submission of microarray data: dealing effectively with data quality issues and information content...
DESCRIPTION
Assumptions -- Sponsor side Sponsor providing data to support IND or NDA Data is part of a larger package and not the sole and only evidence provided Sponsor has on going array efforts with trained staff or will contract with CRO that meets these requirementsTRANSCRIPT
Submission of Microarray Data:Dealing Effectively with Data Quality Issues and Information Content Necessary to Develop an Expression Database
Advisory Committee for Pharmaceutical Science
June 10, 2003
Challenge: Over 50% of drug failures due to efficacy/tox.
Problem: 20 million US patients exposed to drugs withdrawn between Sept. ’97 and Sept. ’98
Problem: Only 1 in 10 IND’s become approved drugs -- Our current methods of candidate characterization are only 10% accurate
Solution: Bridge genomics and chemistry to broadly understand a compound’s effects in genomic terms
Vision and Challenge
The VisionBetter compounds submitted
Safer compounds submitted and approvedLower approval time
Assumptions -- Sponsor side
Sponsor providing data to support IND or NDA
Data is part of a larger package and not the sole and only evidence provided
Sponsor has on going array efforts with trained staff or will contract with CRO that meets these requirements
Assumptions – Agency side
Agency is willing to develop and train staff so that the data is meaningfully interpreted and a balanced view is taken
– Sponsor is concerned about overly reactive view – Sponsor is concerned about the future liability from public disclosure
Agency is able to accept data in a community defined standard format and has capacity to assess quality
- Agency is prepared for all current technologies - Agency will keep up with future technology improvements
Agency desires to deposit submitted data into an internal database for use by staff during future evaluations
Agency understands that context is essential for a balanced view
Array measurements similarities to traditional measurements, but…
ALT elevation is associated with liver damage- Treated group value is 55 ± 7 (SE) [range: 30-75], n=5- Control group value is 50 ± 4 (SE) [range: 15-110] n=15- No treated animals outside range of controls
– Conclusion: ALT not significantly changed by treatment, consistent with good liver safety.
Gene expression data on 5 RNAs associated with liver damage
- No treated animal outside range of controls– Conclusion: 5 RNAs associated with liver toxicity not significantly
changed, consistent with good liver safety.
RNA A RNA B RNA C RNA D RNA E
Ratio to control 0.9 1.0 1.2 0.8 1.4
SEN (Treated, Untreated)
0.1(5,15)
0.2(5,15)
0.1(5,15)
0.4(5,15)
0.5(5,15)
Range(significant at p<0.05)
1.1-0.8n.s.
1.2-0.7n.s.
1.4-0.9n.s.
0.6-1.1n.s.
1.7-0.9n.s.
… but, microarray data has differences from conventional measurements
Both agency and community has lower familiarity with technology
– Will improve with experience
Concern that the survey might uncover “confounding” factors- Sponsor concerned about a “overly reactive view”
- FDA concerned about sponsor missing important findings
Less “scientific agreement” about how to interpret and meaning
– Pattern matching is less familiar to biological community
– Requires a different mindset than single gene focus of the past
Perception that microarray data is lower quality and noisier
– Technology has improved
– Carefully conducted experiments are accurate and predictive
Summary: What should be provided by sponsor to FDA MIAME/MAGE-ML compliant descriptions of experiment(s) and
electronic submission of all data
Minimum experimental design metrics similar to that required for any other biological experiment
“Youth” of technology requires that additional quality data be submitted to demonstrate competency of experimenters
Sponsors interpretation of data in “scientific paper style” format
Comparison to community accepted RNA biomarkers and compared to Bench Mark Drugs and Toxicants
– Interpretation vs. context of current drugs failed drugs and toxicants
Sponsor provided data to assure QUALITYThemes Measurement vs. lab historical values
Measurement vs. external standards» RNA external standard
Measurement vs. internal standards» A few spike-in standards
•Different than other agency submissions due to “youth” of technology•Need to assure competency of experimenter•Need to assure internal/external consistency •Need to assure consistency with historical values
Experiment is complex but has several points where critical evaluations needed
Laboratory Information Management System (LIMS)
In VivoBiology
In VitroBiology
RNAIsolation
TargetPreparation Hybridization Data
Loading
Laboratory Information Management System (LIMS)
RNAIsolation
TargetPreparation Hybridization
286 steps needed to prepare microarray
Several quality control check points
Independent of platform
Experimental design minimums
Minimum of biological triplicate
Minimum of three untreated or mock or vehicle treated controls, processed contemporaneously
Minimum of three external standard RNA’s, processed contemporaneously
Minimum three spike-in RNA’s each sample
Treated
Untreated -Control
RNA standard External Control
3 Spike RNA’s
RNA used in experiments Ratio of 28S to 18S
• Mean• Standard Deviation• Range• Traces for all RNA samples
Data provided for each the following:
Samples in dataset
Historical for similar tissue/cell prepared in lab
RNA external standard, contemporaneous with dataset
Hybridization and array Q.C.
Data for experimental samples Array average signal to background ratio Array average background Array average raw signal Log dynamic range Average Raw signal intensity for each of three spike-in RNA’s
Comparative data (Mean, SD and range) Historical for similar samples (match for tissue or cell type) Historical average for RNA standard Historical average for each of spike-in RNA’s Average for contemporaneous RNA standard Average for contemporaneous controls
Experiment internal and external consistency
Experimental samples, correlation coefficient (CC) with
1. each other
2. contemporaneous controls
3. contemporaneous external standard RNA
4. historical external RNA standards
5. historical similar tissue or cell line samples
Mean, SD and range in all cases
Sponsor provided “Scientific literature style interpretation”
Abstract Significance relative to application Brief methods Summary of quality evidence described earlier Results Discussion Conclusion relative the application under consideration Conclusions in context with other drugs and toxicants
Report helps the agency help you
Summary: What should be provided by sponsor to FDA MIAME/MAGE-ML compliant descriptions of experiment(s) and
electronic submission of all data
Minimum experimental design metrics similar to that required for any other biological experiment
“Youth” of technology requires that additional quality data be submitted to demonstrate competency of experimenters
Sponsors interpretation of data in “scientific paper style” format
Comparison to community accepted RNA biomarkers and compared to Bench Mark Drugs and Toxicants
– Interpretation vs. context of current & failed drugs and toxicants
Internal FDA contextual databaseMicoarray technology requires stepping into the coming age of
electronic data submissions
Paper submission of microarray data is not useful
Used by agency to develop a better understanding of technology
Used by the agency to allow a look at the data in the context of other submissions
Contextual database is highly useful to provide meaning and balance to interpretation
– Same biology is present regardless of data preparation technology
Once the data is available to the agency how should they evaluate?
Objectives Promote balanced view of the data React to truly significant events Ground the analysis in the context of real world
effects of drugs, failed drugs, toxicants and standards
Need a reference database for these purposes
What is necessary to provide contextual grounding? Database containing …
Wide diversity of successful drugs, failed relatives, toxicants and standards
– Needed to understand pharmacologic mechanism– Needed to understand toxicity mechanism
Multiple tissues with dose and time context
Linkage of expression data to orthogonal data including:- Pharmacology including on and off target sites of action- Histology- Clinical Chemistry- Hematology- Chemical Structure
In vivo and in vitro data for successful bridging domains
Benefits of using a reference database to the agency Provides context to interpret events seen with
candidates – An example drawn from DrugMatrix
EGF-Receptor (p<0.01)
cKit-Oncogene(p<0.01)
BCL2 Oncogene(p<0.001)
Meloxicam
Torsemide
Nislodipine
Rosigolitazone
Norethindrone
Grenisteron
Torsemide
Clotrimazole
Valproic Acid
Simvastatin
Beta-estradiol
Carboplatin
Busulfan
Clofibric Acid
Melatonin
Sparfloxacin
Withdrawn Diethylstilbesterol
Oncogenes elevated by approved drugs in liver
Context promotes balanced view of data
Summary: What should be provided by sponsor to FDA MIAME/MAGE-ML compliant descriptions of experiment(s) and
electronic submission of all data
Minimum experimental design metrics similar to that required for any other biological experiment
“Youth” of technology requires that additional quality data be submitted to demonstrate competency of experimenters
– Additional quality data beyond that required of established technologies
Sponsors interpretation of data in “scientific paper style” format
Comparison to community accepted RNA biomarkers and compared to Bench Mark Drugs and Toxicants
– Interpretation vs. context of current drugs and toxicants
Conclusions Looking forward Micro Array technology is ready to contribute today
– Simple assurances of quality are needed
– Contextual databases to allow meaningful interpretation are available
Developing understanding and consensus around meaningful RNA biomarkers
Requirements beyond normal verification of data quality will diminish as community sophistication using technology improves
Analysis of data collected off several platforms is quite possible
- Same biology is found regardless of the platform
Clinical applications to accessible human tissues will become common
Improved predictive power of animal studies making truly predictive animals models a reality
Addresses the Problem: Patients exposure to drugs which are subsequently withdrawn
Addresses the Problem: Only 1 in 10 IND’s become approved drugs -- Our current methods of candidate characterization are only 10% accurate
Result
Helps realize the VisionBetter compounds submitted
Safer compounds submitted and approvedLower approval time