Search the CONFOCAL archive at
http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal

Tami Casavan wrote:

>
> We are using a Zeiss LSM 510 META to perform linear unmixing for eCFP and
> eGFP.  When unmixing a sample with ONLY eCFP using spectra for both
> fluorochromes, the result includes a low intensity grainy textured signal for
> eGFP.  Does anyone know a way to eliminate this erroneous signal?  This
> appears using online fingerprinting as well as linear unmixing with only
> advanced linear unmixing checked for the options.
>

Linear unmixing just means that the system tries to fit (up to) 8 spectral
points to two prestored curves. If your CFP-only sample differs at all from
the stored CFP curve, the difference will be assumed to be due to GFP.
If fact, any CFP-only sample will differ from the stored curve due to
noise, and also because the fluorophore will have some spectral shift if
the environment has changed (eg, pH, temperature, fluidity, lipid vs aqueous,
etc). So if you do linear unmixing on a CFP only sample, you are almost
guaranteed to get a low intensity "GPF" image. By taking a series of such
measurements, you can assess you noise floor, which determines how big a
difference between the two fluorophores can be resolved.

I do not know the Zeiss system, but perhaps it has an option to set a
"significance" threshold for the signals. This simply means you would set
a threshold below which you do not accept the signal.

BTW, this same thing happens when fitting a curve to multiple exponentials,
If it is a monoexponential, and you fit it to two exponentials, you will get
nonzero coefficients for both exponentials (unless the data are artificial).
However, one of the two will be very small compared to the other one.

--aryeh
--
Aryeh Weiss
School of Engineering
Bar Ilan University
Ramat Gan 52900 Israel

Ph:  972-3-5317638
FAX: 972-3-5340697