Albert Donnenberg wrote- >Et tu Howard? > >My guess is that the huge majority of flow done on the planet in which I >live is performed on analytical instruments with fixed optics. Wonders of >modern multiparameter flow aside, 2 color is still more prevalent than 3, 3 >more prevalent than 4, and four is way more prevalent than 5 or more! > >In our core facility (which features 2 identical 4 color analyzers) I preach >constantly for users to label their axes, but nothing like Howard and Paul >have suggested. Antibody, fluorochrome, and detector are what you really >want to use in the label, especially the first two, since this vitally >important information should be included in the header of the listmode file >where it will be linked to the primary data forever. > >How about the other suggested info? >Center of emissions wavelength: Incomplete (it has been already acknowledged >that band-width is important too) and entirely unnecessary on instruments in >which the optics are standard and specified by the manufacturer. These are >constants within experiments, between experiments, and between labs. In >our lab, and in all others with unmodified Coulter XLs cytometers, FL1 can >be used perfectly intelligibly in an axis label as shorthand for: > >The light from a 488 nm argon laser has been focused by 2 beam shaping >lenses; light scattered at 90o within a quartz cuvette was passed through a >collimating lens, a 488 dichroic longpass filter, through a 488 blocking >filter filter, a 550 nm dichroic longpass filter and a 524 bandpass filter >(505 - 545 nm) before being sensed by the photomultiplier tube arbitrarily >designated FL1. The analog signal has been amplified, digitized, compensated >against signals from other detectors, log-transformed and binned into 1024 >channels which is represented on a four decade scale. > >If I have made modifications to the instrument, or if I am using an >instrument which has easily replaceable optics (like our MoFlo), then it >certainly behooves me to detail this information in the materials and >methods, but PLEASE, not in the axis label (see Chart Junk, below). > >Compensation: I agree that the vast majority of published histograms show >compensated data, and that the majority of flow sins are committed in the >name of compensation. But include compensation in the axis label???? >Cytomation has tried this in their Summit software and once you pass 2 >colors it is an unintelligible mess (e.g. FL1 - X% FL2 - Y% FL3....). >Perhaps (perhaps) the compensation matrix and other "cytosettings" should be >routinely included in the materials and methods section (difficult to do if >these settings change slightly each day as the result of QC procedures). I >say perhaps because this information will only provide a rough guide for >others wishing to reproduce the work, even if they have identical >instrumentation. On the positive side, it will allow the manuscript >reviewer to decide if anything particularly egregious has been done to >instrument or the data. As an aside, I am a huge fan of saving high >resolution uncompensated data in the listmode file and performing >compensation in software. > >Summary: >When you are sitting at the instrument acquiring your data: please label >each parameter with the antibody (or dye), fluorochrome (if this applies), >and (do I speak heresy?) the fluorescence channel (FL1, FL2, etc.). This >information will be forever linked to your listmode datafile and you will be >thankful when you pull the data up for reanalysis several months later. > >When your are trying to publish your data the goal is to create a graphic >that allows the reader to understand at a glance what you measured and to >let the data speak for themselves. Axis labels are but a small part of this >process. Read Edward Tufte, "The Visual Display of Quantitative Information, >Graphic Press, Cheshire, Ct, 1983, for an explanation of "Chart Junk" and >why you don't want to clutter your graphic with redundant or superfluous >information. > >But beware, you may still have to please Paul or Howard. Well, one of the more unkind cuts, anyway; Albert and I are much closer in opinion than he would have you think. First, everybody who has to display any information should read Tufte's books (there are two others, "Envisioning Information" and "Visual Explanations", besides "The Visual Display...) and, if there are a few hundred extra bucks in the budget, take his course, which is highly entertaining as well as highly informative. Second, (just to confuse...), I second Albert's call for different information for internal data and data in publications or presentations. When you take the data, you really do need to note down somewhere what physical quantity you are measuring on each channel, and, for a fluorescence channel, that would include excitation wavelength and filter response characteristics. In the case of an instrument such as the original B-D FACScan, with fixed filters, the information is written down once and for all in the manual (one hopes), so the experimenter can look it up if it is needed. Now, there is some tendency to label successively longer wavelength fluorescence channels as FL1, FL2, FL3, etc., particularly in benchtop instruments, and, should I manufacture a flow cytometer that used a red diode laser to excite APC (or Cy5), Cy5.5, and APC-Cy7, my FL1 would be 660 nm, FL2 700 nm, and FL3 750 nm. To tell you the truth, I am not sure what the center or cut on wavelengths wavelengths and pass bands are for FL1, FL2, FL3, etc. on either the FACScan or the Coulter XL; I don't use either of those instruments. That is one reason I very much favor sticking to at least center wavelengths, or using cut-on wavelengths and "LP" on long pass channels, etc. Again, though, this is information the *user* needs to have available, and probably *not* what should be in publications and presentations. There are two reasons for this. The first is that the data in a display intended for publication are (at least in fluorescence channels) either compensated data (e.g., amount of antibody) or data that didn't need compensation (e.g., DNA content). I'll be ecumenical here, and consider a display of cytokeratin, CD71 (transferrin receptor), and DNA content in a cell sample from an epithelial tumor, allowing me to include a couple of antibodies and a DNA stain among my reagents. What labels did I use? Who cares? It depends. If I used a 3-laser system with violet diode, green YAG, and red diode lasers, I could use DAPI as a DNA stain, without RNAse, and PE (which everybody on the list knows means "phycoerythrin") and APC (which almost everybody on the list would take to mean "allophycocyanin" in this context, even though many would also use it to stand for "antigen presenting cell" in other contexts) as antibody labels for cytokeratin and CD71, respectively, and I could pretty well get away with using the uncompensated 395->450 nm fluorescence, 532->575 nm fluorescence, and 635->660 nm fluorescence signals to represent DNA content, cytokeratin, and CD71. The latter are the biologically significant quantities about which I want to inform readers of my paper or attendees at my presentation, so it would make more sense to label the axes with them than with the physical characteristics of the fluorescence channels, but either set of labels would be correct. However, if I used propidium iodide (with RNAse) as the DNA stain, and fluorescein and PE as antibody labels (which might or might not be possible, depending on how much antigen was expected to be in the cells), I'd have to do a lot of compensation of the original fluorescence signals (say 525 nm, 575 nm, and 610 nm) to get the compensated measurements to reflect DNA content, cytokeratin, and CD71. Not only would it make more sense to use these biological characteristics as axis labels in displays of the compensated data, it would be *incorrect* to use "525 nm fluorescence", etc. I did not propose and am not proposing that the details of compensation or even "compensated" be included in axis labels. The whole point of doing multiparameter flow cytometry is that it allows us to detect and, in many cases, to quantify several different biological materials or characteristics of single cells. That's what we present to our outside audience; it is our own internal responsibility to keep track of our materials and methods to assure that we get the right biological data from our physical measurements. If you're giving a 15-minute talk at a meeting, you should be able to skip even the details of which reagent or label you used for which biological parameter. However, if you're attempting to publish a paper, you owe it to your readers (and, if not to them, to the referees, in the increasingly unlikely event that any of them know the difference between the input stream and the waste stream of a flow cytometer) to provide enough information to enable them to understand how the experiments were done and, if appropriate, to repeat them, etc. So information on which label was used for which antibody, and what nucleic acid stain you used, is necessary, as is information about the instrument. And I don't think it's enough to say you measured FL1, FL2, FL3 on a FACScan, not when it takes about two extra sentences to specify characteristics of the fluorescence channels. As I said before, *I* couldn't tell you with 99% certainty what the passbands for FL1, FL2, and FL3 were, so how could I repeat the experiment using my Cytomutt? But all of these details go into the small print in "Materials and Methods", or, in many of the trendier journals, into supplemental material that is kept on the web site and never makes it into print. So, the bottom line is that you have to play to two audiences. Your biologically relevant data should always be intelligible to people with training in the biological field(s) involved, but you also need to provide information to the people who may want to do similar cytometric studies, and they will need more in the way of detail about reagents, techniques, and physical measurements. And read Tufte. There's way too much extraneous junk in a lot of cytometry displays, and, in many cases, things which ought to be there are left out. -Howard
This archive was generated by hypermail 2b29 : Wed Apr 03 2002 - 11:58:09 EST