Re: 2 machines, 2 compensation troubles.

Uriel Trahtemberg wrote-

>
> As usual, when confronted with trouble I turn to you for help. this  
> concerns two machines with similar problems.
>
> 1) The FACScalibur in our lab. I performed a titration experiment  
> for Sytox Green and when checking the compensation needed for it to  
> work with other stains, saw the results shown in the pdf "SytoxG  
> linearity". I have attached one of the files (ApoPMN.NucAcTit. 
> 1mar07.021) if you want to check it out. As you can see, instead of  
> being a nice distribution on the X axis, it looks like a waving  
> serpent! On close inspection, the waviness can already be seen in  
> the uncompensated FL1 vs. FL2 plot. After long thought, I have  
> arrived to the conclusion this is due to faulty log-amp behavior  
> causing deviations from linearity. Certainly, this has important  
> consequences for our results on that machine! In order to confirm  
> my findings I exported the FCS data to excel, sorted the data  
> according to the FL1 intensity and did a FL1/FL2 plot. After	
> accounting for statistical noise, the mean ratio should stay	
> constant - after all, that is the rationale of compensation! As you  
> can see from the "SytoxGreen.xls" plots, that is hardly the case.
> I turn to you to ask for alternative explanations and possible  
> solutions. Is this kind of problem repairable? how easily? We have  
> a service contract with BD for that machine, so would I be right to  
> assume they should fix it? What kind of tests would you suggest to  
> confirm and/or refine my findings?
> PS: SytoxGreen is NOT metachromatic AFAIK.

I'm feeling lazy so I'll only comment on the first "problem", in  
which the reasons for replacing log amps in newer cytometer designs  
become apparent.

See p. 202 of Practical Flow Cytometry and the middle panel of Figure  
4-57 (you can also go to www.analog.com and download AD8307.pdf,  
which is a data sheet on one of the newer log amp modules). A log amp  
is not a log table. The log amps used in most flow cytometers are  
made up of multiple modules and exhibit linearity curves that are  
wavy even in the range in which they perform best. The typical	
deviation from peak to trough over that range is plus or minus 0.5  
dB. Do the math and you will find that is plus or minus 6 per cent,  
if you're talking voltage. Plus or minus 6 per cent is fine if you're  
measuring immunofluorescence, where other biologic and instrumental  
factors typically contribute much more to variance than does the  
inherent, built-in, not-correctable-by-replacing-parts inaccuracy of  
the log amp. If you measure nucleic acid stains, which are  
stoichiometric, with a log amp, you pretty much get a picture of the  
log amp's nonlinearity curve. The solution to the problem is to  
measure such stains using linear electronics. Most flow cytometer  
manufacturers have now opted to build high dynamic range linear  
electronics that typically allow accurate measurement over a four- 
decade range; the resulting data are digitized to enough bits'	
precision to allow conversion between linear and log scales to be  
done digitally. It is, however, worth noting that Cytomation/Dako and  
Cytopeia cytometers do use log amps, converting the log data with  
high-resolution ADCs to allow digital linear-log conversion for  
compensation purposes, and the users of these cytometers are not, in  
general, bothered by the inaccuracy - unless, of course, they try to  
do something like measure DNA on a log scale.

-Howard