DNA Microarraybased test systems:

1
DNA Microarray-based test systems

• Achieving High Quality Data

Dr. Roland Toder r.toder.consulting Germany
2
Biochips or arrays with various biological
material as tagged probes
Array formats are divided on the basis of feature
density. Macroarrays have less than 500
features/sqcm microarrays have more than 500
features/sqcm.
3

• Pharma
• Pharmacogenetics (pharmacokinetic) personalized
  medicine (haplotyping)
• Human Genetics
• neonatal screening
• Infectious disease
• subtyping of pathogens
• Transplantation medicine
• HLA-typing haplotyping - homo-/heterozygotes)
• Cancer
• evaluation of aggressiveness part. tumors

• Pharma
• Elimination of drug candidates likely to produce
  unacceptable side effects.
• Proteins and their operation in biological
  pathways of disease and enlist them as targets
  for the development of new and better medicines
• pharmacodynamics (amount of metabolizing enzymes)
• Cancer
• subtyping of cancers and stage (B-cell lymphoma)
• cancer treatment (chemotherapy metastasis)
• Infection
• stage of disease (infection) depending on the
  enzyme-cascade

Infectious Disease Pathogen screening
(sepsis-chip) Blood bank
4
Expression profiling approach in Diagnostics

• First step Broad approach (many genes) in order
  to identify relevant genes
• this is the most critical step where the quality
  of the data is important.
• Example
• Screening of women which have had breast cancer
  in order to find out whether they do need
  systemic drug therapy to prevent tumor spread
  (metastasis) after surgery.
• 100 women where screened with a chip containing
  25,000 genes
• A70 gene signature was sufficient to indicate
  whether metastasis do appear soon.

5
SNP Detection

• By creating microarrays that detect
  disease-associated DNA sequence changes
  (polymorphisms or mutations), this technology may
  be used to identify groups of DNA sequence
  changes that predispose patients to certain
  diseases.

Human variation at the level of single nucleotide
polymorphisms may hold the key to effective
disease treatment.
6
Examples in medical DNA diagnostics
HLA typing Chip from Genescan Europe AG

• DQA1 of the chip as an example
• HLA-DQA1 22 alleles
• Sequence-specific oligos out of 13 polymorphic
  regions of exon 2 and 3 of the HLA-DQA1-gene on
  the chip combination of 3 oligonucleotides per
  allele allow the typing of the 22 alleles

7
Allel 01021
controls
a b c
Oligo variants
1 2 3 4 5 7 8
9 11 12 13 14 15
Cy5-control
a b c
Perfect match
1 mismatch
8
Allel 0303
controls
a b c
Oligo variants
1 2 3 4 5 7 8
9 11 12 13 14 15
Cy5-control
a b c
Perfect match
1 mismatch
9
NOC-PCR (Genescan Europe AG)
One Amplicon multiple internal solid phase primer
HLA - DQA haplotypes
10
HLA/DQA Genotyping with NOC-PCR
DQA 102/05
DQA 201/05
11
HLA/DQA Exon 2 comp. 0201/05-0102/05
1,2
1
0,8
T
0,6
mA
A
0,4
0,2
0
0102/05
0201/05
Genotype
12
HLA/DQA Exon 2 comp. 0201/05-0102/05
13
Liver-Chip (GeneScan Europe AG)
C
A
B
(Yellow-Experiment) HepG2-RNA-Cy5 HepG2-RNA-Cy3
HepG2-RNA Cy-5 Jurkat-RNA Cy-3
14
Using Microarray Data for Drug Discovery What
is necessary?

• Standards and Controls
• More stringent requirements for process control
  and methodology needed
• Measurable repeatability
• Measurable reproducibility
• Process definition
• Quality metrics
• Process must be robust to stand up to regulatory
  scrutiny

15
(No Transcript)
16
Microarrays are a Closed Loop Process
Chip Chemistry
Statistical Analysis
Closed-loop Process
Probe Labeling
Detection
Standardized Kits Available
Hybridization and Washing
Standardized Kits Available
17
What is necessary to obtain quality data in the
detection process?

• Right Resolution
• For proper sampling of spot data
• Signal to Noise Ratio
• To better discriminate between real data and
  system noise
• Repeatability
• For replicable results
• Statistical Analysis
• Proper metrics for spot quality, and for
  determining data trends

18
Internal controls for each spot will be necessary
for diagnostics

• A false positive or false negative result in
  diagnostics can be fatal
• So what degree of stability, security, and
  safety of data acquisition should be provided
  for an application biochip in medical
  diagnostics?
• (FDA/international standards)

The quality of data acquired through the analysis
system is therefore the most crucial and
critical step.
19
Image Sampling Right Resolution
Simplest Case T-Test of means for 2 spots.

• Assumptions Normal distribution, 2-tailed test.
• Statistical sampling can be determined a priori
  with a power analysis.
• Power 1- ?
• Where ? is the probability of failing to reject
  the null hypothesis when it is false. In other
  words, failing to find a difference that actually
  exists.

20
Image Sampling Pixels/Spot
Effect Size and Power

90 ?m diameter spot
5 ?m
10 ?m
15 ?m
3.25 ?m
Large gt 99.9 Medium 99 Small 85
Large 95 Medium 60 Small 27
Large 58 Medium 27 Small 9
represents the smallest spots made due to
variation in contact printing
Cohen, J. (1988) Statistical power analysis
(2nd ed.). Hillsdale, N.J. Erlbaum.
21
Repeatability and Reproducibility

• Reproducible and predictable performance of a
  system is essential for providing the best data
  and process control.
• Typical laser scanners produce R2 values from.60
  to .98
• CCD
• Sampling statistics
• Constant Gain of the detector
• R2 values from .95 to .98
• slide to slide

Meaning 95 of the data is attributable to the
experimental conditions and only 5 must be
determined with repeated scans.
22
Reproducible Systems
Chip Chemistry
Statistical Analysis
Closed-loop Process
Probe Labeling
Detection
Hybridization and Washing
23
Quality Data - Signal to Noise (SNR)

• Critical for detecting small changes in signal or
  detecting weak signals
• Noise in the acquisition system makes
  determination of significant data at low levels
  difficult.
• Background subtraction methods affect the
  linearity of the data

Low level gene expression is difficult to
determine unless system has low noise and high
data integrity.
24
SNR is the foundation of image quality.

Defined as the mean of the signal divided by its
standard deviation.
25
CCD Integration Time Yields More Photons

• A CCD device can detect approximately 400x more
  photons than a laser scanner.
• Simultaneously sampling 200,000 pixels with the
  CCD allows a much longer collection time, and
  therefore many more photons/pixel are detected.
• This greater number of photons yields a high
  signal-to-noise ratio (SNR), which is a key
  determinant of a high spot and image quality.

26
SNR Test Repeat Scan Comparison
Laser Scanner
arrayWoRxe
Ratio of Cy3/Cy3 0.998 ?0.4569 (CV 0.46)
Ratio of Cy3/Cy3 0.987 ?0.0321 (CV 0.03)

• This test reveals scanner statistical noise
  independent of intensity differences on the
  slide.
• Same slide scanned 2x on each scanner.
• Same area compared from each scan (background, no
  spots).
• Overlay each replicate wavelength and measure the
  pixel ratio coefficient of variation (CV).
• There is good agreement between the predicted and
  actual measurement accuracy for both scanners.

27
Data Quality Coefficient of Variation of the
Ratios (CVR)

• The CVR is a collapsed measure of both the
  quality of the chemistry and the quality of the
  image.
• Statistical analysis of spot intensity gives a
  measurement of the relative expression of the
  gene (the ratio between the 2 probes used in the
  experiment) but also the confidence in the
  measurement.
• The CVR is the normalized variance for the spot
  ratio, expressed as percentage of the ratio.

28
Coefficient of Variation of the Ratios (CVR)
Defined as the standard deviation of the pixel
ratios of the two channels divided by the mean
pixel ratio of the two channels.
A low CVR value indicates high confidence in the
ratio.
29
Coefficient of Variation of the RatioCVR
CVR is a Powerful metric that completes the
Closed-loop Process
CVR VariabilityProcess VariabilityDetection
Surface Chemistry Buffers Labeling Probe
Chemistry Humidity Spotting
Resolution SNR Geometry Uniformity Repeatability C
hannel Alignment
Fixing one without the other does not reduce the
CVR
30
Coefficient of Variation of the RatioCVR

• Strong indicator of spot quality
• Strong indicator of experiment quality
• Can be used to weight spot data for inclusion in
  bioinformatics analysis
• With proper imaging, the ratio variability within
  each spot is predictive of the variability
  between experiments
• Can be used to reduce the need for experimental
  repeats

This tool will be very important in routine
analysis where a maximum reliability of each
array is required.
Brown, et al. (PNAS Vol. 98 no. 16
pp.8944-8949 2001)
31
Conclusions

• DNA-arrays when used in diagnostics and pharma
  have to meet a high degree of reproducibility,
  reliability and quality in order to become a
  standard tool.
• The production of chips and necessary consumables
  such as labeling systems washing and detection
  systems, scanners and software have to meet
  established certification systems such as ISO
  GMP and FDA.