High Content Imaging and Flow Cytometry with Phaedra: can ......PHAEDRA: Search & Report (1) User...
Transcript of High Content Imaging and Flow Cytometry with Phaedra: can ......PHAEDRA: Search & Report (1) User...
High Content Imaging and Flow Cytometry with Phaedra: can we integrate with Knime?”
Frans Cornelissen
Informatics CoE, Janssen Pharmaceutical Companies of
Johnson & Johnson
What’s in a name…
Plate-based
High-
Content
Analysis,
Evaluation
&
Dynamic
Reliability
Assurance
LIMS-integrated screening tool.
Can start from images, or from images + numerical data.
End point is compound result, single-well or dose-response.
Strong focus on user interaction, quality control functions and annotation.
GOALS of PHAEDRA
• Seamlessly integrated data access.
– Images and numerical data.
– Cell level, well level, and plate level data with full drill-down capacity.
– Full experimental context, including e.g. compound structures.
– Dose-Response Curve Fitting
– Upload to Data Warehouse
• Data hub between different systems.
– All types of HCS imagers.
– Experimental setups from different LIMS environments.
– Capable of accepting data from other [imaging-analysis] equipment.
• Bring data analysis to a higher level.
– Highly interactive, customizable environment.
– Statistical tools for univariate & multivariate data analysis.
– Full QC & Annotation options.
– Search & Report facilities.
Architecture & Design
4
Java language
Oracle database
OSGI dynamic module system
• Permits cross-platform deployment.
• Developer familiarity.
• Multiple implementations available.
• Highly efficient image compression.
• Fast extract Images on different ROI/Scales/Quality.
• Rapid incorporation of new tools.
• Dynamic management of modules and updates.
Eclipse/RCP app framework.
• Highly configurable user environment.
• Add functionality from BioClipse, OpenChrom, …
JPEG 2000 images
• Database links with LIMS systems.
OpenGL, ICS, R, MATLAB, Javascript, PDF, JFreeChart, …
(thanks to these standards/Technologies) most of PHAEDRA was developed in ~ 12 months
Experiment Definition
Plate Definition & Preparation : TiboLims, Laplace, …
Plate Definition
Plate Scan
Plate Scan: Mias-2, Opera, Incell,
& Analysis: Dci, Matlab, A
Capella,…
Metadata
Plate Preparation
Images
Image Analysis
Cell + Well Data &
Overlays
ABCD FEDS
PHAEDRA
Import Archival Validation
& Approval
Search, Report,
Heatmaps Charts
Visualization XX50 calcs
Images, Plate, Cell
&Well Annotation
Link
Data Warehouse
Search & Vizualization
PDF files
Rules, Features, Preferences, ..
Create & Edit
Protocols
HCA : Systems & Process Interconnections
Yellow boxes: can be done in or initated from PHAEDRA
Annotations
Features-Kinetics
Image Types
Well Entities
Wells
Plates
Experiments (runs, batches, plate stack)
Protocols (e.g. diff cell lines)
Protocol Classes Neo
Neo Cell Line ABC
Batchname 1
Plate 36570152 Plate 36584432
Well 1 Well 2
Compound, Concentration Well Results
Mean Area, length, perimeter, intensity
Well Annotations
Images
Raw Image Channels
Overlay Result Images
Cell Results
Area, length, perimeter, intensity
Cell Annotations
Well 3
Batchname 2
Neo Cell Line DEF
Batchname 3
PHAEDRA DB : Main Entity Definitions
Drill down
PHAEDRA UI: Extensive, Flexible GUI based on RCP + OSGI, an
Industry-Standard Open Source Framework
Consists of…
All Views can Talk to each Other
Assemble views and create own perspectives
One perspective, 3 views
Perspectives can co-exist
User can have multiple, cooperating perspectives open at
the same time
Eclipse/RCP Framework
8
Dockable GUI
45+ Views
Customizable perspectives
Univariate Tools
Multivariate Tools
PHAEDRA: general GUI Look and Feel
Perspective open/switch
User & data source indicator/selector:
DVL, PPR, PRD
Subsystem Traffic Light
Health Indicator
RAM usage indicator & Garbage Collector
Multiple Views
Heatmaps
charts
Images
Image Detail
EXPLORER
Dynamic Drill-Down from Compound to Cell
10
DR Curve Plate Heat Map
Well Spider
Scatter Plot of Cells
Every Open View reacts immediately to all actions/selections
Quality Control & Annotation
11
• Strong focus on supporting interactive QC of the HCS data set.
• Users can add annotations to:
• Wells
• Cells
• Regions on images
• Invalid wells can be excluded:
• By selecting wells manually.
• By defining filter criteria, (RULES) e.g. on number of cells detected.
• By selecting wells on well scatter plots.
• Exclusion of data points triggers [automatic] recalculation.
• Validation (User), Approval(Manager)
12
PHAEDRA: Validation: Well Selection Rules Queries: e.g. Indicate all well images where #branches > 30
Efficient : • Can apply any rule to select wells, then invalidate all selected • Rules can be stored / reused
Advanced Data Analysis Tools Built-In
• MATLAB integrated to support advanced Image Analysis
3D Colony Formation 3D Neurites & Spines
Analysis of low-resolution, whole-well brightfield Z-stack.
High-resolution confocal Z-stack.
Advanced Data Analysis Tools Built-In
• R engine for statistics and dose-response curve analysis – Supports rejection of outliers or selection of an alternative model.
– Multiple measured features are fitted in parallel.
• Javascript tool – Data transformations to generate ‘Calculated Features’.
PHAEDRA: Protocol Class Configuration
Each aspect of each
Protocol can be fully and individually
configured
Users/ Roles / Access Rights
Each aspect of each
FEATURE can be fully and individually
configured
PHAEDRA: Protocol – FEATURE Configuration
For calculated (derived) features
For dose-response curves
17
PHAEDRA: Protocols: IMAGE Configuration
Extensive, Comprehensive:
Each aspect of each
Image Channel can be fully and individually
configured
Other config panels
PHAEDRA: Analysis & Import: in & outputs
1 : drag input folder
2 : Select output Protocol (defines the analysis)
3 : go to the analysis settings
PHAEDRA: Analysis & Import: settings
check Settings (optionally)
Experimental cluster support
Settings are saved with the data
PHAEDRA: Analysis & Import: preparation/test
Test a few wells or do
Full Analysis of all plates in
this experiment
Preparation: Input Folders are “validated”
PHAEDRA: Analysis & Import: execution & import
analysis: all wells are analyzed ; runs MULTITHREADED in background while continue other work
There is also an import-only perspective, to integrate 3th party analysis results & images
PHAEDRA: Image Visualization
Complete : Linked Hierarchical Image views from
PLATE to WELL to CELL
PHAEDRA: JPEG 2000 Images, JPX format
JP2 compression: • best compression for natural imagery •Less artefacts, very flexible&performant file format
JPX format: • multiple layers stored in 1 file • each layer is optimally compressed
Enables fast random access and on the fly Generation of arbitrary sized thumbnails
Each protocol has its own import +compression settings
PHAEDRA: Show/Add Well & Cell Annotations
Cell rejection & Well recalculation
PHAEDRA: Data Visualization
Statistics, Histograms, Scatter, Trends, Spiders, … , for Experiments, Plate, Well & Cell data
Protocol Experiment Plate
PHAEDRA: Data Visualization
Multiple plate Info Tabs for each Plate
Hide/show , Maximize Views
Cell Histograms & Statistics
• All views from all perspectives can be combined at will • Views can be placed and scaled as needed • Users can save any combination as new personal perspectives
Fitting
Reject outliers & refit
Batch Compound Validation/Approval
Invalidated & Disapproved compounds will NOT be uploaded to the Data Warehouse
Original state = unchecked
Display: 1200 compounds/7200 curves
Invalidated state
Validated state
Column Grouping
Freezing Columns
Images & data are linked : up to individual cell/organell
Bi-directional hightlighting
31
PHAEDRA: Search & Report (1)
User can Combine any number of search Rules, Save searches
for later use, Drill down in Collections, etc.
PHAEDRA: Search & Report (2)
Everything that can be shown on screen can be put in
a PDF Report
EXAMPLES: 3D Tumor Colony analysis
Max. Intensity Projection Images
Colony Size
Colony Location
Max. Intensity Projection Image
Quality Control: Uniformity of colony location
EXAMPLES : 3D Tumor Colony analysis
Colony Size Position & Size
EXAMPLES: 3D Neurite & Spine Analysis
Max. intensity Projection
Thumbnails
Effect of Pilocarpine
EXAMPLES: 3D Neurite & Spine Analysis
ICS: ~ 60 K / image
Multiple Real-time Interactive 3D Views
Multi-channel fluorescence:
Show arbitrary
image & overlay combinations,
set transparency, Stretch individual
channels
Epithelial -Mesenchymal Transition Assay (oncology)
Same file, different channels
3D
3D-Dicom 3D : slices LSM ICS JPG TIFF PNG NIFTI
Analyze GIF
BioRAD …
Phaedra as a multi-format, multi-dimensional image viewer
39
Summary - PHAEDRA: 50 Linked Views
12 Linked Chart Types
PHAEDRA: >150 Context Actions (only highest level shown)
Intuitive, Extensive All actions
available as context menu
items
PHAEDRA: >100 Preferences
Extensive: About Everything that can be shown on screen can be configured - customised
Flexible: Centrally Stored per User ; User can log in on any PC and reuse his preferences
FEDS: JP2 Plate Images,Overlays & Data in PDF
Interactive !
JP2 in PDF: Best of Several Worlds: • Highest Compression • Standard Archival of Image + Data • Interactivity
cell/well data as attachments
PHAEDRA: Development
PHAEDRA: Future • Extend Image Analysis Toolkit for new protocols • Finalize Cell-level Rejection & Recalculation • Kinetics (see below) & Flow Cytometry (next slide)
End
Ari de Groot, Frederick Michielssen, Pieter van Nuffel, Michael Geerts
Flow Cytometry in PHAEDRA: R & Knime
FCM == workflow ! Compensation Transformation Gating strategies Counting
Need to show data & scatterplots at different stages in the gating, HilLite subpopulations in all plots
=> KNIME!
produce high-quality graphics:
R + Reporting
FCM in PHAEDRA: R & Knime
Planning: • develop nodes to :
• read/write • FCS files (FCS 3.1 spec) • gate files (Gating-ML spec) • ACS files (archiving in FCM, ZIP +TOC.xml contains all data files of 1 plate)
• automatically/ interactively define rectangular/polygon gates on scatterplots Lots of Questions: • can we integrate the Knime application with Phaedra ? • can we execute a WF iteratively on all wells of a plate (headless – no UI) • can we execute all R calls within a single (static, persistent) instance (in order to reduce data transport back & forth)