A talk on high-throughput RNai and compound screening data analysis given at Finnish Institute for Molecular medicine (FIMM) March 13, 2008, 12.30—17.30.
1. VTT MEDICAL BIOTECHNOLOGY
Analysis of HTS data
FIMM & Biomedicum Medical Bioinformatics Day,
March 13, 2008, 12.30—17.30
Pekka Kohonen
VTT Medical Biotechnology
FIMM
2. VTT MEDICAL BIOTECHNOLOGY
Presentation overview
1. The high-throughput screening workflow
2. Design considerations in the screens
• Which genes to assay: biological question at hand
• Sources of error in the screens:
• Biological/technical variance (negative controls)
• Transfectability of the cells (positive controls)
• Off-target effects (redundancy and replication)
3. RNAi screening data normalization
4. Hit picking and prioritization
5. New technologies: Cell Arrays and Lysate Arrays
6. Integration of data from other sources
7. Hight Throughput screening database (HTSdb)
• Combines multiple assays and platforms
• plate based, lysate arrays, cell arrays, supporting data(GE, aCGH)
• Based on R/MySQL
• quot;First Lightquot; recently
3. VTT MEDICAL BIOTECHNOLOGY
Screening work-flow
Biological question
Reagents: Libraries of siRNAs, miRNAs,
Biological assay
compounds
Primary screens
Replicating hits
Data integration with gene expression, Investigation of pathways targeted,
aCGH, other screens (cancer/normal) literature mining
Secondary screening
Prioritized hits for further validation
4. Flow-through of a High-throughput screen
VTT MEDICAL BIOTECHNOLOGY
in 384 wells
2) Add transfection
agent
1) Pipet diluted siRNAs
3) 35 ul of trypsinized
cell suspension
384 well plates
4) Incubate
72 hrs
5) Add cell phenotype
stains & incubate
6) Fluorescence
measurement
& data analysis
5. VTT MEDICAL BIOTECHNOLOGY
Design considerations: Off-target effects
• Non-sequence
specific off-target
effects:
– Interferon
response
– siRNA causing
miRNA
machinery
saturation
– Lipid toxicity
• Specific:
– Effects on
related mRNAs
– miRNA
mechanism
based off-target
effects
Off-target effects are usually cell line and siRNA specific
The best way to mitagate them is to have 2-4 siRNAs per gene
6. VTT MEDICAL BIOTECHNOLOGY
RNAi screening data normalization
Edge-effects and B-score normalization
Raw data showing an B-score normalized data
edge effect after removal of the edge
effect
• Edge effect is seen especially with the Cell Titer Blue (CTB) reagent
• Edge effect causes a lowered signal intensity at the edges
• In the B-score normalization estimates of row/column effects are obtained
using a two-way median polish. (Brideau et al., J Biomol Screen. 2003)
7. VTT MEDICAL BIOTECHNOLOGY
Functional screens are used to define the effects of the
siRNAs on cell proliferation
Raw data
CTB
Normalised
data
Cell proliferation hits from the
screens
8. VTT MEDICAL BIOTECHNOLOGY
In red: siRNAs that cause growth inhibition
3
Cell Line 1 Cell Line 2
3
2
2
1 1
0
Z score: growth inhibition
0
0 50 100 150 200 250 300 350 400 450
0 50 100 150 200 250 300 350 400 450 -1
-1
-2
-2 -3
-4
-3
-5
-4
-6
-5 -7
3
(Z score: Growth inhibition)
2
Cell Line 2
1
0
-5 -4 -3 -2 -1 0 1 2
-1
-2
-3
-4
-5
-6
-7
Common Anti-
proliferative hits Cell Line 1
(Z score: Growth inhibition)
9. VTT MEDICAL BIOTECHNOLOGY
Cell Titer Blue (CTB) growth inhibition screens (Blue means growth inhibited)
siRNAs hitting
preferentially the parent
cell line
siRNAs hitting the
variant_1 cell line
siRNAs hitting the
parental cell line
Pan-hitting siRNAs
Parental Variant_1 Variant_2 by Pasi Halonen
10. VTT MEDICAL BIOTECHNOLOGY
I TECHNOLOGY INTRODUCTION - TRANSFECTION CELL ARRAYS
• Up to 46 000 spots with different individual siRNA transfections in single assay plate.
• Arrays with cells growing only on arrayed spots.
• System allows low cost uHTS with minimal infastructure requirements.
• Has five measurement channels for visualization of different antibodies and stains
by Juha Rantala
11. VTT MEDICAL BIOTECHNOLOGY
Image analysis will be a bioinformatics challenge for the
cell array technology
1. Imaging 2. Automated image analysis
• image based cytometry
10,000s of
images from
each experiment
- requiring
terabytes for
storage
• Analysis of antibody staining/ organelle stains
DNA ACTIN Antibody 1. Antibody 2. + Antibody 3. ?
3. Result classification by morphology, intensity, localisation, number etc.
12. VTT MEDICAL BIOTECHNOLOGY
II Cell lysate microarrays for multiple end-point analysis
Protein lysates Pre-miR transfections
siRNA transfections
Multiple protein Lysates from cultured cell lines
microarray slides
Phenotype markers
Proliferation: Ki-67, Cyclin E, Histone H3
Apoptosis: Caspase-3, PARP, Histone H2AX
Cell cycle: Cyclins D, E, A, B1, p-HistoneH3(Ser10)
EMT: E-cadherin, Vimentin, Beta-catenin
Targets & pathways: p53, c-Myc, Met
by Rami Mäkelä
Signal quantification and analysis of functional effects
13. VTT MEDICAL BIOTECHNOLOGY
Integration of data from other sources
Two cell lines: GE+siRNA One cell line: GE+siRNA+aCGH
sirNA growth inhibition difference
Expression ratio to parental
Gene amplification, siRNA
Increased gene expression growth inhibition and gene
and greater siRNA growth expression increase
inhibition
by Henrik Edgren
14. VTT MEDICAL BIOTECHNOLOGY
High Throughput Screening Database:
Multiple Assays of the same Model System
Plate based: HTSdb Lysate arrays:
- CTB - up to 3 channels
- CellTiter-Glo™ - multiple endpoints
- ApoOne™ - use of ratios
- luciferase assays
Supporting Data:
Cell Arrays: - gene expression
- up to 5 channels
- uHTS (10000's) - aCGH
- improved repeatability - miRNA expression
- use ratios for normalization
15. VTT MEDICAL BIOTECHNOLOGY
HTSdb Design Principles
• Pragmatic - focused on analysis needs
• Extensible to new data sources, normalizations and sample
annotation terms
• Different assays done on same biological samples can be
combined (eg. CTB, ApoOne, Lysate Arrays)
• Other data sources (gene expression, miRNA expression)
can be combined with screening datas
• MySQL open source database
• R statistical programming language is used to access the
database and to analyze the datas
• Bioconductor R-libraries are used when applicable
• Ensembl: all identifiers are linked to ensembl genes
quot;First Lightquot; recently - data input,
normalization and retrieval
16. VTT MEDICAL BIOTECHNOLOGY
Database Structure
Annotations of reagents
siRNA, miRNA, compouns
Datas: raw and normalized
Screen Annotations
17. VTT MEDICAL BIOTECHNOLOGY
VTT Medical Biotechnology, Turku, Finland CONFIDENTIAL
Canceromics
• Matthias Nees
• Elmar Bucher
• Henrik Edgren
• Kalle Ojala
• Sami Kilpinen
Biochips • John-Patrick Mpindi
John-
High-throuput screening
• Petri Saviranta • Tommi Pisto
• Rami Mäkelä • Merja Perälä
kelä • Pekka Tiikkainen
• Juha Rantala • Pekka Kohonen
• Arttu Heinonen • Henri Sara
• Niko Sahlberg • Maija Wolf
• Pasi Halonen
• Suvi-Katri Leivonen
Suvi-
Harri Siitari
• Saija Haapa-Paananen
Haapa-
• Vidal Fey
Olli Kallioniemi