Re: Data Analysis Software

From: Mark A. KuKuruga (kukuru@umich.edu)
Date: Wed Mar 17 1999 - 14:07:49 EST


Ray Hicks wrote:

> >MAK wrote: 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 . . .

> Ray Hicks wrote:
> By which model are "cells defined as falling within regions by the DNA fitting
> algorithm"?  I wasn't knocking modelling (nor the model-makers), but was pointing
> out that it doesn't allow you to identify which cell, in an area of overlap,
> belongs to which compartment.

> The reason for deconvolution is to create a perfect world where those assignments
> can be made.  The conspiring spread functions are modelled  and how well the
> model's output matches the real histogram is a measure of perfection.  But the
> nature of modelling still doesn't allow you to identify a cell.

I think we're saying the same thing in different ways . . . by mathematically
defining compartments in the DNA distribution, one can indicate (within the
constraints of the accuracy of "deconvolution") the "probability" of a given number
of cells falling within a specific compartment.  Obviously, one cannot use these
distributions to gate on the compartments, except that one can choose to gate cells
according to where they are likely to be statistically.  This may therefore only
allow you to be confident in defining "only" S-phase cells, limited by , say, 3
standard deviations above the mean of G1 and the same below G2M.  So, in the
context of sorting this can be useful, where these gates now serve as sort
regions.At the same time, the data in discussion here already approaches a
definition of these compartments, given that BrdU incorporation defines S-phase.  I
think the original problem lay in the confusion arising from the observation that
cells with apparent G1 DNA content were incorporating BrdU.  Use of modeling cannot
help this . . . which was the point of my second comment.
Summary:  Use of DNA histogram fitting algorithms can define the probability that a
specific region is of a specific compartment.  This can be used to select regions
that have a higher probability of being in these compartments.  Although these
regions for selection may not necessarily select "pure" populations (it's pretty
good for a limited region of S-phase, but there are overlays in G1 and G2M of
probable S- phase cells), they can certainly define selection limits to improve
your chances of  "sorting" pure populations.  This is used for physical selection
of G1, S and G2M cells.
It is also important to stress the appropriate uses of a given set of data, with an
understanding of its limitations.  Fitting of a DNA histogram will provide
probability statistics that may indicate relative changes in cell cycle
progression.  If one requires more exacting cycle compartment elucidation, use of
alternative methods may be necessary (AO measurements, cyclin expression
measurements . . .)
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