From: W. Roy Overton, Ph.D. (overton@UMDNJ.EDU)
Date: Tue Jul 21 1998 - 22:08:01 EST
Jim Houston wrote: > I have two samples. An isotype control is acquired and an fl2 histogram > is obtained with no events in channel 0/1. A test sample is acquired. > The two histograms look similar. On overlaying, a shift to the right > on the test noted. The following questions arise. > > 1: How to get a percent positive? Do I set the integration cursor at 3 > or 2 or 1 % positive based on the neg. control, then use this for the > positive test? > > 2: Do I subtract the two using a subtraction routine? How many events > are needed for accurate subtraction? Do you show the subtracted > histogram result ( which can look strange) in a publication? > > 3: Which method is more acceptable in the publishing world, and how to > explain this. > > In reality are all the cells very weakly positive. would a shift in the > total mean channel be of help. > What is a significant positive result ( <1%, < .1 or what)? > Jim, The data you describe sounds exactly like the data I was dealing with when I was doing lectin binding studies 15 years ago. That data prompted me to use histogram subtraction which gave a lot of false positive events and unacceptable differential histograms. I eliminated the false positives and improved the differential histograms by developing a modification of the subtraction which I call cumulative subtraction (aka: Overton Subtraction) which was published in Cytometry 9:619-626, 1988. This method is available in most commercial flow software (Cytomation, Coulter, Verity, Phoenix Flow, and MSA, not B-D). In the years since then, I have given this problem a lot of thought. In a presentation that I gave at the CAC meeting in Charleston, in 1991(?), I stated that if a unimodal population shifts slightly to the right, it is more likely that all of the cells are dimly positive, not just a subset. It might be more accurate to say that all of the cells are positive, assuming that the test peak and control peak are statistically different. With regard to publishing your results, an arbitrary gate selected at random seems to be the method most commonly used and accepted. The best solution is to improve the data so that there is no overlap of the positives and negatives. This can easily be done by decreasing the background, increasing the specific fluorescence, and improving the instrument's resolution. (Yes, I'm kidding about it being easy.) Unfortunately, this is not always possible. Roy -- |*****************************************************************| | W. Roy Overton, Ph.D. | | overton@umdnj.edu | |*****************************************************************| | Asst. Professor of Pediatrics | Eastern Regional Sales Manager | | R.W. Johnson Med. Sch.- UMDNJ | Cytomation, Inc. | | Cooper Hospital/UMC | 400 E. Horsetooth Road | | 3 Cooper Plaza, Suite 410 | Ft. Collins, CO 80525 | | Camden, NJ 08103 | Phone: 609-478-0019 | | Phone: 609-342-2894 | FAX: 609-478-6224 | | FAX: 609-338-0262 | Ft. Collins #: 800-822-9902 | |*****************************************************************|
This archive was generated by hypermail 2.1.6 : Thu Jan 01 2004 - 17:35:21 EST
