Helen Ho Field Application Scientist, West Cell: 415-902-8734 Email: helen.ho@thermofisher

GSK, Shanghai New User Training Cellomics ArrayScan VTI HCS Reader August 2008version x.6.1.4 Helen Ho Field Application Scientist, West Cell: 415-902-8734 Email: helen.ho@thermofisher.com

HCS Data “Quality” imaging exposuretime max:min outputwindow imagefore- : background • HCS principles don’t differ from scientific assay development generally: • maximize/optimize signal:noise ratios • maximize/optimize max:min response windows • maximize/optimize speed vs. resolution of S:N & max:min • …these are determined by sample quality, imaging, and analytical input parameters • images’ foreground:background  output max:min responses • images’ foreground:background depends on both… • specimen quality, including labeling reagents and support matrix (e.g., plastic vs. glass well bottoms) -and- • imaging exposure times (eventually, background will rise with foreground, approaching point of “diminishing returns”) • output min/max response windows depend on both … • image foreground:background (sample quality, exposure times, etc.) -and- • analytical input parameters • “GIGO”: “garbage in, garbage out” • HCS assay data’s “legitimacy” depends on the suitability & quality of an assembly of steps. • “A chain is as strong as its weakest link.” • Quality and suitability of… analytical input(parameters) output(features) specimen image • After good sample preparation is achieved, the 1st step in HCS is getting a suitable image. But this is circular/bidirectional between adjacent links… specimen  image  image analysis input  output Typical assay development principles apply: The outputs are not meaningful unless the inputs are. Certain input/outputs can be progressively optimized. For example, Reference Level-based, derived, non-intrinsic features (e.g., %HIGH_, %Responders, %<2N…) can be attended to after “raw”, intrinsic features are solid based on robust input settings.

Object- the item(s) in an image that you wish to analyze. This can be a cell, a nucleus, a colony, or an organism Object Identification- the process of determining which pixels in an image are part of an Object or part of the Background. This is done by creating or modifying a Threshold. Threshold- the intensity value that delineates object pixels from background pixels. Maybe specified directly or indirectly. Field- from microscopy’s “field-of-view’- the area of a Well visible by the camera Channel- a collection of image acquisition settings for a single image; including exposure time, Dye selection (wavelength filter configuration) Image Set- one or more images acquired from a single field, comprising one image per channel Primary Object- generally the objects in the first Channel Sub-objects- an informal term for objects that belong to a primary object- such as neurites or spots assigned to a single cell. Object Segmentation- the process of separating objects that are very close to one another, for the purposes of analyzing them individually Parameter- generally, an input that you make to the software Feature- an output data point based on the image and data analysis process Cell Features- numeric data points describing the measured properties of a single Object Well Features- numeric data points describing the measured properties of a Well of cells/objects- most often, statistical calculations (Mean, SD, % responders, etc.) A Brief Cellomics Glossary

Power on instrument. ArrayScan VTI: Single green switch on right side or front of instrument. For VTI Live cell chamber module, see Appendix. Power on computer. Order of 1 & 2 does not matter. But… Do NOT power off instrument (or components) while ArrayScan software is open. If do, close software & re-boot computer. Log on to Windows on the computer: May not be able to use same login as at your desk PC. May need to be part of a Cellomics domain group. Only YOUR network administrator/IT support staff can help you with this. That person is:____________________________________THIS LOGIN IS VERY IMPORTANT for full functionality. Never let IT staff change name of computer without consulting Cellomics. Some functions (e.g., entire Twister) could be lost & require full re-installation. Launch Cellomics software. Instrument: Shortcut like “Cellomics ArrayScan Instrument”To start software in Disk Mode, have instrument powered off. vHCS:View Either (a) or (b) will prompt you for a Cellomics login. NOT same as Windows login. Default is “cell/cell”. Yours?: ______________________ Your system admin may have/should set up each user with their own Cellomics login (see ArrayScan Administrator tab in the ArrayScan Operator’s Guide) to personalize your files, results, & activities on the Cellomics system. NOT as secure as Windows (or your network’s) login, so keep simple, and know others may see it. Suspect problem? Use HCS Administrator software. If some Cellomics software behavior is suspect, launch your HCS Administrator> Diagnostics> Run Diagnostics. All Results? You’re OK. Otherwise, may need your IT staff’s support. Email Report, or drag down row buttons to select all, then Ctrl+C to copy (then paste and email, etc.). Or PrtSc keyboard button (then paste, etc.) Info also useful for learning AppSvr name, acct name, network bandwidth, server path & free space, database name & location… Getting Started

LIGHT SOURCE: EXFO http://www.exfo-lifesciences.com/products/X-Cite120.html Turns on via main ArrayScan power switch. Stable in 90s. For max life, keep on ≥ 5 minutes. Some users turn lamp off (inside ArrayScan right front door), but leave instrument on, to save bulb life – though potentially confusing. If turned off, and try to re-light before fully cooled, flashing bulb with “no” symbol may appear on lamp display (inside ArrayScan right front door); auto-re-strikes when cooled. bulb rated for 1500h, but note drop-off begins ~1000h in plot. bulb is 100% user-changeable (requires no alignment) instructions come with replacement bulb ordered from Cellomics Recommendation: always have a spare on hand (order one now? P/N = IC001001) CAMERA: Hamamatsu ORCA-ER (for specs see http://www.leedsmicro.com/hamamatsuorcaer1.asp) 12-bit: this means 4096 grey levels Usually images are acquired with the camera Binned 2x2, resulting in an image of 512x512pixels Power is automatically controlled, no user intervention required See Appendix for resolution information ArrayScan VTI Light Source and Camera

Overview of Create Protocol View • ArrayScan / vHCS:Scan software anatomy, top to bottom • Title bar: Software name, version & build (important), current view.Here showingvHCSTM:Scan Version 6.1.0 (Build 6018) – [Create Protocol] • Menu bar: Little redundancy with other controls.File Options View Tools Window Help • Toolbar: The 1st three icon buttons are the main “views” in the software (shown here is Create Protocol view). • Main GUI body (all pale blue background canvas). Notice the Main GUI in this [Create Protocol] view has multiple panes: • Image Acquisition • Scan Limits • Assay • Channel Specific Parameters • Object Identification • Image Display Options • Status bar (gray bar along bottom)Here showing: unlocked protocol icon, Protocol Modified; Operator: cell. 6

What is in an Assay Protocol? Objective (magnification) Imaging resolution What color fluorophores to image Exposure (CCD integration) times Autofocusing settings How many images to take per well BioApplication image processing algorithm settings Kinetic and Motility Feature Setup What’s NOT in an Assay Protocol? Plate Type Form Factor Which wells of a plate to analyze Which field of a well to start in Whether to save images Maximal image displays during scan Reference Well selections Scan software’s 3 views Lesson 1: Assay Protocols How do the 3 main views of ArrayScan/ vHCS:Scan software present Assay Protocols? Create Protocol viewAccess all settings of an Assay Protocol Protocol Interactive view • Interact with sample plate to adjust various imaging input parameters of an Assay Protocol and temporarily acquire image sets. • With temporarily acquired image sets, adjust various analysis parameters and visualize example results. Scan Plate viewExecute an Assay Protocol (run an assay).

Create Protocol View Settings accessible in Create Protocol view are the only settings saved with an Assay Protocol. (Some settings are redundantly accessible in Protocol Interactive view; if changed in that view, they’re also saved when an Assay Protocol is saved because they’re the same settings, just a different view).

Create Protocol View: Image Acquisition pane (except for Configure Focus, these settings accessible only in Create Protocol view) • Camera Configuration- on an Instrument you will have only one choice; on Toolbox clients choose the camera that you have or plan to acquire images with • Objective • All objectives installed on system appear in the drop-down list. • Assay Parameters: PixelSize auto-computed. Understand benefit of UseMicrometers=1 and impact of Objective setting on area- & length-based Object Selection Parameters. See BioApplication Guides. • Acquisition Camera Mode (Generally, “Standard”) • Autofocus Camera Mode: “AutoFocus” is likely specified in Assay Protocols provided by Cellomics. “Standard” may yield faster scan times if the Acquisition Camera Mode is also Standard, but it depends on the samples imaged in the Focus Channel; determine empirically or consult Cellomics Support. • Intra-well Autofocus Interval • Illustrated in ArrayScan HCS Reader User’s Guide (in ArrayScan HCS Reader Operator’s Guide). • Determines the points within a well where a focus adjustment is performed. • Conservative generalizations to start: 4 fields@5x; 3 fields@10x; 2 fields@20x; 1field@40x. • Affects scan times & quality, depending on #fields scanned per well (dependent on Scan Limits). • Therefore, best optimized during development of an assay by careful review of all images acquired during a scan (at least review images from wells with highest # of fields acquired). • Variables to consider include • sample’s cell density (too sparse? “empty” fields take extended time) • sample’s quality (occasional debris or stacked cells? maybe more frequent Interval) • objective magnification and NA (inversely proportional to Interval) • standard- or thin-bottomed plates (thin wavier? maybe more frequent Interval) • plate’s well density (384, 96, etc.; esp. if thin-bottomed, lower density wavier) • users’ and algorithms’ analytical tolerances… • Configure Focus… Generally, use defaults. • Autofocus may consider a Focus Channel image out-of-focus (/sparse) & prevent BioApplication analysis (though save the field’s images). Relax… avoids this prevention (the case for all Disk Scans, since Autofocus isn’t used). • Relax the Pass/Fail Criteriamay be useful when: • High-NA objective on standard thick plates (see User Guide Appendix "Plate Types & Objective Selection"); • Object intensities in Focus Channel vary widely; • Sparse fields are potentially common; • Multi-layer cells are potentially common; • Relax the Pass/Fail Criteria also has additional effects (e.g., focus accuracy may be less), so it’s not recommended as default.

Create Protocol view – Scan Limits pane Scan Limits paneSome logic is built in among checkbox selections. • Scan Limits pane, Intra-Well Stopping Criteria section: • When will the ArrayScan stop scanning a well, and proceed to the next well? • When at least one check-able Intra-Well Stopping Criteria is met or exceeded. • A “Sparse Field” has fewer objects than the Min Objects For Field (always keep >0). • Max Sparse Fields For Well: Sparse Fields must be contiguous to hit this limit. • All Scan Limits can have dramatic impacts on scan time. Think through logic very carefully. • How many objects is enough? • Must be answered by assay developer’s statistical requirements, which can be arbitrary and vary between organizations.. • Cellomics has some recommendations, but these are in the context of specific assay statistical requirements, arbitrarily selected, though common. • Here’s a Good Practice:Early in assay development, scan all possible fields in model wells to create a comprehensive, stored image set. Create a series of batch rescans using the Disk Scan Wizard (or ArrayScan software in Disk Mode) to rescan the images, where the only variable between scans is decrementing # of fields per well. Such a batch is programmed using Plate Protocols, where, in this case, the Plate Protocols differ only in one regard—their Assay Protocols differ only in Max Fields For Well, with all other Scan Limits boxes unchecked. Examine the results of such batches to determine a minimum sample size at which StdDev is stable for principal output well features. • NOTE: An “object” in the context of Scan Limits is BioApplication-dependent. • Scan Limits pane, Intra-Plate Stopping Criteria section:What use is Max Sparse Wells For Plate? • In mass sample prep operations (e.g., automation), if most cells are lost by unsuitable liquid handling or a treatment undergone by all wells, or a necessary labeling regent was not suitably applied… • When scanning a stack of plates, more efficient not to scan “bad” plates.

Create Protocol view – Assay pane Best practice: begin with a default assay protocol already written for the BioApplication you want to use Assay paneAll subsections present in every BioApp, but options within subsections may vary. • Assay Algorithm: NOTE: Generally, do not change the Assay Algorithm selection unless you are familiar with its impact. You will not be prompted that changes not saved before altering this selection are irrecoverable. • No. of Channels: BioApplication- dependent (see BioApplication Guides). • For some, the standard # of channels can be scaled back. • For some, additional “gating” channels can be added. • Focus Channel • Specifies in which channel Autofocusing will be performed during a scan • Generally, choose a channel imaging a dye whose quantitative fluorescence is not critical and/or is relatively photostable (e.g., a nuclear Hoechst stain or demarcating the nucleus). • Features to Store: default, and a good place to start, is to store all features. Select features to Store when screening or doing intensive kinetic assays • Kinetics Checkbox: used to set up time course experiments. Configure button becomes active when checkbox is ticked. • Assay Parameters: List is BioApp-specific (see Guides) & varies with No. of Channels. • Hide Advanced Parameters: BioApp-dependent (see respective User’s Guides). Generally, parameters that refer to Data Analysis (Levels, _CC) are Advanced Parameters • Default Well Feature: see Scan Plate view window (and plate diagram coloring in vHCS™:View).

Channel Specific Parameters- “Dye” and Filter Selection • The Label is an image caption; for your record keeping only (Displayed as Caption in vHCS:View) • Dye refers to the filter configuration to be used in that channel • Consult Appendix B of your ArrayScan User’s Guide tab within your ArrayScan Operator’s Guide binder, which explains why scans are faster if all fluors in a sample can be imaged by the same XF set. • All else being equal, Cy5 channels are usually the weakest (because of both excitation and detection optics). Therefore, use a Cy5-like dye only if a 4th color is needed, and use to label/ report the most abundant target, etc. • Other dyes/fluorophores are certainly possible; consult Cellomics Support or http://probes.invitrogen.com/resources/spectraviewer/ or https://www.omegafilters.com/front/curvomatic/spectra.php.

Imaging: AutoExpose and Fixed Exposure Types of Exposure Methods: • Fixed: Specify a time in seconds for exposure in each channel- that exposure time will be used throughout • AutoExpose: still must specify an Initial Exposure Time. That exposure time will be used as a starting point to identify a time that meets your criteria (see below) • Key image analysis steps are tightly dependent on image characteristics that are, in turn, largely determined by exposure times. • Matching exposure times to existing input analysis parameters is often more productive that resetting analysis parameters. • Instrument components change over time, so autoexpose options that can compensate will prove useful, even when using Fixed exposure during plate scanning. • Even when stored images are re-loaded, their % of dynamic range will be reported by the Autoexpose routine. How bright should my images be? Here are some guidelines for what % of dynamic range may be suitable for a few classes of fluorescence labels. 25%: Good for mainly “structural” labels (e.g., counter-stained nuclei). Still affords some range for differentiation between, e.g., interphase, mitotic/apototic nuclei. ~20% can be tried for minimizing autofocus time & maximizing speed. Typically also sufficient for “binary” assays where the interest is whether the label is there or not, yes/no—e.g., mitotic index, negative viability (“dead dyes”), micronuclei, and others. 35-40%: For “intensity” quantifications. Localized intensity differentials may reflect >40%. Also reasonable starting range for “structural+intensity” labeling—e.g., nuclei in assays needing more discrimination (e.g., DNA content/cell cycle). 55-65%: Good for multi-modal structural labels where accurate identification of more than one intensity level is of interest—e.g., some neuronal cell bodies are considerably brighter than neurites that also must be accurately identified. High dynamic range may also be required for some indicators of cellular metabolism, physiology, health, etc. Best Practice: When trying higher %, be on guard for saturation. Optimal exposure ranges are worth careful investigation—esp. in “intensity” assay, large improvements often emerge. If your BioApplication has multiple channels measuring the same thing, fill them all with variable exposure times during assay development pilots.

AutoExpose Options and Procedures If a scan is executed with an Assay Protocol specifying AutoExpose as Exposure Type, before the whole scan begins, the system… • Goes to 1st well in List of wells to Autoexpose. • Goes to 1st field in List of fields to Autoexpose. • Autofocuses (using Configure Focus…> Autofocus Options> Camera Exposure Time> Focus Exposure Time For Autoexpose—note, AutoFOCUS setting!). If Autofocus fails, so does that field!Proceeds to next. • Execute Autoexposure Method. • Failed any channel? Try in next specified field (i.e., in example, field 5 is used only if field 4 fails). • Succeeded for all channels? Temporarily hold determined exposure time(s) in memory. • Go to next well, as applicable, repeat, hold time. • Compute mean of all “held” times (I.e., where succeeded) & use it for respective channel(s) for whole plate scan. Benefits: • If reagents, optics, etc. vary over time or from plate to plate, this Exposure Type may be compensatory. Caveats: • No automatic way to know if specified wells/fields are in good condition in every plate. And final exposure time(s) is only from successful wells. • Screeners may wantnot to have assay variables masked by system compensations. • Users need to remember what Assay Protocols have set as wells/fields to autoexpose and whether the same for every plate used with that Protocol. An Intermediate – Best Practice: Use Autoexpose in Protocol Interactive on fields deemed representative by you, but specify Fixed ultimately in the saved Assay Protocol. • AutoExpose Options • Whereas Background Target(Mode) Method refers to image histograms, “Peak” does not refer to a histogram peak; after Skipping specified brightest pixels, Peak is X-axis intensity of remaining brightest pixel. “Max Intensity” • Peak methods good for targeting brightest signals as % of camera’s 4095 grey-level dynamic range. • Background… method: When signal is expected to be absent, still can get useful exposure time (typically low % so when fore-ground appears in unknowns, it extends into useful range).

Summary of Create Protocol functions Choose your BioApp by choosing an Assay Protocol If both checked, whichever comesfirst. “Objects” is BioApplication-specific. If both checked, whichever comesfirst. “Objects” is BioApplication-specific. • Saved with the Assay Protocol (User Guide error says “stored with results for the plate”). • Admin. changes to read/write status auto-written here. • BioApplication Event Wizard entries auto-written here. • Max 4000 characters, but keep <255 if possible. • Advanced training topics. Retain settings specified… • in Cellomics-provided Assay Protocol(likely Standard & Autofocus respectively) • …or… by Cellomics support BioApp.-specific. • Advanced training topic • Do not change (default=25)unless advised by Cellomics BioApp.-specific • Depends on plate type, objective (NA & mag), etc. • Best determined empirically. • Starting pointers (conservative):5x, 4 fields10x, 3 fields20x, 2 fields40x, 1 fields Must choose both a Method and a Value here; best optimized in Protocol interactive contiguous “Objects” is BioApp.-specific. See Scan view window(& plate diagram coloring in vHCS:View).

Scan software’s Anatomy Lesson 2: Protocol Interactive view What’s Protocol Interactive view for? • Select a Plate Type Form Factor • Navigate to a well and field (within a well) • Autofocus (& adjust if desired, + DZ’s) • Alter Dyes (filter sets) if desired • Image Fields (1 at a time) • Alter Exposure times (& Auto settings if desired) On an Acquired Image Set… • Alter overlays & composite & colors • Edit Assay parameters • Run Algorithm to see results of current settings • Identify Objects and adjust settings, plus modify Object Selection Parameters • Double-click image to get “pop-ups” and zoom in, see numerical results, toggle overlays. What parts of an Assay Protocol are NOT accessible in Protocol Interactive view? • Objective (magnification) • Camera settings: Acquisition Modes (Binning) & Gain • Autofocus Interval • Scan Limits • Assay Algorithm selected • No. of Channels, Image/Channel Labels • Default Well Feature & Extents • Kinetics Setup • Select Features to Store How do the 3 main views of ArrayScan/ vHCS:Scan software present Assay Protocols? Create Protocol viewAccess all settings of an Assay Protocol Protocol Interactive view • Interact with sample plate to adjust various imaging input parameters of an Assay Protocol and temporarily acquire image sets. • With temporarily acquired image sets, adjust various analysis parameters and visualize example results. Scan Plate viewExecute an Assay Protocol (run an assay).

Protocol Interactive View Screenshot Object Identification Settings Object Selection/ Validation Settings Image Display Settings Most bioapplication-specific image and data analysis parameters are accessed by this button 17

I-V-S : Identify, Validate, Select IN “PRIMARY” CHANNEL(S)** Object Identification • Separate foreground from background. • intensity-based (Some Assay Params can alter raw intensities [e.g., BackgroundCorrection] or shapes/sizes of Identified Objects [e.g., Smoothing].) • Controlled by Object Identification Value • Results depend on Method. Object Validation • “Is this identified object valid?” Criteria are BioApplication-specific. • Objects in primary channel(s) must also be Selected to be Valid. Object Selection • As noted above, in primary channel(s), Object Selection Parameters also Validate objects*Most V3 BioApps have an Assay Parameter for RejectEdgeObjects ** Some BioApps have >1 primary channel (e.g., Cell Spreading, Neuronal Profiling, Micronucleus). Object Identification • “Rejected” overlay* • Must be  in current Ch to see • Algorithm ignores these objects, so no data is reported. Object Validation fail Object Selection fail “Selected_” overlays data results * BioApp-specific (a few cases have no “Rejected_” overlay).

Object Identification Methods Threshold Background Object Number of Pixels Background Peak Object Peak 0 4095 Threshold Background Peak Object Peak 0 4095 IsoData Method Peak Triangulation Method Object Identification Value shifts auto-computed threshold by Value’s %age • How do you think FixedThreshold Method works? • What do you think the impact of BackgroundCorrection is on these methods? • What do you think happens if there is an extraordinarily bright artifact (lint, debris, etc.) in an image?

Always use it, whenever you can! And, when you can,… Best Practice: Use Fixed Threshold for Object Identification Method (if you can). When can I not use it? When any image may have near-confluent fluorescence (maybe): Imagine the Raw Image is full of objects; what would the blurred Background image look like? …And the Background Corrected Image? In sum, has to be some background in every image, in every channel… Why every image/channel?… BackgroundCorrection is a “global” Assay Parameter—i.e., when activated (non-zero Value), it is performed on every image in that channel Adjustable individually in each channel in V3 Bioapplications High impact on all aspects of processing & analyses Performed BEFORE Object Identification All output feature values, etc. vary with Value Choosing a Value “Aggressive” Values (low) ~ diameter largest typical object. for reliably monodisperse objects/signal. affords “localized” correction “Conservative” (large Values) for heterogeneous objects/signals. Greater than 2x largest diameter? Heterogeneity too unpredictable? Consider Max Value (255) Common V3 Assay Params: BackgroundCorrection BackgroundCorrectionBest Practice: Use…Whenever you can! Background image constructed essentially by blurring Raw Image. Value inversely proportional to how much blurring.

Scan Plate View Select which Wells to Scan Select starting Field Type any text desired, for descriptive purposes. When scan completed, this text will be searchable, too. Generally, keep this field <255 characters (though not necessary) and use simple alpha-numeric characters. In the plate Avoid non-simple characters in Plate ID (Barcode) or Plate Name. The Scan software will accept them; however, subsequent searches in vHCS:View using characters like “#” will fail. You can change the feature to display here but it is NOT saved with the Assay Protocol.

Launch via Start> Program Files> Cellomics> vHCS:View. Highly developed software with user-friendly GUI (graphical user interface): Almost complete redundancy between menus, toolbar icons, and right-click context-sensitive pop-up menus. Uses Windows conventional Ctrl+click & Shift+click functions for selections on text lists, spreadsheets, etc. Uses Windows conventional Ctrl+C (copy) & Ctrl+V (paste) from spreadsheets. Full on-line Help menu. Best Practices – Tips for fast, easy use & learning Frequently right-click on GUI elements (graphs, images, diagrams, spreadsheets) for command options on single GUI elements. Use the Windows menu on the menu bar for managing arrangements of multiple windows/views and navigating between them. Check menu bar menus for additional command options, and when multiple GUI element types are displayed (e.g., a view with multiple graphs). Notice the interconnections of interactive GUI elements (e.g., click a point on a graph, and note spreadsheet highlight changes with it). Be fearlessly exploratory—except for Delete Plate(s). Aside from Delete, you cannot change any data using vHCS:View: It’s only a tool for viewing data in different ways. Find a Plate result of interest: Well Details Row/Col mode (so lay out plates accordingly) Use down arrow key to “flip” through Feature list double-click anywhere on graph (except points) to Customize Copy/export spreadsheets or graphs Methods (bar, points, points+lines) File> Report Cell Details toggling overlays Maximize images (and de-max to restore; don’t click  or Cell Details closes) making New plots Scatter plots vs histograms maximize, copy, export spreadsheets/graphs select spreadsheet column heading, right-click, Sort…, useful to find where to use subpop cutoffs or other assay params, as are histograms/scatter plots. show can call Cell Details from multiple highlighted Well Details wells (contiguous) Other common uses besides Well & Cell Details: Multi-Plate View Graph menu> Plot on Single Graph Spreadsheet menu> Export to… Multi-Feature View Graph menu> Plot on Single Graph (only good for similar scales, and rough with multiple rows) Spreadsheet menu> Export to… Reviewing results in vHCS:View

Quiz 1: Assay Protocol Starting with a provided Assay Protocol for the Target Activation BioApplication, 10X objective (NA 0.3), build an Assay Protocol to accomplish the following: • Magnification & resolution (see Appendix pages of this training packet) • ArrayScan 3.5/4.5, ArrayScan VTI (1.0x coupler), 0.645 micron/pixel, 512x512 pixel images • ArrayScan VTI (5.5) with 0.63x coupler, 1.024 micron/pixel, 512x512 pixel images • Analyze 93 cells in each well, but don’t take more than 3 fields per well to do it, and don’t try for a 2nd or 3rd field if the 1st field has <50 cells. • 2 channels, focusing on nuclei • Ch1: Hoechst; Exposure time that will bring foreground intensity to ~1/4 dynamic range (±0.4%, ignoring 0.25% of the brightest pixels) • Ch2: Alexa Fluor 594; Exposure time that will bring background intensity to ~200 intensity units (±0.4%, using smoothing Kernel Size 25) • In RGB composites, make Hoechst appear in blue and AF594 in red. • In Ch1, overlay analyzed objects in blue and Rejected cells in yellow • In Ch2, show Rejected object overlays, plus Modify the Ch1 mask to make it 2 pixels wider in Ch2 and show it in Ch2 as red. • Don’t use Illumination Correction in any channel. • In image captions, call Ch1 “Nuclei” & Ch2 “Target”

Redundant Access in Protocol Interactive View • Redundant access in Protocol Interactive view • These 4 boxed regions in this Create Protocol view window are accessible from within Protocol Interactive view too.Changes made in either view can be/are saved with the Assay Protocol. • Configure Autofocus • Number of Channels • Assay Params. (including Hide…) • Except Kinetics configuration, Select Features to Store • Channel Specific Parameters, • including all Object Identification parameters • including all Image Display Options parameters • except Label, Apply Illumination Correction, Gain Options

Pop-up windows in Protocol Interactive Best Practices rely heavily on productive uses of pop-ups. Use Window menu inside ArrayScan software to manage pop-ups productively. Involves using Windows TaskBar to switch between main software window and popups. Both are shown in this illustration. • Single images at a time: • Draw box to zoom in for details of overlay accuracy, etc. • Use overlay toggle button to see underlying structures. • Brighten images to avoid missing accuracy detail. • Series of same image, varying input parameters. • Don’t move the window once popped up. • Change a parameter, Run Algorithm, & pop up result; stacks on top of prior popup. Now use Windows’ TaskBar to flip back & forth between them to detect subtle changes in overlays. • Repeat with multiple popups, investigating a range of values for the same parameter. Type each value into popup’s text field to track which value is best.[But read note typed into text field in example shown here. Images can be saved, but not text.] • Pop up images from multiple fields for side-by-side comparisons. • Tile Popups Vertically (Ctrl+T). • Select equivalent Bits To Display. • Click Well Features button to compare numerical results. • Same approach for fields from multiple wells—replicates or opposites (positive/max & negative/min). Double-click anywhere onembedded imageto release acopy of the image in separate pop-up window [Quiz 1 answers (2 pages) in Appendix!]

There are only 2 images on this page – top & bottom row. Across each row are different display adjustments of the same image. The images in the top row are from a field right next to the field in the bottom row! In 1st column – Full or Max – do you think the bottom field is dimmer? Do the 12-bit histograms suggest that? The 1st column illustrates the common illusion of Contrast-Stretching—an auto-adjustment of brightness/contrast done on an image-by-image basis. Contrast-stretched images canNOT be legitimately compared for visual intensity differences! Beware: Full Contrast Stretch is the default display adjustment! For image-to-image intensity comparisons, Contrast-Stretching should be deactivated, and Bits-To-Display selected. Histograms under images are 8 bits—256 gray levels—the maximum # of grays display-able on a monitor. Bitshift 3 (i.e., bits 3-10, or grays 8-1024 from the 12-bit source image) illustrates the best spread of the gray levels across the 8 display bits. For the top field, Bitshift 4 resembles the Contrast-Stretched image, whereas for the bottom field, it’s Bitshift 5. Contrast Stretching – Common Illusions Display Options or 8-bit histograms 12-bit histogram y stretch zoomed 12-bit histogram y auto-scaled 12-bit histogram y stretch zoomed to see low frequency pixels 12-bit histogram y auto-scaled • In the bottom field, the bright spot shows up on the 12-bit y-stretch histogram. When auto-re-mapped to 8 bits, those high values prevent expansion of the display histogram.

FAQ’s- Protocol Interactive View Windows TaskBar hidingStatus Bar/Tray Status Bar/Tray info whenpointing at image Status Bar/Tray info whenmoving stage (incl. Focus) • Q. How can I see my numeric data (cell and well details) interactively? • Use icons below image (shown at right) after Run Algorithm. • All cell and well details are shown, regardless of what you’ve selected to Store • Field details are only available if the BioApplication reports them. Consult the User’s guide. Well details Field details Cell details Q. How can I tell what the intensity in a region of the image is? Where can I tell at what Z position I am? A. Status Bar/Tray: Lower left corner of GUI. Information is only displayed for a few seconds. • Top fig.: Window's TaskBar hiding Status Bar/Tray. Right-click empty space on Windows TaskBar> Properties> Lock…, Auto-hide,  Keep…on top… Thenceforth, move pointer all the way to edge to get TaskBar. • Status Tray when pointing at image: Intensity (x,y)= raw value (uncorrected) intensity of pixel under mouse pointer. (x.y) are that pixel’s coordinates. • Status Tray when moving Stage/Focus (z) motor: • Position = [x,y,z]. Since only displayed transiently, to see again without moving, use 0 (zero) for Step Size (microns): and click [–] or [+] button. • Manual Focus buttons are opposite what some think: [–] moves up, toward/into sample; [+] down, away from sample. • System’s current z-position is 3rd Position coordinate shown (in example here, it's -9023.500).

Your trainer shall have helped design an Assay Protocol suitable for an available training sample and your BioApplication interest. When you move(d) from the Create Protocol view to the Protocol Interactive view, 3 new elements of the toolbar become active: Name them.Are these 3 settings saved with an Assay Protocol? There are 2 top-level panes: Name them, & the 2 sub-panes of each. There are two “orphan” tables with bold headings: Name them. What do you see in the Status Bar/Tray? Load the training plate: How? Go to well ___ (B02?), watching the Status Tray. What do you see there? What Field are you in? Move two fields left; now what Field? Go back to the original Field without using Select Field buttons. Acquire an Image in Ch1, foreground achieving 25% dynamic range, ±2%, skipping brightest 0.05%.What steps were necessary? Is the top portion of the image fully visible? If not, make it so.How? Keep this exposure time Fixed for all subsequent images acquired.How? Acquire an Image in Ch2, again keeping the Exposure method Fixed, but getting 25% of dynamic range.How? Go to another field and get images in Ch’s 1 & 2 in a single click.How? Run the Algorithm to see the results using the current settings in Ch1. What’s your Object Identification Method and Value? Can/should you use BackgroundCorrection on this specimen (see page elsewhere)? If so, use your pointer and the Status Bar in the lower left of the window to determine a “conservative” value. Then use the same to determine a suitable FixedThreshold (Object Identification Method) Value. (Hint: What will BackgroundCorrection do to the background pixel values in this raw image). Adjust the Contrast-Stretching (brightness binning). Display the Well Features. Go back to the Raw Image. Then back to the overlay, then raw, then overlay. Double-click the image, then, on this “image pop-up window,”… Zoom in on a specific region of interest (hint: drag mouse pointer to draw box around the region). In a single click, restore the zoomed image to its 1:1 resolution. Keep this pop-up open and click back on the main window. Change the color of the Rejected_ overlay, and turn on the Rejected_ overlay in Ch2.Are the changes immediate? De-Focus the Ch1 image one Manual Step Size, Acquire an Image Set, then set a Focus Offset (DZ) in Ch2 to restore focus in Ch2, but not Ch1. Go to a different well and re-examine the results of your settings after Running the Algorithm on an Acquired Image Set. Compare these results to the one you just did. How? Exercise: Interacting with Protocol Interactive

Object Identification Methods & Values mL mR • Object Identification is, fundamentally, identifying foreground objects as patches of pixels that are brighter than background pixels. • An image comprises pixels; each pixel has an intensity value. • A frequency distribution histogram of these pixel intensity values is called the “image histogram” and is one form of information in an image. • Does an image histogram represent spatial information from the image? • Analyze image histogram info  intensity threshold  identify objects as foreground vs. background. Advantage: • 100’s or 1000’s of images from one assay have variations in foregrounds and/or backgrounds, so histograms vary from image to image. • Flexible, systematic method of analyzing each can still  comparable object identification results amidst the variations. • Method: IsodataThreshold • Assumes bimodal histogram, determines T by iteratively computing threshold between means on left (mL) and right (mR) sides of histogram until T does not change any more. • Good for bimodal histograms (expect every image to have similar densities of foreground objects?) • What happens in fields that are sparse, or fluorescence very low? • What happens in fields with bright fluorescent outliers, debris, or artifacts? • Method: TriangulationThreshold • On image histogram, hypotenuse “drawn” from mode (tallest peak) to max intensity. • T is value on x-axis where longest normal (perpendicular line) between hypotenuse & histogram’s curve occurs. • Good for potentially sparse fields. • What are possible outcomes when fields are potentially dense? • What are possible outcomes when fields have bright fluorescent outliers (consider size of such objects, and “tightness” [kurtosis] of their peaks)? • Method: 3sigmaThreshold- calculated threshold is 3 standard deviations above the calculated Mean of the distribution. • Value: Shifts auto-computed threshold (Tcalc) according to equation below. Tfinal = Tcalcx (1+Value)

Common V3 Assay Params: ObjectSegmentation Image illustrates some caveats of ObjectSegmentation. Some of such debris segments can be Rejected if Object Selection Parameters are optimized, but not always. Also, other caveats less obvious include over-segmentation of a large or elongated single object. ObjectSegmentationJust when you must! ObjectSegmentation • Use it only when you must. • It’s an approximation, so subject to imperfections. • When must I use it? • When a prohibitive majority of cells are simply too crowded to identify individually, and individual identification is required. • e.g. let your biological assay needs guide you. • How else can I deal with clustered cells? • Seed wells to achieve lower final densities, and/or more appropriately dispersed cells (some tips are provided in Cellomics’ FAQs). • Use a different cell type, if possible (a handful of cell types are known for their tendencies to cluster; e.g., HEK293, HepG2, MCF7). • Reject them using Object Selection Parameters. In such a case, weigh… • [time to acquire more images] versus [risk of inaccurate Object Segmentation]; • whether Rejection of clusters introduces bias (e.g., target phenotype expressed only in clusters?). Compare results using [Object Selection Parameters that Reject all but single cells] versus [Rejection of only/mostly single cells]. • What Value should I use? • Radius (1/2 the diameter) of a typical Object. • Clusters of >2 or 3 cells may best be Segmented using a negative Value.

Caveats & benefits of different Object Identification Methods and Assay Parameters Images show caveats of Isodata (or PeakTriang or other histogram-based, dynamic) Object Identification Methods, as well as a caveat of the ObjectSegmentation Assay Parameter, and a benefit of the BackgroundCorrection Assay Parameter.

I-V-S in non-1° (2°) channel(s) • “Rejected_” overlay, usually*. • If “Rejected_” because of 1°Ch, • excluded from Valid_Count •  from Selected_Count •  from %Selected_ • If “Rejected_” because of 2°Ch, • included in Valid_Count • excluded from Selected_Count • reflected in %Selected_( = Selected_  Valid_Count) 2° channels(a.k.a. “downstream/dependent”, “ChN”) “1° object mask” “2°/sub-object masks” 1° Object Identification Object Identification(often + Assay Params) MaskModifierChN(Assay Parameter) 1° Object Validationincludes 1° Object Selection fail fail Sub-Object Selection(same list ) Object Selection(a.k.a. “gating”) “1° objects”with“1° masks” fail Selected_ no overlay overlay data results * BioApp-specific (a few cases have no “Rejected_” overlay).Also, overlay must be .

Interactive identification and Selection of Objects Protocol Interactive view after Identify Objects This is an interactive Object Selection mode: Clicking Identified objects makes them temporarily Selected, adding their values to the Object Selection Set Statisticstemporary table that can be transferred to the Object Selection Parameters that are applied to all objects each time you Run Algorithm. To transfer only a subset, click on the row(s) of the Statisticstable to toggle graying out. Restore Protocol Values is NOT an undo button!! Restores values from last saved version of the Protocol only! • Common mistakes • Notice scrollbars. Unseen Statistics may be transferred! • If zeroes transfer as Max for certain Parameters, all objects will be rejected! • Esp. common in BioApps with parent objects + “child” sub-objects: 1st one selected determines mode of accumulating only object or only sub-object Stats. • Statistics table is NOT CUMULATIVE! Many actions buttons clear it (incl. Run Algorithm, Identify Objects itself, Deselect…, Acquire…, etc.)

More FAQ’s Q: “Does my Form Factor let high field#s get outside my wells?” A: When you pass the borders specified by the selected Plate Type Form Factor, the displayed Field # goes blank. To check a corner of the well, select the max# in the list, then learn what corner it is by moving 1 field in each direction using the Select Field arrows until you find the 2 blanking directions. You can then use the arrows to step to the opposite corner where a blank appears, etc. Q: “I want to look at the same non-1 field# in many wells on my plate. Do I have to select it every time I change wells?” A: No. The remedy is illustrated on this page! • Movement by Select Field arrows auto-updates this #. • And/or, a field # can be selected from list to move there. • Starting Field Offset dialog changes the default value here, but must move out of well and back in to “freshen” changes made in that dialog.(This temporary interactive Starting Field Offset reverts to default 1 when (a) a new Plate Type Form Factor is selected, (b) each time the software is started, and (c) may also when an Assay Protocol is saved, (d) a scan is completed, or other actions, so monitor closely). 34

Plate Protocols Plate ID / Barcode • Plate Protocol • Remaining scan settings not in Assay Protocol: • Plate Type (Form Factor) • Scan Area (which wells on plate) • Starting Field Offset (within each well) • Store Images (saves space for quick tests) • Display More (reduced GUI graphics) Assay Protocol

Anatomy of a Generic Cell Ch 1 Downstream Channels Circ Spots Circ Ring Ring Spots Object Primary Object = Nucleus Special Design for Morphology Explorer Ch 1 Ch 2 Ch 3 or 4 Object SpotFibers (low LWR) Process Members SpotFibers (high LWR) Primary Object = Whole Cell

Masks and Modifications Mask Modifier + 2 Mask Modifier + 4 • Mask: the graphic representation of an object that is applied to separate channel in order to make a measurement in that channel • Mask Modification: the enlargement or shrinking of the original mask • Positive values for the Mask Modifier parameter enlarge the mask (dilation) • Negative values for the Mask Modifier parameter reduce the size of the mask (erosion) • Three major types of Mask Modifications: • Primary object mask (MaskModifier) • Circ Mask (CircModifier) • Ring Mask (RingDistance and RingWidth) • The value of the Modifier is the number of pixels to dilate (positive) or erode (negative) the mask • Adjustment of Mask and Circ Mask looks like this: Primary Object (Nucleus) Mask Modifier -2.0 Mask Modifier -4.0 Channel 1

Adjustment of Circ & Ring Regions • Ring Mask is described by 2 parameters: Ring Distance and Ring Width • Adjustment is always relative to the Primary Object Mask

APPENDIX: ArrayScan resolutions & field sizes

“I’m trying to scan fields at the edges of scan-able areas, but corners are outside the wells.” See text below, right. Pattern always counter-clockwise, but field#2’s position depends on well’s total scan-able field#: System uses width of field at selected magnification. to compute how many fields fit in Form Factor’s Well Height (= Well Width), yielding a square grid of fields. When total field# is odd, field#1 is center; when even, offset by half field. Largest square grid in circular well: Well’s diameter = hypotenuse of grid. Pythagorean theorem says Well Width = Well Height ~ 70% of diameter. On a micron scale, plates may vary widely from specs, even in the same lots. Seating on stage and temperature expansion/contraction also cause variation. If you have verified… system is calibrated, Form Factor’s First Well X&Y optimally approximates A1’s center, Well Pitch X&Y are optimally approximate across a plate (use Status Bar/Tray coordinates in Protocol Interactive noting X’s and Y’s for A1 vs. distal and corner wells)… …may be advisable to decrease Well Width & Height APPENDIX: Field Scan Patterns

Many terms in HCS have historical roots that are blends of photographic imaging and cell-based screening. Often the vocabulary works well, but is not a restriction on capabilities. Plate level features- E.g., exposure times, z-offsets, Reference Levels. This “plate” structure is rooted in cell-based assays in multi-well microtitre plates. In HCS, it can mean a microscope slide, too. Well level features- E.g., Z-position, CellCount, MEAN_ObjectAvgIntenCh2, SD_ObjectTotalInten, %HIGH_RingSpotCountCh2. A well is merely a subdivision of a plate. It constitutes a container for materials delivered to it. It can also be a subdivision of a micro-slide. Fields- A term borrowed from microscopy’s “field of view”. When a Field size is smaller than the image-able surface of a well, multiple fields may be imaged per well. Some BioApplications report Field features. All report Cell Features linked to Wells+Fields. Cell level features- E.g., ObjectSizeCh1, RingSpotCountStatusCh2. Historically, it was cells that were plated into wells of a plate. In HCS, the fundamental unit of analysis (primary Object) is most often a Cell, and multiple features of it are measured. Some of these features may be sub-objects (e.g., spots, micronuclei, perinuclear rings, neurites), though not necessarily. So, Cell Features are a class in this hierarchy, but Cell Features may correspond to properties of colonies or whole organisms (e.g., C. elegans). “Intrinsic/raw” Cell Features- e.g., RingSpotCountCh2 is a mere count of how many spots have been ascribed to an object based on sub-objects (spots) identified by Ch2’s label within a sub-compartment (ring) of an object. “Simple Statistical” Well Features based on intrinsic/raw Cell Features- e.g., MEAN_TargetAvgIntenCh2 represents that mean response of that well’s population of objects. That mean has a SD_TargetAvgIntenCh2 about it, too. ObjectCount is another example; it’s the sample size, n, on which the mean, stdDev, etc., are based. Status Cell Features- e.g., RingSpotCountStatusCh2. In some BioApplications, an object can be assigned to one of some Cell Feature phenotypes. For example, if an object’s TargetIntenCh2 exceeds the corresponding Reference Level, it’s classified by TargetIntenStatusCh2=1; else =0. Population Characterization Well Features based on Status Cell Features- Rather than a metric of a mean response, these report the fraction of a well’s sampled population that fall in a phenotypic class. E.g., %HIGH_TargetAvgIntenCh2 reports what percentage of the well’s sampled population exceeded a minimum Response Level. It’s simply computed by tabulating the corresponding Status frequencies. The meaning & legitimacy of intrinsic/raw features depends on the quality of input settings that generate them. The same is true of Population Characterization features. the %HIGH value is not meaningful if relevant Reference Level inputs were not used. Appendix: Data Hierarchy and Tiers

Quiz 1 Solution

Helen Ho Field Application Scientist, West Cell: 415-902-8734 Email: helen.ho@thermofisher