SLIDE 1
- utline
Pre-amplification critically analysed Jo Vandesompele professor, - - PowerPoint PPT Presentation
Pre-amplification critically analysed Jo Vandesompele professor, - - PowerPoint PPT Presentation
Pre-amplification critically analysed Jo Vandesompele professor, Ghent University co-founder and CEO, Biogazelle Advanced Methods in RNA quantification (London, UK) May 21, 2009 outline introduction 450 miRNA pre-amplification
SLIDE 2
SLIDE 3
pre-amplification – the one and only quality criterion
preservation of differential expression (fold changes)
before (B) and after (A) sample pre-amplification
[no introduction of bias]
(G1S1)B/(G1S2) B = (G1S1) A/(G1S2) A G1B/G2B < > G1A/G2A gene G, sample S, before B, after A
SLIDE 4
pre-amplification – the scene
pre: before actual qPCR amplification: make large amounts of RNA/(c)DNA from limited input
single cell – picograms – nanograms >> micrograms
transcriptome wide
all long RNA molecules advantages
- no prior knowledge of target genes is needed
- study can grow
disadvantages
- somewhat more expensive
focused pre-amplification
predefined set of sequences advantages
- fast and simple
disadvantages
- all targets need to be known in advance
SLIDE 5
pre-amplification – the players
transcriptome wide
Eberwine method
- T7-RNA polymerase based in vitro transcription > antisense RNA
- home brew protocols
SMART method (Clontech)
- template switch mechanism sense RNA
- T7-RNA in vitro transcription
Phi29 based
- rolling circle amplification – strand displacement
- cDNA
SPIA technology (NuGEN)
- hybrid RNA/DNA SPIA primer
- cDNA
focused pre-amplification
limited cycle PCR (10-14 cycles)
SLIDE 6
5’ 3’ 5’ 3’ 5’ 5’ 3’
AAAAA UUUUU UUUUU AAAAA AAAAA TTTTT
1 streng cDNA-synthese (met T7-oligo(dT) primer)
ste
1 streng cDNA-synthese (met random hexameren)
ste
2 streng cDNA-synthese
de
antisense RNA-amplificatie (T7 RNA polymerase)
TTTTT TTTTT
5’ 3’ 3’ 5’ 3’ 5’ 3’ 5’ 5’ 3’
AAAAA TTTTT
5’ 5’ 3’ 3’ 5’
UUUUU
antisense RNA-amplificatie (T7 RNA polymerase)
3’
2 streng cDNA-synthese (met T7-oligo(dT) primer)
de
NNNNNN NNNNNN
5’ 3’
1 ronde
ste
2 ronde
de
Ebermine method
SLIDE 7
SMART
SLIDE 8
SPIA
SLIDE 9
microRNA pre-amplification
stem-loop megaplex reverse transcription using 20 ng total RNA limited-cycle pre-amplification (14) qPCR profiling 450 miRNAs and controls higher sensitivity minimal amplification bias (Mestdagh et al., Nucleic Acids Research)
SLIDE 10
1 2 3 4 5 6 7 8 10 15 20 25 30 35 0.5 1 1.5 2 2.5 10 15 20 25 30
- 0.6
- 0.5
- 0.4
- 0.3
- 0.2
- 0.1
- 1.4
- 1.2
- 1
- 0.8
- 0.6
- 0.4
- 0.2
1 2 3 4 5 6 7 8 10 15 20 25 30 35
Average CqNP (NBL-S, IMR-32) Average CqNP (NBL-S, IMR-32) ∆∆Cq (|∆CqNP - ∆CqP|) NBL-S, IMR-32 Average ∆∆Cq ∆∆Cq (|∆CqNP - ∆CqP|) NBL-S, IMR-32 Average ∆∆Cq
minimal pre-amplification bias
SLIDE 11
5 10 15 20 25 30 2 4 6 8 10 12 14 5 10 15 20 25 30 35 2 4 6 8 10 12 14 5 10 15 20 25 30 2 4 6 8 5 10 15 20 25 30 35 2 4 6 8
Cq-value Cq-value
Cq-value Cq-value
total cell number total RNA input (pg) miR-18a R2 = 0.975 miR-20b R2 = 0.993 miR-92 R2 = 0.998 miR-19a R2 = 0.996
A B
1 2 4 8 16 32 64 128 1 2 4 8 16 32 64 128 0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14 30 25 20 15 10 5 30 25 20 15 10 5 35 30 25 20 15 10 5 35 30 25 20 15 10 5
single cell profiling
total cell number total RNA input (pg)
SLIDE 12
Mestdagh et al., Nucleic Acids Research, 2008
SLIDE 13
- utline
background research & goals neuroblastoma prognostic marker selection study design and workflow
RNA quality control sample pre-amplification normalization
data-analysis and results
SLIDE 14
biomarker signature based stratification
SLIDE 15
biomarker signature based stratification
SLIDE 16
aim
development and validation of a robust prognostic gene signature for
neuroblastoma using real-time qPCR
identifying patients with
increased risk in the current low risk and high risk group good molecular signature in the current high risk group
better choice of risk-related therapy
SLIDE 17
neuroblastoma
most frequent extra-cranial solid tumor in
children
originates from primitive (immature)
sympathetic nervous system cells
1:100,000 children (< 15 years)
20 cases/year Belgium | 700 cases/year
USA
15% of childhood cancer deaths prognosis is dependent on
tumor stage (localized vs. metastatic
disease)
age at diagnosis (< or > 1 year) genetic defects: amplification MYCN, ploidy,
loss of 1p, gain of 17q
SLIDE 18
prognostic classification
misclassifications resulting in overtreatment or undertreatment need for additional tumor-specific prognostic markers current microarray gene expression studies
data overfitting unstable gene lists lack of overlap biological & technical noise much more genes than samples probe annotation / platform different risk definition different data processing and analysis
SLIDE 19
- meta-analysis of 7 published microarray gene
expression studies
- literature screening of almost 800 abstracts from
single-gene studies
selection of a top ranking list of 59 prognostic markers
- two PCR-based assays
- capillary gel electrophoresis (Experion)
RNA quality control 423 samples sample pre-amplification (WT-Ovation) analysis of 366 primary untreated neuroblastoma tumours using real-time qPCR
- Prediction Analysis of Microarrays
- Kaplan-Meier
- Cox proportional hazards
data-analysis
study workflow
SLIDE 20
towards real-time PCR signature profiling
100 ng total RNA
30 ng quality control 10 ng unbiased amplification
WT-Ovation (NuGEN)
PCR assay design and validation
sensitivity, specificity and efficiency
RTPrimerDB (Pattyn et al., 2006, NAR; Lefever et al, 2009, NAR)
absolute standards
real-time PCR using 384-well format
sample maximization strategy
(Hellemans et al., Genome Biology, 2007)
366 tumors and 1 gene/plate
SLIDE 21
WT-Ovation reproducibility
10,00 15,00 20,00 25,00 30,00 35,00 ACTB RPL13A 18S YWHAZ B2M GAPDH UBC HPRT1 SDHA HMBS mean Cq (n=3) genes
Stratagene cell line A cell line B cell line C
mean of 5, 15 and 50 ng of total RNA amplified
SLIDE 22
WT-Ovation – no amplification bias
median bias = 0.36, 90%tile bias = 0.61
10 20 30 40 50 60 70 80 90 100 0,00 0,10 0,20 0,30 0,40 0,50 0,60 0,70 0,80 0,90 1,00
cumulative distribution bias (Cq)
SLIDE 23
WT-Ovation – no amplification bias
median bias = 0.36, 90%tile bias = 0.61
10 20 30 40 50 60 70 80 90 100 0,00 0,10 0,20 0,30 0,40 0,50 0,60 0,70 0,80 0,90 1,00
cumulative distribution bias (Cq)
no need for DNase treatment no need for cleanup of amplified products
SLIDE 24
preservation of differential expression
SLIDE 25
qPCR reproducibility
5 10 15 20 25 30 35 40 50 100 150 200 250 300 350 400
10 # 100 # 1,000 # 10,000 #
within a 384-well plate: 4 x 96 replicates
SLIDE 26
qPCR reproducibility
between two identical 384-well plates maximum ΔCq: 0.45
15 20 25 30 35 40 15 20 25 30 35 40
SLIDE 27
synthetic control
55 nucleotides PAGE purification blocking group 5 points dilution series: 15 molecules > 150.000 molecules
RCRP absolute standards stuffer FP
SLIDE 28
absolute standards
reproducibility across master mixes (5) and instruments (2)
5 10 15 20 25 30 35 1000000 100000 10000 1000 100 10 MM1 MM2 MM3 MM4 MM5
SLIDE 29
absolute standards cross lab comparison
5 standards (triplicates) 5 reference genes + 5
- ther genes
366 samples
SLIDE 30
5 standards (triplicates)
absolute standards cross lab comparison
average ΔCq standards correction Cq samples Cq qPCR instrument 1, mastermix 1 Cq qPCR instrument 2, mastermix 2
16 18 20 22 24 26 28 30 32 34 36 16 18 20 22 24 26 28 30 32 34 36
SLIDE 31
ARHGEF7 gene
366 samples use of 5 standards (triplicates) for correction
absolute standards cross lab comparison
Cq qPCR instrument 1, mastermix 1 Cq qPCR instrument 2, mastermix 2
16 18 20 22 24 26 28 30 32 34 36 16 18 20 22 24 26 28 30 32 34 36
SLIDE 32
SPUD assay (Nolan et al, 2006): detection of inhibitors Computed gel analysis (Experion, Biorad): evaluation of total RNA quality 5’-3’ assay (HPRT1): evaluation of mRNA integrity
rigorous control of RNA quality
423 primary untreated NB (100 ng total RNA)
30 ng
366 RNA samples
SLIDE 33
differences in reference gene ranking between intact and degraded RNA
(Perez-Novo et al., Biotechniques, 2005)
impact of RNA quality on expression stability
SLIDE 34
RNA quality parameters
2 4 6 8 10 10 20 30 40 50 5 10 15 20 20 40 60 80 5 10 15 20 25 20 40 60 80
RQI 5’-3’ dCq AluSq Cq frequency
SLIDE 35
delta-Cq 5’-3’ vs. RQI
RNA samples ordered by average rank (good -> worse)
- 50
50 100 150 200 250 300 350 400 450 50 100 150 200 250 300 350 400 450 RQI delta-Cq
SLIDE 36
RNA quality control & sample selection
423 samples:
4 samples < DOT/DOO without event 5 samples < presence of enzymatic inhibitors (SPUD) 20 samples < lack of mRNA integrity (no ΔCq 5’-3’)
- 12/14 failed WT-Ovation
- all low RQI values
28 samples < poor RNA quality (RQI + ΔCq 5’-3’)
366 best samples (86.5 %)
RQI:
- average = 7.4
- median = 7.6
- 90%-tile > 6.1
ΔCq 5’-3’:
- average = 2.36
- median = 2.06
- 90%-tile < 4.75
SLIDE 37
normalisation using geNorm technology
framework for qPCR gene expression normalisation using the reference
gene concept:
quantified errors related to the use of a single reference gene
(> 3 fold in 25% of the cases; > 6 fold in 10% of the cases)
developed a robust algorithm for assessment of expression stability of
candidate reference genes
proposed the geometric mean of at least 3 reference genes for
accurate and reliable normalisation
Vandesompele et al., Genome Biology, 2002
SLIDE 38
http://medgen.ugent.be/genorm
geNorm software
automated analysis
ranking of candidate reference genes according to their stability determination of how many genes are required for reliable normalization
SLIDE 39
0.003 0.006 0.021 0.023 0.056 NF4 NF1
cancer patients survival curve
statistically more significant results
geNorm validation (I)
log rank statistics Hoebeeck et al., Int J Cancer, 2006
SLIDE 40
mRNA haploinsufficiency measurements
accurate assessment of small expression differences
geNorm validation (II)
Hellemans et al., Nature Genetics, 2004
patient / control
3 independent experiments
95% confidence intervals
SLIDE 41
use of multiple references is now well established
> 1250 citations of our geNorm technology in PubMed > 8000 geNorm downloads in 100 countries
normalisation using multiple stable reference genes
SLIDE 42
data analysis using qbasePLUS
based on Ghent University’s geNorm and qBase technology up to fifty 384-well plates multiple reference genes for accurate normalization detection and correction of inter-run variation dedicated error propagation automated analysis; no manual interaction required
data analysis
http://www.qbaseplus.com
SLIDE 43
data analysis using qbasePLUS
based on Ghent University’s geNorm and qBase technology up to fifty 384-well plates multiple reference genes for accurate normalization detection and correction of inter-run variation dedicated error propagation automated analysis; no manual interaction required
59 prognostic markers + 5 reference genes 366 samples hierarchical clustering survival analysis
Prediction Analysis of Microarrays Cox proportional hazards modeling Kaplan-Meier
data analysis
SLIDE 44
Prediction Analysis of Microarrays PAM training test 15 low risk 15 high risk 334 samples PFS OS
SLIDE 45
classification of patients with respect to PFS and OS
50 100 150 OS total SIOPEN cohort (n = 313)
survival probability (%)
LR n=245 (5) HR n=68 (27)
p = <0.001 (log-rank)
20 40 60 100 80 time (months) 50 100 150 PFS total SIOPEN cohort (n = 312)
survival probability (%)
LR n=245 (42) HR n=67 (35)
p = <0.001 (log-rank)
20 40 60 100 80 time (months)
SLIDE 46
value of the classifier in relation to currently used risk factors: PFS
50 100 150 20 40 60 80 100 10 20 30 40 50 60 70 LR n=94 (17) LR n=152 (25) LR n=234 (39) LR n=222 (31) LR n=8 (1) LR n=24 (11) HR n=48 (24) HR n=31 (18) HR n=34 (21) HR n=34 (15) HR n=34 (17) HR n=20 (12)
p = <0.001 (log-rank) p = <0.001 (log-rank) p = <0.001 (log-rank) p = <0.001 (log-rank) p = 0.12 (log-rank)
time (months) 50 100 150 time (months) 50 100 150 time (months) 50 100 150 time (months) time (months) time (months) 100 80 60 40 20
survival probability (%)
100 80 60 40 20
survival probability (%)
100 80 60 40 20
survival probability (%)
100 80 60 40 20
survival probability (%)
100 80 60 40 20
survival probability (%) survival probability (%)
PFS age <=12 months (n = 172) PFS not stage 4 (n = 256) PFS stage 4 (n = 58) PFS amplification (n = 42 ) MYCN PFS single copy (n = 265) MYCN PFS age > 12 months (n = 142) p = 0.22 (log-rank)
100 80 60 40 20
SLIDE 47
Cox multivariate analysis independent predictor (age, stage, MYCN) multivariate cox analysis
PAM classifier strong independent predictor: patients with high molecular risk have a 19-fold higher risk to die from disease a 4-fold higher risk for relapse/progression compared to patients with low molecular risk
SLIDE 48
RNA quality control cut-off
depends on the application
microarray vs. qPCR expression difference of the target(s) abundance & stability of the target(s) fresh frozen vs. FFPE
based on the performance of our classifier
ROC AUC accuracy analysis
bad good RQI <3: 0.27 3: 0.82 5’-3’ dCq >7: 0.43 7: 0.79 AluSq Cq >15: 0.13 15: 0.81
SLIDE 49
conclusions (I)
validation matters – quality control along the entire workflow
assay performance template quality normalization data-analysis
SLIDE 50
conclusions (II)
largest qPCR gene-expression study (rigourous RNA quality control)
optimized workflow using minimal amounts of RNA (100 ng) use of absolute standards (cross-lab comparison) selected gene list (59) on a large panel of tumours (366 + 223)
robust multigene expression prognostic classifier
validated on an independent set of tumours independent after controling for other known risk factors suitable for routine lab tests
this study might form the basis for future research, i.e. prospective
studies
cDNA library source for future qPCR gene expression studies
SLIDE 51