Validation of Biochip Products
based on Microarray Platform
Jen-pei Liu, PhD
Division of Biometry, Department of Agronomy National Taiwan University
and
Division of Biostatistics and Bioinformatics National Health Research Institutes
at
Outline
Current Issues
Validation and Analytical
Performance
MAQC Project
Two-gene expression ratio predicts
clinical outcome (Ma, et al, 2004)
Tamoxifen for breast cancer
A competitive inhibitor of estrogen binding to estrogen
receptor (ER)
Reduction of 40%-50% in annual risk of recurrence 5.6% improvement in 10-year survival
ER and progesterone receptor (PR, an indicator of a
functional ER pathway) currently the best predicator of tamoxifen response
25% of ER+/PR+, 66% of ER+/PR-, and 55% of
ER-/PR- fail to respond
To predict tamoxifen treatment outcome in early-stage
Issues
Reid et al. (2005, JCLO) reported that they
can not reproduce the Ma’s results
Simon (2005, JCLO) criticized the MA’s
methodology
False + by chance = 5475x0.001 = 5.4 Observed + = 9 ⇒ 60% false + rate Unlikely to have an accurate predictor
Cross-platform and intra- and
inter-laboratory reproducibility
Different designs
Different concepts
Different referenced samples
Different procedures for sample
acquisition
Different experimental protocols
Cross-platform and intra- and
inter-laboratory reproducibility
Recognition and evaluation of reproducibility
Only Recently:
Dobbin et al. (Clinical Cancer Research, 2005) Larkin et al. (Nature Methods, 2005)
Irizarry et al. (Nature Methods, 2005)
Members of the Toxicogenomic Research Consortium
(Nature Methods, 2005)
Tan, et al. (Nucleic Acids Research, 2003) Yauk, et al. (Nucleic Acids Research, 2004)
Evaluation of Commercial
Platforms (Tan et al, 2003)
Platforms
Affymetrix (U95Av2, GeneChips, 25mer oligo probe sets)
Agilent (Human 1, cDNA probes)
Amersham (Codelink UniSet Human I Bioarrays, 30mer Oligo probes)
A group of 2009 common genes present on three platforms
Crossplatform and intra- and
inter-laboratory reproducibility
Conclusions from cross-laboratory and
cross-platform studies are meaningless
when there are fundamental problems in
achieving acceptable intra-laboratory
reproducibility
Conclusions from microarray data analysis
are meaningless when a minimum
requirement for data quality is not met
and provided
Quality Assurance and
Quality Control
Four types of factors affecting
microarray experiments
Technical
Instrumental Computational Interpretive
A single hidden and uncontrolled factor can completely negate a microarray experiment
Quality Assurance and
Quality Control
Technical Factors
Microarray manufacturing Sample collection RNA extraction cDNA/cRNA synthesis Labeling with fluorescent dye Hybridization
Quality Assurance and
Quality Control
Instrumental Factors Imaging acquisition Quantification Computational Factors Data processing Normalization Analysis Interpretive FactorsMicroarray QC Metrics and
Thresholds (MAQC) Project
Sponsored by the US FDA
Establish QC metrics and thresholds to
objectively assessing the performance
of microarray platforms and merits of
various data analysis methods
Two RNA samples from three species
Microarray QC Metrics and
Thresholds (MAQC) Project
Assess the precision and cross-platform
comparability
Nature and magnitude of systematic
bias assessed by QT-PCR
Microarray QC Metrics and
Thresholds (MAQC) Project
Four government agencies
10 platform providers
3 RNA sample provider
27 test sites
10 data analysis sites
200 people from more than 70
Microarray QC Metrics and
Thresholds (MAQC) Project
Completion of MAQC main study – Oct. 2005 Submission of manuscript and release of
MAQC datasets – Feb. 2006
Publications – July-Sept. 2--6
Public meeting on microarray quality control
and data analysis – Dec. 2006
Guidance on microarray quality control and
Performance Evaluation
Analytical Performance
Accuracy Precision
Clinical Effectiveness
Diagnostic accuracy and variability
Clinical Utility
Analytical Performance
Validation assessed functional performance
Precision (Reproducibility)
Assay Sensitivity (Limit of Detection) —ability to
accurately identify positive samples
Assay specificity (Accuracy)
Interfering substances (endogenous and
exogenous)
Validation of cut-off, reference range, or medical
decision point
Roche AmpliChip CYP450 2C19 Test
Analytical Performance
Precision (Reproducibility)
CYP2C19
genotype No. Tested
Genotype Calls Correct Calls Correct Call Rate (95% CI) *1 / *1 134 134 (100.0) 133 0.99 (0.97) *1 / *2 135 135 (100.0) 135 1.00 (0.98) *1 / *3 135 135 (100.0) 135 1.00 (0.98) *1 / *1 135 135 (100.0) 135 1.00 (0.98) *1 / *2 135 134 (99.3) 134 1.00 (0.98) *1 / *2 135 134 (99.3)) 134 1.00 (0.98) Total 809 807 (99.8) 806 1.00 (0.99)
Roche AmpliChip CYP450 2C19 Test
Analytical Performance
Roche AmpliChip CYP450 2C19 Test
Analytical Performance
Roche AmpliChip CYP450 2C19 Test
Analytical Performance
Limit of detection
Lowest and highest concentration of input sample that yields a consistent and accurate result
DNA Amount (ng) Number of Arrays Number of Correct Calls Positive Rate 95% Confidence Limit 50 144 144 100% 97.5 – 100% 25 144 144 100% 97.5 – 100% 2.5 144 144 100% 97.5 – 100% The lowest level of genomic DNA at which a ≥ 95% positive rate was obtained for correct detection of the CYP2C19 gene was 2.5 ng of input DNA.
DNA Amount (ng) Number of Array s # Correct Calls Positivity Rate 95% CI 50 144 144 100% 97.5 – 100% 25 144 144 100% 97.5 – 100% 2.5 144 134 93.1% 87.6 – 96.6%
Roche AmpliChip CYP450 2D6 Test
Analytical Performance
The lowest level of genomic DNA at which a >= 95%
positivity rate was obtained for correct detection of CYP2D6 (*4DxN/*41 and *4/*5 samples) was 25ng.
Potential Interferences
Endogenous and exogenous common substances Commonly prescribed drugs
Molecules similar to the analyte
AmpliChip CYP450 Test
z 10 unique patient samples were tested with and without spiking
of albumin, bilirubin and triglycerides.
z Albumin – 6000 mg/dL; Bilirubin - 60 mg/dL; Triglycerides –
3000 mg/dL (approximately 10-fold greater than normal).
z Elevated levels of lipids, bilirubin and albumin in specimens did
Five-Year View
Calibration methods to systematically correct
ratio under-estimation
Minimization of technical variation to
reproducibly detect subtler changes (e.g. 1.4-fold)
High-throughput tools for a large number of
samples (instead of large number of genes) to identify a small set of biomarkers for drug discovery and development and patient
Five-Year View
A technology platform for assay a small to
medium number (dozens to hundreds) of established biomarker genes in a
high-throughput fashion in terms of samples is needed for diagnostic purposes (Genomic Composite Biomarker Classifier)
A balance between high-density microarrays
Measures of Similarity
Correlation coefficient
A measure for association
Not a measure for similarity (or agreement)
Euclidean distance
A measure for agreement
Measures of Similarity
Example
Case I Case II Case III
X1 X2 X1 X2 X1 X2
1 1 1 2 1 4
2 2 2 4 2 8
3 3 3 6 3 12
Methodological Research
Differentially expression genes
Statistical significance based on hypothesis of difference does not take into
consideration of biological significance Fold change does not take into
Methodological Research
Statistical hypothesis for identification of differentially expression genes should take into consideration biological significance
Ho: μT - μC ≥ δL and μT - μC ≤ δU Vs.
Methodological Research
Hypothesis of no correlation can not
prove the consistency
With 5000 genes, a correlation of 0.05
is statistically significant at 1% level
Use of concordance correlation for the
consistency
Methodological Research
Genomic Composite Biomarker (GCB)
Classifier
The number of genes in GCB classifier is a random variable
The expression level for each selected gene in GCB classifier is also a random variable
Methodological Research
Selection of threshold
Evaluation of systematic bias at
threshold
Evaluation of random error
Measurement error model
Expression levels of genes are not
Courses
Design and Analysis of Microarray
Experiments – Spring, 2006
Statistical Genomics – Spring, 2006 Biological Assays – Every two years Statistical Methods for Biotechnology
Products – Every two years
I – QC/QA of Biotechnology products
II – Clinical Trials, Target Trials, BA/BE studies,
Research Products Developm ent