1. Compensation The process of correcting for flourescence crosstalk
In-house customer training seminar on compensation
Compensation The process of correcting for flourescence crosstalk

2. History of compensation by traditional flow cytometry
Compensating Image Stream Data
Troubleshooting compensation in IDEAS
AMNIS CORPORATION-Compensation

3. Why is compensation necessary?
Broad emmision spectra
Imperfections in fluorescence filtering cause leakage into other channels
Tandem conjugates
Imperfections in brightfield, scatter and fluorescence filtering cause leakage into other channels
AMNIS CORPORATION-Compensation

4. The electromagnetic spectrum
Flourochromes have a specific absorption and emmision sprectra and although some report the peak absorption or emmission, the actual absorptions and emmisions can be broad depending on the flourochrome.
The strategic use optical filters and mirrors dissect the wavelengths of light so that we can detect specific flourochromes.
AMNIS CORPORATION-Compensation

5. Resonance Energy Transfer
Resonance energy transfer is distance dependent and occurs without radiant release from the donor molecule
AMNIS CORPORATION-Compensation

6. Spectral overlap
Here is an example of 3 flourochromes with absorption in the 488 nm range with 3 different emmision peaks. Note that the peak emmisions are broad and do “crosstalk” into adjacent channels.
AMNIS CORPORATION-Compensation

7. Band Pass Filters
A filter with a transmission that is high for a particular band of frequencies, but that falls to low values above and/or below this band
Filters with transmission that is high for a particular band of frequencies, but that falls to low values above and/or below this band
AMNIS CORPORATION-Compensation

8. Beam splitter
A beam splitter is an optical device that splits a beam of light in two.
A dichroic mirror is a type of beam splitter that is able to split light of different wavelengths.
In its most common form, a beam splitter is a cube, made from two triangular glass prisms which are glued together at their base. The thickness of the resin layer is adjusted such that (for a certain wavelength) half of the light incident through one "port" (i.e. face of the cube) is reflected and the other half is transmitted.
AMNIS CORPORATION-Compensation

9. Traditional flow cytometer flow cell
BD FACSCalibur Optics
Traditionally compensation was done pre-acquisition because flow cytometry was designed to sort cells while collecting data.
AMNIS CORPORATION-Compensation

10. The ImageStream® System
Ch1 nm
Ch2 nm
Ch3 nm
Ch4 nm
Ch5 nm
Ch6 nm
AMNIS CORPORATION-Compensation

11. Spectral DecompositionStack
6 Spectral Channels
nm Scatter nm PE
nm DAPI nm PI, 7-AAD
nm FITC nm Cy5, DRAQ5
0.3 degree separation per channel
Deep Blue Corrector
Spectral DecompositionStack
Petzval Lens Sets
Field Lens
Objective &
Quartz Cuvette
Post Mag Lens
AMNIS CORPORATION-Compensation

12. Flourescence Microscopy
The fluorescent microscope utilizes filter cubes that narrow the wavelenth of excitation and emission designed specifically for each fluorochrome.
Fluorescence microscopes typically do not apply compensation because each color is taken with a different set of optimized filters.
Flourescence microscopes typically do not use compensation because each color is taken with a different set of optimized filters. Get transmission data to demo.
From SemRock
AMNIS CORPORATION-Compensation

13. The crosstalk can be quantified and corrected
Fitc PMT PE PMT
Givans
AMNIS CORPORATION-Compensation

14. Compensation of flow data during acquisition
Classic flow cytometry, compensation is done pre-acquisition for the purposes of sorting cells. Easy to compensate in real time on a flow cytometer because it is computationally faster
Figures from BD FACS Academy
AMNIS CORPORATION-Compensation

15. Formula for traditional compensation
xFn = the amount of signal in detector n that originates from fluorophore x
Dn = the measured signal in detector n
For each detector D=the sum of fluorescence from each fluorophore
D1 = 1F1+ 2F1 +3F1 +…nF1
D2 = 1F2+ 2F2 +3F2 +…nF2
Dn = 1Fn+ 2Fn +3Fn +…nFn
A matrix is defined such that each detector measured value is the sum of the peak fluorescence plus the spillover amount from every other fluorochrome.
The matrix is used to remove the contribution of the non-peak fluorochrome. This gives you the ‘true’ value for the peak fluorochrome.
AMNIS CORPORATION-Compensation

16. Factors affecting matrixes
Crosstalk is fluorochrome specific
Tandem conjugates can have dual band fluorescence due to the inefficiency of RET to the acceptor or degradation of the conjugate
Bright vs Dim fluorescence
Autofluorescence is cell specific
Biexponential plots, properly compensated data may have negative intensities and it’s impossible to plot a negative value on a log plot. The linear region of the biexponential plot allows you to do this.
AMNIS CORPORATION-Compensation

17. Compensation of dim to bright cells
AMNIS CORPORATION-Compensation

18. BD Fluorescence Spectrum Viewer
AMNIS CORPORATION-Compensation

19. BD Fluorescence Spectrum Viewer
AMNIS CORPORATION-Compensation

20. BD Fluorescence Spectrum Viewer
AMNIS CORPORATION-Compensation

21. Spectral Compensation
Post-acquisition compensation is applied to images on a pixel by pixel basis in IDEAS.
Single color control samples used to calculate a 6x6 matrix.
SSC Brightfield FITC PE PE-Alexa Draq-5
Channel imagery consists of signal from the desired channel as well as leakage from other channels due to:
Broad emission characteristics of probes
Limitations in optical filtration
Compensation removes the overlap such that channel imagery reflects only the appropriate probe
AMNIS CORPORATION-Compensation

22. Compensating ImageStream data pixel by pixel
Corrections are applied before spectral compensation.
ASSIST values are used for:
1.Darkcurrent correction
2.Brightfield gain correction
3.Spatial registration
Brightfield compensation is done using the background around the objects and is automatically computed and applied when the data file is loaded in IDEAS.
Single fluorochrome compensation control files are used for fluorescence crosstalk compensation.
Each pixel on a CCD detector has a characteristic baseline output known as dark current offset and a responsivity to light exposure, known as the pixel gain. Correcting for these offsets although low allow for increased sensitivity of dim flourescence. These values are collected and saved when running ASSIST and automatically applied to the data in IDEAS.
Spatial registration errors between image channels is measured by imaging the same object(speedbead) in all 6 channels simultaneously and comparing the location of the images.
AMNIS CORPORATION-Compensation

23. Spectral Compensation Dark Current Correction Pixelated Imagery
Read Out
1023
500
800
700
1000
475
775
675
950
425
725
625
1023
Spectral
Overlap
Signal
Signal
Signal 40
Spectral
Overlap 35 25
Dark Current
Green Channel
Orange Channel
Red Channel
AMNIS CORPORATION-Compensation

24. Darkcurrent correction
Each pixel on a CCD detector has a characteristic baseline output known as dark current offset.
AMNIS CORPORATION-Compensation

25. Brightfield gain correction
Each pixel on a CCD detector has a characteristic responsivity to light exposure, known as the pixel gain.
AMNIS CORPORATION-Compensation

26. Spatial registration
Spatial registration errors between image channels is measured by imaging the same object (SpeedBead) in all 6 channels simultaneously and comparing the location of the images.
AMNIS CORPORATION-Compensation

27. IDEAS Tools
The corrections are available to the user in IDEAS during data analysis.
AMNIS CORPORATION-Compensation

28. Spectral Compensation: Matrix Development
Cross talk matrix is determined by calculating best fit linear regression for each dye into each channel. Slope of linear regression is matrix coefficient. A 6x6 martix of linear equations is solved for each pixel in every image to remove cross talked light.
AMNIS CORPORATION-Compensation

29. Potential pitfalls to watch for
Saturation
Mis-alignment
Coefficients calculated on poorly fit lines
Autoflourescence
Camera staging
AMNIS CORPORATION-Compensation

30. Misalignment
AMNIS CORPORATION-Compensation

31. Saturation
Saturation occurs when the pixel can no longer quantify the available light.
The 10 bit detector provides 1024 bins and once 1023 is reached the pixel can no longer quantify the signal and compensation becomes impossible.
It is therefore critical that events with saturated pixels be eliminated
During data acquisition the laser power, camera sensitivity and cell classifiers are used to reduce saturation
Saturation When a pixel can no longer quantify the available light. We have a 10 bit detector that gives us 1024 bins and once you reach 1024 the pixel is saturated. To apply compensation to an unknown quantity of light is impossible. Imagery with saturated pixels cannot be used to create a matrix nor be compensated. Figure with saturated image and a plot with saturation. Images with saturated pixels and point to crosstalk showing that the numbers do not change.
AMNIS CORPORATION-Compensation

32. Saturation
AMNIS CORPORATION-Compensation

33. Undercompensation due to saturated pixels…
AMNIS CORPORATION-Compensation

34. Staging the camera Charged Coupled Device
Active Channel
96 pixels wide by
512 pixels tall X6
Inactive area
12 pixels wide by
256 FWD Stage Selection
Charged Coupled Device
AMNIS CORPORATION-Compensation

35. Properly compensated single color controls
AMNIS CORPORATION-Compensation

36. Mis-compensated single color controls
AMNIS CORPORATION-Compensation

37. Compensation variation with laser power
Jurkat cells stained with NFkB FITC were run in the absence of brightfield and a 1000 images were collected with 488 laser excitation ranging from 20mw to 200mw at 20mw increments. Compensation values were calculated in IDEAS graphed over changing laser power. Compensation variation with increasing laser power shows little to no correlation to changing laser power.
AMNIS CORPORATION-Compensation

38. Compensation variation with laser power
20 mw positive mw Blank
100 mw positive mw Blank
200 mw positive mw Blank
AMNIS CORPORATION-Compensation

39. Two methods for calculating the matrix
The Means method is used for uniform objects like beads.
The Best Fit is used for objects that have a varied level of fluorescence.
AMNIS CORPORATION-Compensation

40. Coefficient error
Check the matrix
AMNIS CORPORATION-Compensation

41. IDEAS Compensation
Open IDEAS for live demo of compensation from this slide. Return for discussion about saturation, problems
AMNIS CORPORATION-Compensation

42. Workflow
Collect files of single color fluorescent controls with brightfield turned off
Open IDEAS and start a New Matrix under compensation
Add the single color control files to the analysis (open and load files)
Select the single cells using the scatter area vs. aspect ratio dot plot to use as the compensation population
Assign the positive populations to the appropriate channels
Create the compensation matrix
Validate the matrix
Save the matrix
Use the matrix to open data files
A compensation matrix is calculated using single color fluorescent control files that are collected
on the ImageStream in the absence of brightfield illumination. One file of between 500 and 1000
events should be collected for each fluorochrome in the experiment. Once the matrix is created it
can then be applied to the experiment data when batch processing or opening a raw image file.
1. In the compensation drop down menu select New Matrix.
2. Click the Add button in the lower left portion of the window, and navigate to the single
color no brightfield compensation control files.
3. Highlight each fluorescent control file using Ctrl-select and press open. A list of each
compensation control file to be loaded appears in the directory.
4. When all control files have been added to the directory, click Load Files.
5. A prompt to combine the control files comes up. Select no to save on disk space or yes
to combine each control sample into a single duplicate data file.
6. After the files open, a dot plot of scatter area vs. aspect ratio comes up. Using a region
tool identify the single cells to use as your compensation population.
7. Select the newly created population from the compensation population drop down list,
or use the “All” population if few outliers are present.
8. IDEAS will automatically identify the single color positive cells for each channel and
present adjacent channel scatter plots with these events displayed in their respective
colors.
9. Under “positive populations” is a list of each color channel. From the drop down list next
to the channel names, select the appropriate single color positive population and do
this for each fluorochrome used in the experiment. For example, if FITC was used, in
channel 3 (green) select the “3_Positive” population.
10. Once the positive control populations are assigned to their corresponding channels, click
the Create Compensation Matrix button.
11. To view the matrix, click on the Compensation Matrix tab in the upper right hand corner.
12. Validate the matrix by right-clicking each cell of the matrix.
13. The Matrix Coefficient Intensity Plot is displayed. If the green line of best fit does not
exactly overlay the data points on the plot, click the Add Graph to Analysis Area button.
14. Using a region tool select a new positive population that excludes the
outliers.
15. Assign the new population to the appropriate channel using its drop down list.
16. Click Create Compensation Matrix again to re-calculate the matrix using the new
populations.
17. Select Save Compensation Matrix.
AMNIS CORPORATION-Compensation

43. Go forth and compensate!
AMNIS CORPORATION-Compensation

44. AMNIS CORPORATION-Compensation

45. AMNIS CORPORATION-Compensation

46. Calculating the compensation matrix
Two channels of interest X and Y
Compute the variance of X (VarX), variance of Y (VarY) and the covariance of X and Y (CovXY). CovXY measures the degree to which 2 values vary together
Define a 2 x 2 matrix:
Eigen Value =
The major Eigen value is the positive value
Slope =
Eigen values can be found for square symmetric matrices. There are as many eigen values as there rows (or columns) in the matrix. Conceptually they can be considered to measure the strength (relative length) of an axis (dervied from the square symmetric matrix).
Each eigen value has an associated eigen vector. An eigen value is the length of an axis, the eigen vector determines its orientation in space.
AMNIS CORPORATION-Compensation

47. Pixelated Imagery Green Channel Orange Channel Red Channel Read Out
1023
500
800
700
1023
Signal 40
Spectral
Overlap 35 25
Dark Current
Green Channel
Orange Channel
Red Channel
AMNIS CORPORATION-Compensation