sub-G1 cells are debris and not apoptotic, and that many of the apparent G0/G1 cells are in early apoptosis, and thus using DNA content alone as a judge of apoptosis gives inaccurate results.
- If further immunophenotyping of S-phase cells is important to your study, consider adding a second parameter that can be used to identify S-phase cells to your assay. You can then switch to our bivariate model, and will have the ability to gate much more accurately on S-phase cells. The univariate model can give you a reasonable estimate of the number of S-phase cells, but has trouble picking out specifically which cells are in S-phase.
For information on the univariate algorithm used in the software, please see:
Watson, Chambers, & Smith: A Pragmatic Approach to the Analysis of DNA Histograms with a Definable G1 Peak [Cytometry 8:1-8 (1987)].
For more information see the related links listed below: