Re: histogram stats, which nos best?

From: David Novo (dnovo@ucla.edu)
Date: Wed Oct 25 2000 - 17:59:26 EST


Hi Roy,

It depends on what type of data you want to show. If you are correlating
your fluorescence level to number of antibodies using one of the common
methods you could just report the number of molecules on the surface. Of
course, there are caveats to doing the quantitation properly i.e. ensuring
the F:P ratio on your antibodies is exactly known, etc.

The easiest thing to do would be to express your data as a percentage
increase. That would be easy to do comparing the means or medians of the
control and treated cell in every type of cell. Including the CV's of
course to let people get a sense of how wide your distribution was.

There were some recent articles on KS statistics in cytometry and
elsewhere. They can be easily found on medline. There are certain caveats
to those as well, including how to bin the data and what to use as the
threshold for significance. With 1024 degrees of freedom two samples from
the same tube will probably be statistically different just due to the noise.

If you just want percent positive and the populations are closely
overlapping you may want to use an overton subtraction type method
(Overton, WR.  Modified histogram subtraction technique for analysis of
flow cytometry data.  Cytometry, 1988 Nov, 9(6):619-26.)

If you are going to be doing a lot of this type of analysis, the software
package FCS Express (www.denovosoftware.com) will automatically do
quantitation as well as Overton subtraction.

-Dave


At 08:03 AM 10/24/00 -0700, Meenakshi Roy wrote:


>Hi everybody,
>     My question is probably related to the isotype control/autoflourescence
>question.  I am comparing several different cell types, for expression of
>of the same marker X.  The catch is that the intensity of staining of the
>same isotype control is different in each cell type.
>   The different cell types clearly show different intensities of staining
>with marker X, and this would correlate with what I see functionally.
>However, as I said before the isotype controls also increase somewhat,
>though not as strongly in intensity.
>   What is the best way to compute numbers for this case?
>1.  Show median intensity for the total populations, without subtracting
>the isotype?
>2.  Use K-S stats to compare marker X and isotype control on the same cell
>type?  Would someone please enlighten me on this one? Which figure is
>relevant?  D/S(n) or the channel figures?
>3.  Cell quest has an option whereby I can get stats using the marker X as
>a numerator and the isotype control as a denominator?  Is this an option?
>
>   Forgive my ignorance, and any suggestions/explanations would be much
>appreciated.
>
>Meenakshi Roy



This archive was generated by hypermail 2b29 : Wed Apr 03 2002 - 11:56:10 EST