Flow Data Analysis Qianjun Zhang Application Scientist, Tree Star Inc. Oregon, USA
Flow Data Analysis From Cells to Data - data measurement and standard Data display and gating - Data Display options - Gating strategy - Gating controls Statistics
From Cells to Data http://flow.csc.mrc.ac.uk/?page_id=5
From Cells to Data Informations collected Light Scatter parameters - Forward Scatter: proportional to cell size - Side Scatter: proportional to cell granularity or cell internal complexity Fluorescence parameters Time parameters
Flow Cytometry Standard FCS All those information about the cells or particles will be collected and write into a list mode file. The principal goal of the Standard is to provide a clearly defined and uniform file format allowing files created by one type of acquisition instrument to be analyzed by another type on another computer
The History of FCS Format Version YEAR FORMAT SOURCE NOTES 1984 FCS 1.0 Robert F. Murphy et al., Cytometry 5:553-555 First data format 1990 FCS 2.0 ISAC, Cytometry 11:323-332 Data File Standard committee of ISAC revised the FCS1.0 and adopted the format 1997 FCS 3.0 L. C. Seamer et al., Cytometry 28:118-122 Handle large files (>100MB), support unicode text as keyword for the need of internalization. http://murphylab.web.cmu.edu/fcsapi/fcs3.html
The FCS Standard Requires that Each Data Set in a File Contains Three Segments: HEADER, TEXT and DATA, with an Optional ANALYSIS Segment Contain all of the measurements(fsc-h, FSC-A, SSC-H, SSC- A,FL1-H ) for each individual cell processed in a given sample http://www.isac-net.org/content/view/101/150/#3.1
Scatter Profile of Whole Blood Granularity Side Scatter! Smallest and least granular population 0! 200! 400! 600! 800! 1000! 0! 200! 400! 600! 800! 1000! Forward Scatter! Size Largest and most granular population Granulocytes Monocytes Lymphocytes Slide credit to Dr. Holden Maecker Stanford University
Fluorescence Data Display Histogram Bivariate plots Multidimensional plots
&./01!-"!"" - "!"#$%!"&'((%) * +. * + - * +, * + * Histograms - Histograms display the distributions of the events for one parameter -The X--fluorescence intensity -The Y--number of cells per channel " * + +! " "! "!! " #! " $! " % &'()*+,& * + + * + * * +, * + - * +. Histogram overlays Control $-./0$-123304 ( ), ( ) + ( ) * ( ) ( ) ( ) ) ( ) ( ( ) * ( ) + ( ), ( ) )!"#$%&'! ( ) ) ( ) ( ( ) * ( ) + ( ),! " "! "!! " #! " $! " %
Bivariate Displays-Dot Plots Bivariate plots display data with any two parameters, Scatter parameter or Fluorescence parameter Adjunct Histogram!"#$%!&'(($) 8 6 9 8 6 2 8 6 8 8 6 6 *+,-./&+($0 12345!"#$!60745 1 2 3 1 2 * 1 2 1 1 2 2!"#$%$&'(() *+,#-!".$%$!'(() #/,0-!"#$!40567 1 2 3 1 2 * 1 2 1!"#$%$&'(() *+,#-!".$%$!'(() #/,0-6 7 6 866 876 266 Forward Scatter 1 2 2 1 2 1 1 2 * 1 2 3!".$45 1 2 2 Each dot corresponds one or more cell events - Pro: show rare events very well - Cons: Saturate in areas of high cell density, no indication of the relative density of the events 1 2 2 1 2 1 1 2 * 1 2 3!".$67 A perfect way to show both histogram and the dot plot. Showing the histogram of each parameter along with traditional dot plots, provide more info to your data
Bivariate Displays-More Options Density Plot Contour plot Pseudocolor Plot Pseudocolor Plot(smooth) Dot Plot Zebra Plot
Bivariate Displays - Show Outliers Density Plot Contour plot Pseudocolor Plot Pseudocolor Plot(smooth) Dot Plot Zebra Plot
3D
3D movie i!ustration was generated by FlowJo V7.6 Song by The Black Keys Things Ain t like They Used to Be
Flow Data Analysis From Cells to Data - data measurement and standard Data display and gating - Data Display options - Gating strategy - Gating controls Statistics
Gating: Define Population(s) of Interest(POI) Gating basic - Gate on population of interest and gate out population not of interest - Gate out debris, dead cells,doublets/cell aggregates, Focus on POI - Is both objective and subjective Gating tools - Hierarchical gating - Show ancestry& Backgating
Debris Debris are particles smaller than cells, they can easily distinguished by size %"!# %"!# %!!# %!!# ''()* $"!# $!!#!"#$%&'()*+,-./0 ''()* $"!# $!!# "!# "!#!!! "!# $!!# $"!# %!!# %"!# &'()*! "!# $!!# $"!# %!!# %"!# &'()*
Doublets Discrimination What are doublets and why are they harmful? - Doublets are two cells stick together - Doublets can lead to higher background and false positive and especially harmful for cell cycle analysis. Single Cell Doublets Gate on Area Vs Height 1.-/ 2!346 1--/ '!"#$%' ()'*+,-.'!"#$%&'(&)*+,-& 2!345 0.-/ 0--/!"#$%&'( )*+*,.-/ -!"#$%&!' Flight of time!"#$%./!& -.-/ 0--/ 0.-/ 1--/ 1.-/ 2!346
Does It Matter Which Parameter to Use in Doublets 8546 8546 8446 8446 *!+9. 7546 7446!"#$%&'()*!+),-. /0123 *!+9: 7546 7446!"#$%&'()*!+),-: /0173 Rule to live by Choose what works best for you 546 546 4 4 4 546 7446 7546 8446 8546 *!+9, 4 546 7446 7546 8446 8546 8546 8546 8446 8446!!+9. 7546 7446!"#$%&'()!!+),-. /21;3!!+9: 7546 7446!"#$%&'()!!+),-: /81;3 546 546 4 4 4 546 7446 7546 8446 8546!!+9, 4 546 7446 7546 8446 8546
Viability Check Dead cells are sticky and can bind to anything, including your fluorochrome conjugated Ab, generating background and false positive. DNA dyes: penetrate compromised membranes DAPI Hoechst 33342 Hoechst 33258 7-AAD PI Other types Annexin V: ca2+dependent phospholipid binding protein, apoptosis cells Invitrogen's violet-excited live/ dead fixable viability dyes:binding to cellular amine
Dead Cells Exclusion /2-= /--=!"#$%&$''( )*+*, >>37<.2-=.--= 2-= - -. - /. - 0. - 1. - 2 34567!"#$89$:;7<%)<<9
Gating Strategy Gate out Debris Gate out doublets ''()* %"!# %!!# $"!# $!!#!"#$%&'()*+,-./0 ''()* %"!# %!!# $"!# $!!# Debris(NOT) 2!345 1.-/ 1--/ 0.-/ 0--/ Singlets!"#$%&'( )*+*, "!# "!#.-/!! -! "!# $!!# $"!# %!!# %"!# &'()*! "!# $!!# $"!# %!!# %"!# &'()* -.-/ 0--/ 0.-/ 1--/ 1.-/ 2!346 FlowJo hierarchical gating tree %"!# %!!# Plot on FSC&SSC POI /2-= /--=!"#$%&$''( )*+*, Gate out dead Live cells randparent Parent Child ''()* $"!# $!!# >>37<.2-=.--= "!# 2-=!! "!# $!!# $"!# %!!# %"!# &'()* - -. - /. - 0. - 1. - 2 34567!"#$89$:;7<%)<<9
Gating Strategy - Continued %"!# %!!# Monocytes POI ()"% "! & "! % "! $ "! #! ''()* $"!# $!!#! "! # "! $ "! % "! & '()$# "!#! #&!' #!!'! "!# $!!# $"!# %!!# %"!# &'()* (()*+ "&!' ),$* "$6"7 ),$-./.01223 456%7 "!!' Lymphocytes &!'!! "! # "! $ "! % "! &.),$
Backgating as a Gate Validation Tool Loosen Gate Show Ancestry Backgating Tighter Gate Backgating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
Summary I - Gating Clean up the cells by gating out debris, doublets, dead cells before proceeding to POI Use hierarchical gating and show ancestry to easily track your gating trees Use backgating tool to validate your gates
Yet, to Properly Analyze Data, We Need Controls! Compensation controls - To correct spectra spillover FMO controls : Fluorescence minus one - To define positive population Biological controls - Experimental controls(untreated vs treated) - Unstained controls - Isotype controls(for non-specific binding) - Viability controls
Compensation Controls-To Correct Spectra Spillover FITC PE Uncompensated Compensated Wavelength True PE =total FL2 -FITC spill over FITC median fluorescence PE median fluorescence ---------------------------- ---------------------------- Negative Positive Negative Positive ----------- ---------- ----------- ---------- Uncompensated 125 3,540 185 1,650 Compensated 125 3,560 135 135 1,650 185 % Spillover = X 100 3,540 125 %spillover remains the same for the same fluorochrome
Compensation Rule#1 Compensation is about fluorochrome, not about particular antibody or cell type To set up compensation control Don t have to use the same Abs used in the experiment with the exception of Tandem Dyes Don t have to use the same cell type used in the experiment, alternatives are other cell types or antibody capture beads antibody capture beads are great alternatives due to its brightness and sharp peaks(small CV). there are caveats: beads are not available to every antibody type and beads staining has to be as bright or even brighter than you sample staining.
Myths About Compensation Save the compensation setting and use it next time - Almost never, there are exceptions, but you have to know what you are doing Mix positive and negative population of different cell type in one comp control tube - The negative and positive population must have the same autofluorescence, thus the same type - Don t mix comp beads and cells to compensate individual parameter - OK to compensate some parameters with beads and others with cells Set compensation by eye - Never!
Ways to Compensate Instrument compensation Out: old school Subjective: Inaccurate manual adjustment Complicated Acquisition Permanent: the matrix wrote into the FCS file and not removable with FCS 2.0 Software compensation In: new generation Objective: Accurate calculate mathematically Simplifies Acquisition Non-permanent: create and modify matrix with ease
Biexponential Transformation - to Better Visualize Data After Compensation Y=X Y=log(x) y=arcsinh(x) - Transformation is a bi-exponential display tool by introducing linear scales into your axes to better visualize your data - It does not change your data! Herzenberg et. al. Nature Immunology 2006
Use Fluorescence Minus One (FMO) Controls.45*! 67! *2867! *2967!!"!!"! &!!"! %!!"! $!!"! #!!"!!!"! "! Isotype!"! "!!"!!!"! #!!"! $!!"! %! #$%&! - to Identify Positives :! :! :! :! The cells are stained with all the reagents except onefluorescence minus one!"#$%&"'()*+"$,+-!./0)*+"$,+-!.1--2)3$%&"'(! *;<! *;<! :! *;>! *;=! *;=! *;>8?0! *;>8?0!!"#$%&'()#*+," FMO!"#$%&'()*!"! "!!"!!!"! #!!"! $!!"! %!!"! "!!"!!!"! #!!"! $!!"! %! Slide credited to Dr. Mario Roederer
FMO Compensated data exhibits spread Bright single positives may change threshold levels between dim and background in other dimensions Isotype is not useful in many cases Unstained cells or complete isotype control stains are improper controls for determining positive vs. negative expression in multi-color experiments. FMO should be used when the antigen expression is low or dim The Best control!! Graph credit to Dr. Cynthia Guidos Hospital for Sick Children Research Institute Toronto
Summary II - Controls Always run proper compensation controls and use software compensation instead of hardware Get used to the idea of Biexpoenential display to better visualize and present your data FMO is the best control to distinguish positive from the negative
Statistics - Mean, Median & Mode Mean: the sum of the linear values divided by the total number of events. - Require normal distribution - In most case, mean refer to linear armetric mean, if you work on log scale use Geometry mean Median: the point in the distribution where 50% of the population lies to either side - Do not have assumption about the data distribution, it can be in any shape, the same applies to mode - Is less affected by outliers and probably is a better measurement for most flow cytometry data Mode: the value that is of most frequently found in the data MEAN = MEDIAN = MODE MEAN = MEDIAN = MODE > > MEAN MEDIAN MODE MEAN < MEDIAN < MODE
SD and CV SD=Standard Deviation CV=Coefficient of Variance Both measure the width of the peak and require a normal distribution - SD has the same units of the mean, is a absolute value - Low SD means the data is tightly distributed around the mean. - Normal distribution, 68% is within 1SD of the mean. with 99.7% of the data within 3 SD of the mean - CV=SD/mean - A relative number, best for comparing between samples MEAN
rsd and rcv rsd=robust standard divination rcv=robust coefficient of variance Do not require a normal distribution rsd is based upon the deviation of individual data points to the median of the population rcv=rsd/median Robust statistics is a method to produce estimates not affected by small departures from the assumed model The data is less affected by outliers Is more appropriate for most flow cytometry data MEDIAN
MFI=Mean Fluorescence Intensity? MFI=Median Fluorescence Intensity? There is no standard definition for MFI, it must be defined within context From what we learn so far, median fluorescence intensity can be a more appropriate measurement to avoid outliers to skew your data But again it depends on your data
Take Home Messages Flow data is prettier than you think! Backgating as a validating tool to check your gates Always use proper gating controls (eg,fmo) and biological controls(eg, viability controls) Never underestimate the importance of compensation Use Median! Credit was given if a slide or part of the slide was borrowed or modified from other people s work. Flow Cyte training materials and Google search were utilized during the preparation of this talk
Online Resources Purdue University Cytometry Laboratories http://www.cyto.purdue.edu/ Purdue University Cytometry Laboratories E-mail Archive http://www.cyto.purdue.edu/hmarchiv/cytomail.htm Invitrogen s Fluorescence Spectra Viewer http://probes.invitrogen.com/resources/spectraviewer/ BD s Fluorescence Spectra Viewer http://www.bdbiosciences.com/external_files/media/ spectrumviewer/index.jsp The University of Chicago Flow Cytometry Facility Blog http://ucflow.blogspot.com/
FlowJo Session