High Content Imaging and Flow Cytometry with Phaedra

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
HighContent
Analysis,
LIMS-integrated screening tool.
Can start from images, or
from images + numerical data.
End point is compound result,
single-well or dose-response.
Evaluation
&
Dynamic
Reliability
Assurance
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
Java language
• Permits cross-platform deployment.
• Developer familiarity.
• Multiple implementations available.
OSGI dynamic
module system
• 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
• Highly efficient image compression.
• Fast extract Images on different ROI/Scales/Quality.
Oracle database
• 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
4
HCA : Systems & Process Interconnections
Plate Definition & Preparation : TiboLims, Laplace, …
Experiment
Definition
Plate Preparation
Plate
Definition
ABCD
FEDS
Data
Warehouse
Search &
Vizualization
PDF files
PHAEDRA
Plate Scan
Images
Import
Link
Create &
Edit
Protocols
Search,
Report,
Print
Validation
&
Approval
Archival
Metadata
Plate Scan:
Mias-2, Opera, Incell,
& Analysis:
Dci, Matlab, A
Capella,…
Cell + Well
Data &
Overlays
Image
Analysis
Heatmaps
Charts
Visualization
XX50 calcs
Rules, Features, Preferences, ..
Yellow boxes: can be done in or initated from PHAEDRA
Images,
Plate, Cell
&Well
Annotation
PHAEDRA DB : Main Entity Definitions
Protocol Classes
Neo Cell
Line ABC
Protocols (e.g. diff cell lines)
Experiments (runs, batches,
Plates
Plate 36570152
Wells
Well 1
Compound,
Concentration
Batchname 2
Well 2
Well 3
Images
Raw Image
Channels
Features-Kinetics
Cell Results
Overlay
Result
Images
Mean Area, length,
perimeter, intensity
Well
Annotations
Batchname 3
Plate 36584432
Well Results
Image Types
Annotations
Neo Cell Line
DEF
Batchname 1
plate stack)
Well Entities
Drill down
Neo
Area, length,
perimeter, intensity
Cell
Annotations
PHAEDRA UI: Extensive, Flexible GUI based on RCP + OSGI, an
Industry-Standard Open Source Framework
Consists of…
Perspectives can co-exist
User can have
multiple, cooperating
perspectives open at
the same time
Assemble views and
create own perspectives
One perspective, 3 views
All Views
can Talk to
each Other
Eclipse/RCP Framework
Dockable
GUI
Customizable
perspectives
45+
Views
Univariate
Tools
Multivariate
Tools
8
PHAEDRA: general GUI Look and Feel
 Perspective
open/switch
E
X
P
L
O
R
E
R
Heatmaps
charts
Multiple
Views
User & data source
indicator/selector:
DVL, PPR, PRD
Images
Image Detail
Subsystem
RAM usage indicator &
Traffic Light
Garbage Collector
Health Indicator
Dynamic Drill-Down from Compound to Cell
DR Curve
Plate
Heat Map
Scatter Plot
of Cells
Well
Spider
Every Open View reacts immediately to all actions/selections
10
Quality Control & Annotation
• 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)
11
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
12
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
PHAEDRA: Protocol – FEATURE Configuration
 For calculated (derived) features
Each aspect of each
FEATURE can be fully
and individually
configured
 For dose-response curves
PHAEDRA: Protocols: IMAGE Configuration
Extensive,
Comprehensive:
Each aspect of each
Image Channel can be
fully and individually
configured
17
PHAEDRA: Analysis & Import: in & outputs
2 : Select output Protocol
(defines the analysis)
1 : drag
input folder
3 : go to the analysis settings
PHAEDRA: Analysis & Import: settings
Experimental cluster support
check Settings (optionally)
Settings are saved with the data
PHAEDRA: Analysis & Import: preparation/test
Preparation: Input Folders are “validated”
Test a few
wells or do
Full Analysis of
all plates in
this
experiment
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
Each protocol has its own import +compression settings
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
PHAEDRA: Show/Add Well & Cell Annotations
Cell rejection & Well
recalculation
PHAEDRA: Data Visualization
Protocol
Experiment
Plate
Statistics, Histograms, Scatter, Trends, Spiders, … , for Experiments, Plate, Well & Cell data
PHAEDRA: Data Visualization
Hide/show , Maximize Views
Multiple plate Info Tabs for each Plate
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
Column Grouping
Display: 1200 compounds/7200 curves
Original state = unchecked
Validated state
Invalidated state
Invalidated & Disapproved compounds will NOT be uploaded to the Data Warehouse
Freezing Columns
Images & data are linked : up to individual cell/organell
Bi-directional hightlighting
PHAEDRA: Search & Report (1)
User can Combine any number
of search Rules, Save searches
for later use, Drill down in
Collections, etc.
31
PHAEDRA: Search & Report (2)
Everything that can be shown
on screen can be put in
a PDF Report
EXAMPLES: 3D Tumor Colony analysis
Colony Size
Max. Intensity Projection Image
Max. Intensity Projection Images
Colony Location
EXAMPLES : 3D Tumor Colony analysis
Colony Size
Position & Size
Quality Control:
Uniformity of colony location
EXAMPLES: 3D Neurite & Spine Analysis
Thumbnails
Effect of
Pilocarpine
Max. intensity Projection
EXAMPLES: 3D Neurite & Spine Analysis
Multiple Real-time
Interactive 3D Views
ICS: ~ 60 K / image
Epithelial -Mesenchymal Transition Assay (oncology)
Multi-channel
fluorescence:
Show arbitrary
image & overlay
combinations,
set transparency,
Stretch individual
channels
Same file, different channels
3D
Phaedra as a multi-format, multi-dimensional image viewer
3D : slices
LSM
ICS
JPG
TIFF
PNG
NIFTI
Analyze
GIF
BioRAD
…
3D-Dicom
Summary - PHAEDRA:
50 Linked Views
12 Linked Chart Types
39
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
cell/well data as attachments
Interactive !
JP2 in PDF: Best of Several Worlds:
• Highest Compression
• Standard Archival of Image + Data
• Interactivity
PHAEDRA: Development
Ari de Groot, Frederick Michielssen, Pieter van Nuffel, Michael Geerts
PHAEDRA: Future
• Extend Image Analysis Toolkit for new protocols
• Finalize Cell-level Rejection & Recalculation
• Kinetics (see below) & Flow Cytometry (next slide)
End
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)