Transcript Slide 1

Comparison of Microarray Data Generated from Degraded RNA using Five Different Target Synthesis
Methods and Commercial Microarrays
#120
Scott Tighe and Tim Hunter
Microarray Core Facility, Vermont Genetics Network at the University of Vermont, Burlington, Vermont, USA
Results
Abstract
As the need to collect gene expression data on degraded RNA continues, an evolution of
new reagents and arrays now on the market may improve these data sets. The focus of this
study is to evaluate the effectiveness of five different target amplification strategies with
three types of Affymetrix GeneChips employing RNA at four different levels of
degradation. For this study, reference RNA was non-chemically degraded to final RIN
values of 10, 6, 4, and 2 as determined by the Agilent Bioanalyzer. Each RNA was amplified
using standard Affymetrix protocols, ExpressArt by AmpTec, and NuGEN’s Ovation V2, WT
Pico, and WT FFPE kits. Preliminary results indicate that data generated from RNA with a
RIN of 9, 6, and 4 performed well on Exon 1.0 ST arrays, Gene 1.0 ST arrays, and 3’ U133a
2.0 arrays using the ExpressArt, NuGEN WT Pico, and FFPE reagents. For RNA with a RIN
value of 2, preliminary data suggests that the NuGEN WT Pico and FFPE provide the
highest correlation to data generated by higher quality RNA followed by data generated by
the ExpressArt reagents. RT-qPCR was also performed and revealed a similar trend to the
microarray data.
3’ Array data: U133a 2.0 Images
Human Gene Array 1.0 ST Images
Std Affy
(cRNA)
Affymetrix 100ng
100ng Affy Prep
NuGEN Ovation
cDNA)
Amptec TR
AmpTec TR
AmpTec TR Micro
(cRNA)
NuGEN FFPE
NuGEN FFPE
(cDNA)
NuGEN FFPE
NuGEN Pico
RIN 10
NuGEN Pico
(cDNA)
RIN 10
RIN 6
RIN 4
RIN 2
% present and Background
Percent present indicates the number of genes detected above background. Low
background is desirable. These data indicate good results for all samples except
AmpTec RIN 2.
Methods
Human Exon Array 1.0 ST Images
Oligo[dT]
RIN 6
RIN 4
RIN 2
Area Under the Curve (AUC) Values, MAD residuals, and RLE Mean
AUC is the statistical measure of the percent true positive signal above the
negative signal. AUC values must be above 0.8 to be considered good. MAD
residual means are values that indicate how a dataset deviates from the
expected mean. High values are an indication of problematic data. RLE means
values also indicates variability of data. High values indicate problematic data.
NuGEN
Pico
RIN 10
RIN 6
RIN 4
RIN 2
Area Under the Curve (AUC), MAD residuals, RLE Mean
These data suggest that AmpTec RIN 10-4 and all NuGEN Pico and FFPE have
consistently good values.
Random Priming
80.00
Moderate scaling factors increases are noted for NuGEN RIN 6-2 RNAs.
Data for AmpTec TR reagents were very good.
80
600
GAP 3-5-ratio
SF
B-actin 3-5-ratio
50
Array
300
300
250
200
200
3’ U133a2.0
1000
NO
AmpTec-RT - ENZO IVT
3’ U133a2.0
475
NO
NuGEN Ovation
3’ U133a2.0
80
NO
NuGEN Pico
3’ U133a2.0
50
NO
NuGEN FFPE
3’ U133a2.0
50
NO
Differential Gene Detection
Number of differentially detected genes as a function of Target preparation method, degradation level,
and Genechip type determined by X-ray software. Note that differential detection can be skewed by a
low call rate.
Percent Venn Diagram
When analyzing degraded RNA using X-Ray
software at the gene-level, the Exon 1.0 ST
array had more genes detected than the Gene
1.0 ST Array.
Affymetrix Exon 1.0 ST and Gene 1.0 ST Arrays
NuGEN-Pico-4-EX
NuGEN-Pico-3-EX
NuGEN-Pico-1-EX
NuGEN-FFPE-4-EX
NuGEN-FFPE-3-EX
NuGEN-Pico-2-EX
NuGEN-Pico-4-EX
NuGEN-Pico-3-EX
NuGEN-Pico-2-EX
Affy-1-EX
NuGEN-Pico-1-EX
4-NuGEN-PicoZ-GA.rma
3-NuGEN-PicoZ-GA.rma
2-NuGEN-PicoZ-GA.rma
1-NuGEN-PicoZ-GA.rma
4-NuGEN-Pico-GA.rma
3-NuGEN-Pico-GA.rma
2-NuGEN-Pico-GA.rma
1-NuGEN-Pico-GA.rma
4-NuGEN-FFPE-GA.rma
3-NuGEN-FFPE-GA.rma
2-NuGEN-FFPE-GA.rma
1-NuGEN-FFPE-GA.rma
4-Affy-GA.rma
3-Affy-GA.rma
2-Affy-GA.rma
1-Affy-GA.rma
4-Amptec-GA.rma
3-Amptec-GA.rma
2-Amptec-GA.rma
rRNA reduction
Std 3’ Affy- ENZO IVT
Conclusions
Preliminary QC results are no substitute for transcript-level bioinformatics. Complete
analysis of these data are in progress but not yet available. We have demonstrated
that microarray analysis of degraded RNA is possible and usable and Exon Arrays
appear to be most forgiving, but more costly.
- No one statistical value demonstrates data quality for Exon, Gene, or 3’ Arrays
- Standard Oligo[dT] target prep methods readily reveal elevated 3’- 5’ ratios for B-actin
and GAPDH for degraded RNA. B-actin is more sensitive to degradation than GAPDH.
Standard Affymetrix Protocol
NuGEN Pico
NuGen FFPE
AmpTec Micro with Affy labeling
- AmpTec 3’ to 5’ ratios remain low regardless of RNA degradation and may not be an
appropiate indicator of data quality and must be considered with Background Ave and %
present.
Method
Array
Input(ng)
NuGEN Pico
Gene/Exon
50
NO
NuGEN FFPE
Gene/Exon
50
NO
Affymetrix WT
Gene/Exon
100
NO
AmpTec-RT-Affy-Label
Gene/Exon
500
NO
Std 1ug Affy Protocol
0
NuGEN-FFPE-4-EX
50
1-Amptec-GA.rma
4-NuGEN-Pico
4-NuGEN-FFPE
4-Amptec-Enzo
4-NuGEN-Ovation
4-Affy
3-NuGEN-Pico
3-NuGEN-FFPE
3-Amptec-Enzo
3-NuGEN-Ovation
3-Affy
2-NuGEN-Pico
2-NuGEN-FFPE
2-Amptec-Enzo
2-NuGEN-Ovation
2-Affy
1-NuGEN-Pico
1-NuGEN-FFPE
1-Amptec-Enzo
1-Affy
0
1-NuGEN-Ovation
10
NuGEN-FFPE-3-EX
0
NuGEN-FFPE-2-EX
100
NuGEN-FFPE-1-EX
150
100
20
Input(ng)
NuGEN-FFPE-2-EX
350
400
30
Method
bgrd_mean
400
40
Standard Affymetrix protocol with ENZO IVT
NuGEN Ovation V2
NuGEN Pico
NuGEN FFPE
AmpTec Micro with ENZO IVT
pm_mean
450
500
60
3’ Affymetrix U133a 2.0 Arrays
500
pm_mean
bgrd_mean
good arrays have high pm_mean and low bgrd_mean
1:RIN=10 2: RIN=6 3:RIN=3 4:RIN=1
Amptec-4-EX
Target Preparation Method
These data indicate very good signal to noise (PM mean-BKGD Mean) for
AmpTec RIN 10 and all NuGEN preps.
QC Metrics Gene Array 1.0 ST
700
70
Perfect Match and Background Means
Amptec-3-EX
RIN 2
Amptec-2-EX
RIN 4
Amptec-1-EX
RIN 6
Perfect Match and Background Means
The values of perfect match probes verses background probes results in a
direct correlation with signal to noise ratios. The greater the differential,
the “cleaner” the signal. High background can negatively affect a true signal.
Good results are indicated by AmpTec RIN 10-4, NuGEN FFPE, and NuGEN
Pico.
Affy-4-EX
RIN 10
High 3-5’ ratios indicate degradation of input RNA. High Scaling Factor indicates a
dim chip or reduced signal. These results show that Oligo[dT] priming is negatively
impacted by degradation indicated by high 3-5’ ratios for RIN 6, 4, 2 RNA.
Affy-1-EX
4-NuGEN-Pico
3-NuGEN-Pico
2-NuGEN-Pico
1-NuGEN-Pico
4-NuGEN-FFPE
3-NuGEN-FFPE
2-NuGEN-FFPE
1-NuGEN-FFPE
4-Amptec-Enzo
3-Amptec-Enzo
2-Amptec-Enzo
1-Amptec-Enzo
4-NuGEN-Ovation
3-NuGEN-Ovation
2-NuGEN-Ovation
1-NuGEN-Ovation
4-Affy
3-Affy
2-Affy
1-Affy
3’-5’ ratios and Scaling Factor
NuGEN-FFPE-1-EX
0.00
Amptec-4-EX
10.00
Amptec-3-EX
20.00
Amptec-2-EX
30.00
Amptec-1-EX
40.00
Affy-3-EX
Ambion First Choice™ Human Brain reference RNA was degraded to four levels determined by
the Agilent Bioanalyzer 2100 using open tube method.
rle_mean
Affy-4-EX
50.00
mad_residual_mean
Affy-3-EX
60.00
pos_vs_neg_auc
Affy-2-EX
BG Avg
Affy-2-EX
RNA Source
1.2
1.1
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
%P
70.00
- All methods including Affy, AmpTec, NuGEN Pico and FFPE generated acceptable Exon
Data for RIN values 10, 6 ,4 based on AUC data, Residual Mean, and RLE values. However,
when the exact same hybridization mixes were applied to Gene arrays, background means
were elevated for the Affy-100ng method.
rRNA reduction
Not tested- rReduction CV to high
Data Analysis
Quality control data analysis was performed using GCOS, Expression Console 1.1, and X-RAY version 3.99385
- The Affy 100ng and AmpTec preps demonstrated much higher Background Means compared
to other methods
3’ Array –U133a 2.0
Correlation Plots
Correlation Plots were generated
using Affymetrix Expression Console
software. These plots indicate
similarity between samples. Red
indicates higher similarity.
Gene 1.0 ST Array
Exon 1.0 St Array
- Good results were obtained for with RIN 2 RNA using NuGEN Pico and FFPE methods
when analyzed by RMA Gene levels for Exon Arrays
.
Acknowledgements
This research was funded through the Vermont Genetics Network at the University
of Vermont (UVM) through Grant Number P20 RR16462 from the INBRE Program of
the NIH National Center for Research Resources. We Gratefully acknowlegde John
Burke at X-Ray Biotique Systems for outstanding software support.