>Dear Flow Cytometry list reader, > >as a disclaimer, I'm quite new to the FCS techniques, >and hoping not to upset anyone with my question. I >looked through the archive but couldn't find a answer. > >I just discovered the the software package "Summit" >form Cytomation, which is "free" to use as a offline >analysing tool. I was surprised to read the following >statement in the online tutorial: > >" By using the Summit software, there is no need to >adjust compensation on the >cytometer before acquiring data. Summit provides the >ability to adjust compensation after the >raw data has been acquired and saved as a file" > >What is your opinion to this statement ?. Is this true >?. Yes, in general. There is no need to compensate data in real-time unless you doing sorting. However, compared to software compensation there are some subtle artifacts which occur in analog real-time compensation (mainly occurring on signals whose values are close to 0) which make it "look" better than software compensation, but in some cases can give misleading results. >FlowJo, FCSPress also give the possibility to >"software" compensate samples. Will I get the same >result when I do the compensation with these packages, >or will the result depend on the underlying >algorithmen ?. You should get the same result if each package uses the same model to compensate the data. The standard model is a linear one - the uncompensated and compensated data are assumed to be in two vector spaces which can be mapped to each other by an invertible linear transformation, i.e. a matrix. The coefficients of the matrix are determined from single color controls. The mapping from the raw data to the compensated data is called the compensation matrix, and its inverse is called the spillover matrix. From the single color controls the coefficients of the spillover matrix are determined, and it is inverted to do compensation. There are several items to consider in this general scheme. First is that the raw data, with the exception of the new fully digital systems, is usually collected with logarithmic amplification. The scaling of these log amps is usually not know precisely, so the scaling of the data typically could be off by up to 5%, leading to inaccuracies. Second, in some systems one works with the spillover values, which are easy and intuitive, and some systems work with the inverted compensation matrix values, which you think are intuitive but are not. FlowJo is a system which works with the spillovers. That is, you give it the controls and it calculates the compensation matrix. Summit, I believe, as well as most systems of analog compensation, you work directly with the coefficients of the compensation matrix, e.g you select N, where N is the coefficient in the equation FITC-comp = FITC-raw - N*PE-raw This works well for compensating pairs of colors, but leads to inaccuracies when doing large numbers of colors for two reasons. First, you can only do pair-wise subtraction. Second, there are multiple interactions between colors which are not intuitive, so, for example, compensating FITC into PE and PE into PE-TR may lead to over-compensating FITC into PE-TR and you have no way of backing that out. This can only be overcome by working with the spillover values and inverting the matrix. Lastly, of course, this assumes that Summit (and FCS-Express) do use the linear model correctly. In an older version of Summit I used, overcompensation resulted in data "bouncing" off the axis instead of being smashed down on the axis. One should note that the difference between spillover and compensation matrix values is confused by many, due to the long history of working with compensation coefficients in analog as described above. For example, on the new BD DiVa software, the user enters the spillover values for every parameter pair, but the interface in which it is entered looks like: FLx - N*FLy That is, it gives the impression that one is entering compensation values, which is not the case. A clearer and more detailed discussion of this has been posted by Mario Roederer at http://www.drmr.com/compensation/index.html. There are also some articles in Cytometry on the subject. Marty > >I appreciate if anyone can comment on my questions. >I'm also interested to know if someone knows review >articles on compensation/how to do (software and >hardware). > >Many many thanks > >Frank Mattes >PhD student >Royal Free Hospital and University College Medical >School >London >f.mattes@ucl.ac.uk > >__________________________________________________ >Do You Yahoo!? >Yahoo! - Official partner of 2002 FIFA World Cup >http://fifaworldcup.yahoo.com -- Marty Bigos Director, Flow Core Gladstone Institute of Virology and Immunology Building 3 SFGH Rm 509 415-695-3832
This archive was generated by hypermail 2b29 : Sun Jan 05 2003 - 19:26:12 EST