Mark, Autofluorescence compensation was published back in the mid-1980's, doing it both by software (Roederer & Murphy, Cytometry) and by hardware (Alberti, Parks & Herzenberg, Cytometry). Basically, you just treat one of your fluorescence channels (for which there is no stain, for example, FL2 in the absence of a PE stain) as the channel that measures autofluorescence. Then you just do standard compensation based on removing the spillover between this channel and the others. There is no need to collect data at different voltages (but if you do, make sure to collect the "compensation controls" (including the unstained autofluorescence control) at each different voltage and adjust comp settings for each voltages separately). For Macintosh, FlowJo can perform this compensation on your data, assuming you have collected all of the appropriate compensation controls already. For PC, a variety of programs will let you do this. I don't think you should use a generic "algebraic" approach, because what you are doing is compensation, NOT background subtraction. Therefore, you need to use the linear algebra mathematics that are built into the compensation algorithms of these programs. mr >I have a separate question: "After-the-fact compensation" > > We have numerous FCS files (generated on a Coulter Elite using Expo >software but typically analyzed using CellQuest). We would like to try >different kinds of compensation, after the experiment is over. The object >is primarily to remove autofluorescence in order to improve the signal to >noise ratio of our samples. In other words we would like to run "induced" >and uninduced" samples at various voltage settings for the different PMTs >and, at a later point, try correcting for autofluorescence by subtracting a >fraction of one channel from another channel. > Ideally this would be a program which inputs one or more FCS files >together with an algebraic expression representing the operation to be done >on the files and outputs FCS files containing the altered parameter (or >maybe a new parameter representing the transformed data). The only trick >might be that log data would have to be converted to a linear signal before >subtraction and then converted back to log data. > >Mark Poritz, Ph.D. >Senior Scientist >Arcaris Inc. >615 Arapeen Dr., Suite 300 >Salt Lake City, UT 84108 >801-303-0300, 0333 fax >(formerly Ventana Genetics)
This archive was generated by hypermail 2b29 : Sat Mar 10 2001 - 19:31:02 EST