A pre-sorting data analysis is usually performed in order to determine the thresholds and parameters for the sorting logistics. In conventional FACS and µFACS, data acquisition is divided into two parts, the hardware and the software. The electronic hardware involving PMTs and their amplifiers was discussed in the Chapter 2. I now discuss how the data were analyzed using software that Hou-pu Chou[47] and I have
written.
A LabView Program, LV40, was used to capture signals from the PMTs at 5000 Hz through a data acquisition board (Lab-PC-1200; National Instruments, Austin, TX). This program is able to capture and digitize data from as many as eight different PMTs into eight channels (Figure 3.11).
A C++ Builder program, Pulse Analysis2, was used to determine the positions,
heights and widths of the pulses from the data captured by LV40. Initial baseline 2written by Hou-pu Chou
LABVIEW LV40 Digitizes data at 5000 Hz
PMT1 PMT2
Ch 1 Ch2
Pulse Analysis threshold
max width min width
# of data max deviation new data weight,k
Butterworth IIR filters
smoothes out high frequency noise
Initial Baseline
Initial Standard Deviation sqrt ( ( m = S
s = data/# of data
Sdata) -2 2)
m
Search for a pulse
if data >m +threshold*s
Updated baseline
Updated standard deviation
m = k * + (1 - k) ( m) s = k * ( - m) + (1 - k) ( s)
data old data old
2 2 2
maximum & position
full-width Time Voltage true false Ch 1 Maximum width baseline positions Ch 2 Maximum width baseline positions Output Files Set by user
Figure 3.11: Data processing for pre-sorting analysis. See text for details. µ and its standard deviation σ for each channel was determined by averaging the first 5000 points (preset) from the data. If a data point had a higher voltage signal than (µ+threshold ∗σ), the sum of the baseline µ and the product of a preset threshold and standard deviation, σ, of the baseline, it was counted as a part of a pulse. The maximum value of the pulse was found by iterative search through the data points. The height of the pulse was then determined by subtracting the maximum data value by µ. The width of the pulse (full width of half maximum) was calculated by subtracting the positions of the data points at the first half maximum and the second half maximum of the pulse. The position of this pulse was then determined at its maximum value (pulse height). In between pulses, µ and σ were
updated as a feedback mechanism due to the decline of the background fluorescence. Such decline may be attributed to photobleaching of channel materials and the other particles adsorbed onto its surface. The new data weight κ was preset to be 0.001. The new µ was calculated as µ = κ∗data+ (1−κ)∗old µ and the new σ, σ2 =
κ∗(data−µ)2+(1−κ)(old σ)2. The maximumσ of the baseline should not exceed the
value of threshold*σ. A digital Butterworth IIR (infinite impulse response) filter of various orders and frequencies was also designed into this program to smooth out any high frequency noise in the pulse data. But the filter was not used for any cell analysis. The program then output a file for each channel with the height, width, position and baseline of pulses. These output files were then imported into a worksheet program, such as Microsoft Excel or Sigma Plot, to plot histograms of these pulse data and perform statistical analysis.
Another program, Pulse Ratio Analysis3, was written to further analyze the pulse
data from Pulse Analysis. The purpose of this program is to compare the ratios of pulse heights from two or more PMTs. Two pulse data files from two channels were analyzed. The positions of the peaks from one file were first compared to the positions of the peaks from the other file. If the position of one peak lies within the full-width of half maximum of the other peak, or vice versa, these two peaks were counted as signals from the same cell observed simultaneously by the two PMTs and the ratio of their pulse heights was recorded.
The objective of these analyses is to set up sort gates to screen for target cells. In any cell sorting application, a sort gate, or a set of criteria, has to be set in order to capture the target cells. Typical sort gates may include: forward and side light scatter, viability stains and fluorescent stains for target cell functions of interest. For many conventional FACS, a gate can be set up in a twofold way. For example, a cell has to pass a certain forward scatter gating before it can be considered for the fluorescence gating. In Chapters 2 and 3, I used a simple voltage threshold as the criterion for sorting fluorescent from nonfluorescent cells. Although sorting according to ratios of the signals from two PMTs is not presented in this chapter, Chapter 4
discuss sorting of target cells using ratios of the signals from two PMTs.