The WEB version of these slides can be found on Detectors, Electronics, Data Analysis BMS 633/BME 695Y - Week 3 J. Paul Robinson Professor of Immunopharmacology School of Veterinary Medicine, Purdue University Title page The title page of the lecture. The WEB version of these slides can be found on http://tinyurl.com/385ss Hansen Hall, B050 Purdue University Office: 494 0757 Fax 494 0517 email: robinson@flowcyt.cyto.purdue.edu WEB http://www.cyto.purdue.edu Material is taken from the course text: Howard M. Shapiro, Practical Flow Cytometry, 3nd edition (1994), 4th Ed (2003) Alan R. Liss, New York. 3rd Ed. Shapiro p127-133 4th Ed. Shapiro p160-256

Learning goals Students will lean about the nature of detection systems of flow cytometry Their use, characteristics, benefits and problems The types of detection systems used The way data points are collected and used The principles of data analysis and reporting

Detectors Light must be converted from photons into volts to be measured We must select the correct detector system according to how many photons we have available In general, we use photodiodes for scatter, and absorption and PMTs for fluorescence Detectors This slide introduces the concept of detectors to students.

Characteristics of Light Detection Red sensitive PMT UV line

Silicon photodiodes A silicon photodiode produces current when photons impinge upon it (example : solar cells) Does not require an external power source to operate Peak sensitivity is about 900 nm At 900 nm the responsivity is about 0.5 amperes/watt, at 500 nm it is 0.28 A/W Are usually operated in the photovoltaic mode (no external voltage) (alternative is photoconductive mode with a bias voltage) Have no gain so must have external amps quantum efficiency ()% = 100 x ((electrons out)/(photons in)) Silicon photodiodes The first type of detectors to be covered is silicon photodiodes. This slide covers the properties that are involved in photodiodes.

PMT Produce current at their anodes when photons impinge upon their light-sensitive cathodes Require external powersource Their gain is as high as 107 electrons out per photon in Noise can be generated from thermionic emission of electrons - this is called “dark current” If very low levels of signal are available, PMTs are often cooled to reduce heat effects Spectral response of PMTs is determined by the composition of the photocathode Bi-alkali PMTs have peak sensitivity at 400 nm Multialkali PMTs extend to 750 nm Gallium Arsenide (GaAs) cathodes operate from 300-850 nm (very costly and have lower gain) PMT This slide covers the properties behind photomultiplier tubes. This slide may have too much information on it and might be broken down in the future.

Signal Detection - PMTs Secondary emission Cathode Anode Photons in Amplified Signal Out Signal Detection - PMTs A visual representation of PMTs that shows how PMTs work. End Window Dynodes Requires Current on dynodes Is light sensitive Sensitive to specific wavelengths Can be end`(shown) or side window PMTs

Photomultiplier tubes (PMT’s) The PMTs in an Elite. 3 PMTs are shown, the other 2 have been removed to show their positions. A diode detector is used for forward scatter and a PMT for side scatter. © J.Paul Robinson © J.Paul Robinson © J.Paul Robinson Photomultiplier tubes (PMT’s) An example of PMTs. PMT The Bio-Rad Bryte cytometer uses PMTs for forward and wide angle light scatter as well as fluorescence

PMTs High voltage regulation is critical because the relationship between the high voltage and the PMT gain is non-linear (almost logarithmic) PMTs must be shielded from stray light and magnetic fields Room light will destroy a PMT if connected to a power supply There are side-window and end-window PMTs While photodiodes are efficient, they produce too small a signal to be useful for fluorescence PMTs This slide covers how PMTs need to be handled with care.

High Voltage on PMTs The voltage on the PMT is applied to the dynodes This increases the “sensitivity” of the PMT A low signal will require higher voltages on the PMT to measure the signal When the voltage is applied, the PMT is very sensitive and if exposed to light will be destroyed Background noise on PMTs is termed “dark noise” PMTs generally have a voltage range from 1-2000 volts Changing the gain on a PMT should be linear over the gain range Changing the voltage on the PMT is NOT a linear function of response High Voltage on PMTs This slide explains the properties of PMTs in regards to voltage.

Diode Vs PMT Scatter detectors are frequently diode detectors Sample stream Diode Vs PMT This slide visually compares diodes and PMTs. Back of Elite forward scatter detector showing the preamp Front view of Elite forward scatter detector showing the beam-dump and video camera signal collector (laser beam is superimposed)

Spectral Imaging

Review of Electronics Based on Ohm’s Law, the flow of a current of 1 Amp through a material of resistance of R ohms () produces a drop in electrical potential or a voltage difference of V volts across the resistance such that V=IR DC - direct current - the polarity of a current source remains the same when the current is DC AC - Alternative current - this is generated by using a magnetic field (generator) to convert mechanical into electrical energy - the polarity changes with motion V(t) = Vmax sin (2ft) A wire loop or coil exhibits inductance and responds to alternative current in a frequency dependent fashion. AC produces a changing magnetic field - generates a voltage opposite in polarity to the applied voltage In an inductance of 1 Henry (H) on a voltage of 1 volt is induced by a current changing at the rate of 1 Amp/second - this property is called reactance Review of Electronics This slide briefly covers the basic concepts of electronics.

Review of Electronics Reactance like resistance provides an impediment to the flow of current, but unlike resistance is dependent on the frequency of the current If a DC current is applied to a capacitor a transient current flows but stops when the potential difference between the conductors equals the potential of the source The capacitance measured in Farads (F) is equal to the amount of charge on either electrode in Coulombs divided by the potential difference between the electrodes in volts - 1 Farad = 1 coulomb/volt DC current will not flow “through” a capacitor - AC current will and the higher the frequency the better the conduction In a circuit that contains both inductance and capacitance, one cancels the other out The combined effect of resistance, inductive reactance and capacitive reactance is referred to as impedance (Z) of the circuit Impedance is not the sum of resistance and reactance z=(R2+(Xl-Xc)2)½ (Xl = inductive reactance, Xc = capacitive reactance) Review of Electronics Another slide that is used to briefly cover key concepts that are used in electronics.

The Coulter Principle Cells are relatively poor conductors Blood is a suspension of cells in plasma which is a relatively good conductor Previously it was known that the cellular fraction of blood could be estimated from the conductance of blood As the ratio of cells to plasma increases the conductance of blood decreases The Coulter Principle This slide covers the biological parameters of blood in regards to electrical properties.

The Coulter Principle 2 chambers filled with a conductive saline fluid are separated by a small orifice (100mm or less) Thus, most of the resistance or impedance is now in the orifice. By connecting a constant DC current between 2 electrodes (one in each chamber), the impedance remains constant. If a cell passes through the orifice, it displaces an equivalent volume of saline and so increases the impedance. The Coulter Principle This slide shows how the electrical properties of blood are manipulated via a Coulter filter for analyzing blood cells.

Electrical Opacity This is similar to impedance, except that you use an AC current across the electrodes of a coulter cell When the frequency used is in the radio frequency range (RF) the parameter measured is known as electrical opacity This reflects the AC impedance of cells and is dependent on cellular structure and less on size Electrical Opacity This slide covers how cells impede AC currents.

Linear and Log circuits Linear circuits Logarithmic circuits Dynamic range Fluorescence compensation Linear and Log circuits This slide changes the topic of the lecture into the electronics involved behind light collection. In this slide linear and log circuits are introduced and going to be compared.

Why use linear amps? Factor reduction 10 100 1000 10000 pulse output The problem with compensation is that it needs to be performed on linear data, not logarithmic data. Thus, either the entire electronics must be built in linear electronics, which requires at least 16 bit A-D converters, or a supplementary system must be inserted between the preamp and the display. We need the dynamic range for immunologic type markers, but we can’t calculate the compensation easily using log amps - certainly not without complex math. Flow cytometers amplify signals to values ranging between 0-10V before performing a digital conversion. Assuming this, with 4 decades and a maximum signal of 10 V we have: Why use linear amps? Factor reduction pulse output 10 100 1000 10000 1v 100mv 10mv 1mv

Why use linear amps? Factor reduction 10 100 1000 10000 pulse output The problem with compensation is that it needs to be performed on linear data, not logarithmic data. Thus, either the entire electronics must be built in linear electronics, which requires at least 16 bit A-D converters, or a supplementary system must be inserted between the preamp and the display. We need the dynamic range for immunologic type markers, but we can’t calculate the compensation easily using log amps - certainly not without complex math. Flow cytometers amplify signals to values ranging between 0-10V before performing a digital conversion. Assuming this, with 4 decades and a maximum signal of 10 V we have: Why use linear amps? 10 100 1000 10000 1 100mv 10mv 1mv Factor reduction pulse output

How many bits? Assume we convert linear analog signals using an 8 bit ADC - we have 256 channels of range (2n) (28-256) corresponding to the range 0-10 V Channels difference is 10/256=40mV per channel 1V 10V 100mV How many bits? 0 50 100 150 200 250 Channels

Ideal log amp Linear Log amp Log Channels 0 50 100 150 200 250 1 V 10 V 100 mV Linear 0 50 100 150 200 250 Log amp 1 mV 10 mV 100 mV 1 V 10 V Ideal log amp Log 0 50 100 150 200 250 Channels

Log amps & dynamic range Log amps and dynamic range Compare the data plotted on a linear scale (above) and a 4 decade log scale (below). The date are identical, except for the scale of the x axis. Note the data compacted at the lower end of the the linear scale are expanded in the log scale.

Log/lin display Log/lin display

Ratio circuits Ratio circuits are analog circuits which produce an output proportional to the ratio of the 2 input signals. They are usually made from modules called analog multipliers. Examples are calculation of surface density or antigenic receptor sites by dividing the number of bound molecules by the cell surface area. e.g. Could use 2/3 power of volume to obtain surface area - but few cytometers make this parameter so can use the square of the cell diameter of scatter instead to approximate. pH can also be measured using ratio circuits Calcium ratio (using Indo-1 we can ratio the long and short l) Ratio circuits

Data Acquisition operations which are required to make measurements of a specified physical characteristic(s) of cells in sample Each measurement from each detector is referred to as a variable or “parameter” Data are acquired as a “list” of the values for each variable (“parameter” ) for each event (“cell”) Purpose is to store data And to convert data to numerical form Data Acquisition

} System management Operational Steps Sample Preparation Data Acquisition Data analysis Data Reporting } We will only deal with these in this lecture

Data Analysis Issues to define Data acquisition vs. data analysis Data analysis software Data display Establishing Regions of Interest (ROI) and gating Analysis methods that can change results Data Analysis

Data Analysis Main tasks Cell counting Population discrimination A-D conversion of data Dynamic range must be appropriate DSP for pulses if appropriate Data rates and data acquisition Preprocessing for data acquisition Data Analysis

Data Analysis Output goals Frequency Distributions Distributions (Gaussian/normal) Statistical components Skewness and Kurtosis Compensation/crosstalk Reporting Data Analysis

Data Analysis Histograms Bivariate displays Comparing histograms K-S Cumulative (Overton) subtraction constant CV analysis Bivariate displays dot plots linear regression/Least-squares fits Isometric (2 parameter histogram) Data Analysis

Flow Cytometry Computer Files Listmode files -correlated data file where each event is listed sequentially, parameter by parameter -large file size Histogram files uncorrelated data used for display only Flow cytometry standard (FCS 2.0, FCS 3.0) format used to save data use other software programs to analyze data Note: No cytometry manufacturer abides strictly by the FCS standard Flow Cytometer Computer Files

Data Analysis Software Instrument Software Elite 4.0 Coulter Bryte HS 2.0 Bio-Rad Lysis II Becton Dickinson Commercial Sources WinList & Modfit LT Verity Software ListView & Multicycle Phoenix Software FloJo Treestar Software FCS Express Ray Hicks Flow Explorer Ron Hoebe Free Flow Software WinMDI Joe Trotter MFI Eric Martz Data Analysis Software

WinMDI WinMDI or Windows Multiple Document Interface -requires Windows 3.1, Windows 95, Windows NT or OS/2 Developed by Joe Trotter at the Scripps Institute Available FREE from Internet: http://facs.scripps.edu/software.html Excellent Tutorial developed by Dr. Gerald Gregori http://www.cyto.purdue.edu/flowcyt/labinfo/labinfo.htm WinMDI

Precision - C.V. Precision: CV Sensitivity MESF Units Accuracy and Linearity Noise Background Laser noise Precision - C.V. Shapiro’s 7th Law of Flow Cytometry: No Data Analysis Technique Can Make Good Data Out of Bad Data!!!

Data Acquisition - Listmode v e n t P a r a m 1 P a r a m 2 P a r a m 3 P a r a m 4 F S S S F I T C P E 1 59 1 8 9 2 58 1 1 1 5 9 5 54 6 8 3 3 66 4 6 8 3 5 112 6 8 3 6 115 6 8 3 Data Acquisition - Listmode n etc

Statistical Calculations Number of events – we always collect this Mean: is a measure of central tendency Standard Deviation: is a measure of variability Coefficient of Variation Statistical Calculations

One parameter (frequency) histogram # of events for particular parameter One parameter frequency histogram establish regions and calculate coefficient of variation (cv) cv = st.dev/mean of half peak

Coefficient of Variation %CV Definition = St.Dev x 100 MEAN CV=3.0 CV=3.0 MEAN Coefficient of Variation Crucial in establishing: alignment Fluidic stability Staining of cells

Coefficient of Variation Calculation Statistical (Subjective) Formula (not boundary dependent Objective) • • Coefficient of Variation • • • • • • • • • Least-Squares (Accurate, non- subjective) • • • • • • • • • • • •

Histogram Comparisons The question here might be: Is there a difference between these two data sets? Histogram Overlays We compare histograms to determine if there is a difference between them. If there is, we can make a statement of difference based on statistics. Since we are usually measuring biological phenomena, our conclusion will be related to the biological difference perhaps.

Kolmogorov-Smirnov K-S Test 100 50 Cumulative Frequency Distribution Fluorescnece Intensity Kolmogorov-Smirnov 0 50 100 50 100 Channel Number A good technique for estimating the differences between histograms

Histogram Analysis Normalized Subtraction Match region Histogram Analysis Normalized Subtraction False Negatives • Very accurate • Assumption that control & test histogram are same shape • Match region finds best amplitude of control to match test histogram

Histogram Analysis Integration “Positive” histogram Frequency Histogram Analysis Integration False Negatives False Positives • Very subjective analysis • Not easily automated • Not good for weakly fluorescent signals

Histogram Analysis Accumulative Subtraction Actual Negatives Negative Control Test Number of Events Histogram Analysis Accumulative Subtraction Actual Positives Cumulative Events • Very accurate • Assumption that control & test histogram are same shape • Match region finds best amplitude of control to match test histogram

Basic Histogram Operations Gating or Region of Interest (ROI) selection 1. A gate is a region of interest Gates can be applied to any histogram Gates or ROI can also be applied to mult-parameter plots Gates are applied to select out cells with a desired characteristic. Gates can be additive – this means the results are compounded in the data analysis

Gating Example We have here a histogram By definition it is single parameter Total cells -5000 Gate M1 determines a region from point A to point B on the X axis (log FITC) Within the boundaries of A-B, the gate M1 gives is the total number of cells within the range A-B – the number of cells is 4900 A B

Gating Example We have here a histogram By definition it is single parameter Total cells -5000 Gate M2 determines a region from point A1 to point B1 on the X axis (log FITC) M2 Within the boundaries of A1-B1, the gate M2 gives is the total number of cells within the range A1-B1 which is 4,700 A1 B1

Multiple Gates Total cells -5000 Any number of gates can be applied to a histogram. Gates can be inclusive, exclusive or “either or”. For example, you could select all cells that satisfy gate M6, excluding gate M3 – (M6-M3) would give you the same result as adding gates M1 and M4 (M1+M4). M1 M3 M4 M5 M6 M2

Multiple parameter displays Following display are important in flow Dot plot Density dot plot Contour plot Isometric plot 3D projection Complex displays – TIP and TIG displays Two parameter displays Note: TIP – Tube identifier Parameter – allows the display of data points for multiple samples TIG: Time Interval Gating – allow the display of multiple samples over time.

Isometric Plot - 3 Parameter view - simulated surface is created - 2 parameter data plus cell number - # of particles used as 3rd parameter - 3-D space

Density Dot Plot Contour Plot A B Density Dot Plot / Contour Plot A: Color of dots gives an indication of the identify of subpopulations. e.g. in the above plot the green dots are high density and the mauve are low density areas (FS is Forward Scatter and 90ls is Ninety Degree light scatter or orthogonal light scatter.) B: The color of lines in each contour provides an indication of the number of events in that level of the plot. e.g. in the above plot the green are high density and the mauve are low density with proper contour lines. The data sets of A and B are identical.

More displays Color coded dot plots Here, the multiple colors are in the lymphocyte gate. All of the se cells are identified on the left plot. When applied to the scatter plot, there is a region with multiple colors. In this display, each population has been identified by a different color

Kinetic Analysis 2 D plots 50 ng PMA 0 ng PMA Stimulated Unstimulated Fluorescence 0 ng PMA Unstimulated TIME (seconds) 1800 450 900 1350 Figure 9.3.4 This figure shows an example of stimulation of neutrophils by PMA (50 nm/ml). On the left the unstimulated cells show no increase in DCF fluorescence . On the right, activated cells increase the green DCF fluorescence at least 10 times the initial fluorescence.

Some Multi-data display formats --- --+ -++ -+- +-- +-+ ++- +++ FITC Fluorescence Mo1 CD4 CD8 CD45 leu11a CD20 Tube ID 1 2 3 4 5 6 7 8 9 This is the “Phenogram” format which displays all of the possible binary combinations of a set of fluorochromes – in this case there are 3 colors (n) so there are 2n =8 combinations. Multiple histograms displayed in a combination format J. Paul Robinson, K.Ragheb, G. Lawler,S.Kelley, & G. Durack: Rapid Multivariate Analysis and Display of cross-reacting antibodies on Human Leukocytes. Cytometry 13:75-82,1992 Robinson, J. Paul, Durack, Gary & Kelley, Stephen: "An innovation in flow cytometry data collection & analysis producing a correlated multiple sample analysis in a single file". Cytometry 12:82-90,1991.

The first distribution demonstrates forward gating The first distribution demonstrates forward gating. Cell fluorescence is gated based on their scatter characteristics. Below fluorescence is used to “backgate” the fluorescence signal onto the scatter dotplot Forward gate Back gate Gating log PE 2P Scatter 1P Fluorescence 2P Fluorescence

Specific Cases - DNA analysis Doublet Discrimination 8 x 125  m laser beam shape 16 x 64  m laser beam shape Peak Fluorescence Peak Fluorescence Doublet Discrimination Clumps Integral Fluorescence Integral Fluorescence Slide 18, 11/11/96 of DNA.ppt

From Duque et al, Clin.Immunol.News. Negative Positive HLA-DR T CD13,33 CD19 TdT CD10 CD20 Mu B,T AMLL AML T-ALL AML-M3 AUL ? PRE-BI PRE-BII PRE-BIII PRE-BIV PRE-BV Decision Tree in Acute Leukemia Decision Tree in Acute Leukemia An example of how data analysis can result in a decision process for a data set

Multi-color studies generate a lot of data 4 color 3 color 2 color 1 2 3 4 5 6 7 8 9 10 Log Fluorescence QUADSTATS ++ -- +- -+ Multi-color studies generate a lot of data This example shows how complex the analysis can become for a large set of data with many variables. Represented are the number of dual plots that would have to be displayed to represent the possible number of combinations. It should be noted of course that you cannot display 3 or more dimensions in 2 dimensional space!!

Summary of Material There are 2 primary types of detectors used in flow cytometers These have different sensitivities and applications We collect data in log space mostly because we need a large dynamic range (this is difficult to do in linear space because of limits and costs of hardware) Data acquisition and analysis Types of data formats and presentation formats Data analysis techniques such as gating, forward and back gating