Ray Hicks wrote: > At 11:22 am -0500 16/3/99, Mark A. KuKuruga wrote: > >Jill Martin wrote: > > > >> . . . I have an investigator who is looking at proliferation and would like > >> to see the uptake of Brdu in S-phase. We would like to do an accurate > >>cell > >> cycle analysis which will give us not only the percentage of each > >>phase but > >> the exact boundaries so that we can gate on them. > > > >Cell cycle fitting programs, like ModFit (verity) or MultiCycle (Phoenix) will > >fit your DNA histogram data and effectively define limits of your cycle > >compartments. These limits can then be transferred as gating criteria. > > > > Hi Jill, > > be aware that the fitted one-dimensional DNA distributions won't let you assign > any particular cell to its compartment. The reason for deconvolving the histogram > is indeed to "deblur" those boundaries, but it's the histogram that gets > deblurred not the cell data, and you end up with probability information - you > may find out how likely cells are to be in each compartment, but not which ones > are. The probability information is the point of my comment. I use this "probability" in sorting, where one can determine the most probable regions for sorting G1, S, and G2M. In this context, cells defined as falling within these regions by the DNA fitting algorithm are certainly specific to that compartment . . . unless you suggest there is no real relationship between DNA content and DNA-fluorochrome signal intensity . . . > You'll notice that the results of modelling have overlapping distributions; some > of the S-phase generally ends up in the G0/G1 and G2/M "areas", and vice versa - > the closer the output is to the input, the better the fit (generally). It's not the fault of the modeling, yet the nature of the measurement that forces these apparent overlaps. In a perfect world, perfect resolution may result in the ability to define a single channel for G1, a specific channel for transition to S-phase, channels that designate the interface of G2 and Mitosis . . . but, our flow cytometers, labeling techniques, operators, cells, dyes, all conspire to dither these measurements so that what we actually measure requires accommodation through use of this broadened S-phase polynomial function. Logic predicts that there will be S-phase cells having DNA content close to G1, and at the other end, close to G2 and Mitosis (keeping in mind here that we're actually measuring PI content, of course). The "overlap" allows for this. > You might be better off gating on the bivariate plot and forgetting the > deconvolution, at least you're just making two assumptions (that all of the > S-phase cells are labelled, and none of the G phase cells) rather than the wide > range of assumptions used in deconvolving and modelling. I concur here . . . as I commented, one can take the S-phase measurements directly from the bivariate DNA-BrdU histogram, perhaps more accurately than from the DNA histogram fit. This is not always practical, however, considering that treatments to compartmentalize via this technique will often preclude other goals, such as viable cell sorting based on cell cycle position. So, the histogram fit serves as the best approach in many applications, regardless of the necessary assumptions.Besides, the historical record will show the extensive effort applied by the inventors/developers of these models (Dean, Rabinovich, Bagley, Watson, others) towards their validation. MAK. -- Mark A. KuKuruga, Managing Director University of Michigan Core Flow Cytometry kukuruga@medmail.med.umich.edu
This archive was generated by hypermail 2b29 : Wed Apr 03 2002 - 11:53:15 EST