Skip to content

Cell cycle

A Cell Cycle node plots the DNA histogram of the population feeding it and fits a cell-cycle model to it - Dean-Jett-Fox or Watson - reporting the fraction of cells in G1, S and G2/M.

Create one with Cell Cycle on the project panel toolbar.

Setting it up

Click the X axis and choose your DNA channel - the area parameter of your DNA stain (PI, DAPI, Hoechst, FxCycle and so on). Then pick the model in the box above the results grid: Dean-Jett-Fox unless you have a reason to use Watson (see The models). Everything else is derived from those two choices and from the node's input on every replot.

Feed the node from a gate that has already removed debris, doublets and dead cells. The fit assumes it is looking at single, intact nuclei, and it is only as good as the population you hand it.

Gate out doublets first

A doublet of two G1 cells has the DNA content of a G2 cell. Gate on DNA area against DNA width or height before the cell-cycle node, or the G2/M fraction will be overstated.

The models

A DNA histogram is three populations:

  • G1 cells carry one genome copy - a Gaussian peak.
  • G2/M cells carry two - a second Gaussian at twice the DNA content.
  • S phase cells are somewhere in between, spread over every DNA content from one copy to two.

Both models treat G1 and G2/M as Gaussian peaks. They differ in how they handle S phase. Pick one in the box above the results grid. The choice is saved with the project, and a new node starts with whichever model was last picked for the same data.

Dean-Jett-Fox

Dean and Jett described the S-phase spread as a smooth curve across the G1..G2 interval; Fox made it a second-degree polynomial broadened by the same measurement noise the two peaks carry. That is the form fitted here: an S cell sitting at a given DNA content is smeared by the instrument that widened the two peaks, and by exactly that much, with the width running linearly from the G1 width to the G2 width.

Use it for ordinary asynchronously growing cells, where S phase is a smooth plateau between the peaks.

Watson

Watson's pragmatic model makes no assumption about the shape of S phase. Each peak is fitted only on the side that S phase does not reach: G1 from three standard deviations below its mean to one above, and G2/M from one below to three above. S phase is then whatever the histogram holds between the two means once both peaks are subtracted.

Use it when S phase is not a smooth curve: synchronised cells, a drug block, or an S phase with a bump in it that a polynomial cannot follow. On simulated blocked cells (60% G1, 25% S, 15% G2/M), Watson reads 61 / 24 / 15 and Dean-Jett-Fox reads 48 / 35 / 17.

The trade-off is that S-phase cells the instrument smears onto a peak are fitted as part of that peak. A small G2/M peak with S phase running right up to it absorbs the most. On an ordinary asynchronous sample, Watson therefore reads S phase low and G2/M high. On simulated data with 55% G1, 30% S, 15% G2/M and 4% CVs, Watson reads 58 / 22 / 20 and Dean-Jett-Fox reads 55 / 30 / 15. Use Dean-Jett-Fox unless S phase is visibly not a plateau.

Watson needs each peak to span at least a few histogram bins. If a sharp histogram shows no fit under Watson, raise the bin count.

The results

The grid beside the plot reports:

Row Meaning
RMSD Root-mean-square deviation between fit and data - the goodness of fit. Under Watson it covers only the parts of the two peaks that were fitted: between the peaks, S phase matches the data by construction
%G1, %S, %G2 The three fractions
G1 Mean, G2 Mean Peak positions, in the parameter's own units
G1 CV, G2 CV Peak coefficients of variation
% less G1 Events below the G1 peak - sub-G1, typically apoptotic or debris
% greater G2 Events above the G2 peak - polyploid, aggregates or doublets that got through

Reading the numbers

  • G1 CV is the headline quality metric for the staining and the instrument. Under 5% is good; much above that and the G1 and G2 peaks start to overlap the S phase, and the three fractions become model-dependent rather than measured.
  • RMSD tells you whether the model fits at all. A large value with clean peaks usually means something other than cycling cells is present - apoptotic debris, or two cell populations with different ploidy.
  • % less G1 and % greater G2 are the model's honest account of what it could not place. Large values there mean the input needs better gating.

Bin count

The histogram bin count comes from the global setting on the Options page. More bins resolve the peaks better but make the fit noisier; the default is a reasonable compromise for the usual few tens of thousands of events.

Multiple conditions

Make one cell-cycle node per condition and feed each from its well gate. Put them in a row on the canvas and the effect of a treatment on the three fractions is immediately visible, with all the fits computed the same way.