Applications of Genomic Technologies

1
Applications of Genomic Technologies to Improve
Recognition, Understanding, and Assessment of
Pharmaceutical Actions A Focus on Integrating
Gene Expression Data Sets into Regulatory Practice
Frank D. Sistare Office of Testing and
Research Center for Drug Evaluation and
Research Food and Drug Administration April 9,
2003
2
Presentation Outline

• Overview of the technology and medically related
  applications of pharmacogenomics relevant to
  CDERs responsibilities
• Concerns and issues (technical, procedural,
  biological) raised at the drug development,
  regulatory oversight, patient care interface
• What CDER is doing to address these issues

3
(No Transcript)
4
PCR (Polymerase Chain Reaction)
Modified from http//www.accessexcellence.org/AB
/GG/polymerase.html
5
(No Transcript)
6
Dual-color Microarray Platforms for Gene
Expression Profiling

• cDNA microarray
• spotted oligo microarray
• in situ synthesized oligo microarray

7
Anatomy of a GeneChip Probe Array
Probe Array
18µm
1.28cm
500,000 specific 25mer oligonucleotide probes
Hybridized Probe Array Image
8
GeneChip Product Sales Growth

• Majority of arrays sold for expression analysis
• Top 20 pharma and most top biotech are customers
• Customers in US, Canada, Europe, Japan, Pacific
  Rim

9
A Powerful Protein Function Detector
PERTURBATION
10,000s of REPORTERS
Database
Identifying Changes in the Function of many
Proteins Simultaneously
10
A Gene Expression Signature As A Predictor of
Survival in Breast Cancer NEJM 347 (12/19/2002)
1999-2009
The Netherlands Cancer Institute Departments of
Pathology, Molecular Carcinogenesis,
Radiotherapy, Biometrics Laura van t Veer, Marc
van de Vijver, Guus Hart, Hans Peterse, Karin
van der Kooy, Dorien Voskuil, Tony van der Velde,
Douwe Atsma, Anke Witteveen, Ron Kerkhoven,
Molecular Pathology, René Bernards Emiel Rutgers,
Harry Bartelink, Sjoerd Rodenhuis Rosetta
Inpharmatics/Merck and Co.Inc Kirkland WA,
USA Hongyue Dai, Yudong He, Mao Mao, Matthew
Marton, North Creek, Tracy Erkkila, Mark
Parrish, George Schreiber, Chris Roberts, Peter
Linsley, Stephen Friend
11
Can gene expression profiling be used to improve
prediction of clinical outcome?

• Aim
• to identify patients at risk to develop distant
  metastases
• to accurately select for adjuvant therapy (who to
  treat, avoid over-treatment)

12
Unsupervised Clustering Analysis of Breast Cancer
Expression Profiles
5000 genes
Lymphocytic Infiltrate
Agioinvasion
BRCA1
ER
PR
Grade 3
98 Tumor Samples
Contains ESR1 and enriched for estrogen-responsive
genes
Enriched for lymphoid genes (B cells, monocytes)
13
Supervised Classification Prognosis
Leave-one-out cross-validation
70 significant prognosis genes
good signature
78 tumors
poor signature
threshold set with 10 false negatives 91
sensitivity, 73 specificity
14
Independent validation set of 73 tumors patients
lymph node negative
73 tumors
70 prognosis genes
Patients metastases 5
yrs
7 false negatives (n1) 94 sensitivity, 55
specificity
15
Independent validation set of 114 tumors patients
lymph node positive
114 tumors
70 prognosis genes
Patients metastases 5
yrs
7 false negatives (n2) 93 sensitivity, 49
specificity
16
Independent cohort of 295 tumors patients yrs, lymph node negative or positive
295 tumors
70 prognosis genes
Unselected series, mean follow-up 8yrs
17
Identifying Patterns Associated with a Complex
Trait
Identify Phenotypic Extremes

• Identifying patterns of expression associated
  with a trait is enhanced if we focus on the
  phenotypic extremes

18
Sub-classifying the Phenotype Using Microarray
Data
Identify Expression Patterns Associated With
Extremes
Identify Expression Patterns Associated With
Extremes

• The subtypes of the phenotype are identified
  using the gene expression data
• For each subtype there are patterns (signatures)
  of expression that serve to enhance
  identification of these more homogenous groups
  (homogenous with respect to the processes
  underlying the phenotype)

19
1.00
0.90
0.80
0.70
TS
0.60
DPD
0.50
Estimated Probability of Surviving
0.40
TS4, or TP18, or
0.30
DPD2.5 (n21)
0.20
0.10
p
0.00
0
6
12
18
24
30
36
Months since start of chemotherapy
20
40 Gene Cluster Discriminates between Good and
Poor Outcomes in Patients with Renal Cell
Carcinoma
Takahashi et al. PNAS 98 9754, 2001
21
(No Transcript)
22
SPS Allows Accurate Clustering of Compounds
Allows Identification of Sub-parts of a
Compounds Effect
Signature type
Compounds?
DNA crosslinker
Statin
Fibrate
NSAID
Signatures?
Hepatotoxicty
Estrogen
2-D Hierarchical Clustering of 29 signatures vs
28 compounds
Signatures relate expression changes to
pharma-cology, toxicology, pathology, chemistry
and biology
23
Microarray Design Comparisons
Oligo
GeneChip
cDNA
24
Knowledge of Biological Truth!
Bioinformatics, Cellular processes/ biochemical
pathway analyses, GO ontologies, referencing
previous experiences, database comparisons,
homework
Information
Data transformation, normalization, scaling,
statistical selection
Microarray Expression Data
Sample integrity, enzymatic processing, amplificat
ion, labeling, array hybridization, washing,
array scanning
25
The Effect of Simvastatin on Expression Levels of
Genes in the Cholesterol Biosynthesis Pathway
Oral administration of Simvastatin results in
significant upregulation of genes involved in
cholesterol biosynthesis in the liver.
26
Concerns and issues raised at the drug
development, regulatory oversight, patient care
interface

• Technical
• With so many variables/options in data capture
  (probes, platforms, RNA sample processing, hyb
  processing), data analysis (true signal,
  normalization, statistics, clustering) are we
  measuring true biology reproducibly and
  accurately or is there too much systematic
  experimental error to reasonably contend with?
  are the data misleading or can we get the same
  true answer from datasets off different platforms
  and different laboratories? can universal
  reference standards help us?
• What is a reasonably detailed and practically
  useful relational data set to constitute a
  regulatory submission that defeats healthy
  skepticism concerning data integrity? will these
  data be appropriate for an FDA database?

27
Concerns and issues raised at the drug
development, regulatory oversight, patient care
interface

• Procedural
• Would all (or any) of these tests ...have to
  be done under GLP conditions? be interpreted by
  FDA as relevant to human safety? Or is a
  research information package approach feasible?
  If such an approach becomes reality, what data
  are appropriate to submit as such? what data
  are inappropriate?
• How will the FDA prepare itself to work with
  these huge data sets in a timely manner? .and
  ensure that individual reviewers do not
  prematurely interpret, generate hypotheses, or
  over interpret those expression alterations whose
  biological significance have not been
  scientifically established?
• How will the FDA communicate which expression
  alterations have reached a scientifically mature
  level of understanding and can rationally be
  considered relevant safety biomarkers?

28
Concerns and issues raised at the drug
development, regulatory oversight, patient care
interface

• Biological
• How would the Agency react if an oncogene was
  activated? would the sponsor have to notify the
  FDA, their clinical investigators, and IRBs?
• Will ?more sensitive? gene expression changes
  drive drastically lower clinical trial starting
  doses and prolong Phase I clinical trials?
• Which expression alterations are reliable and
  biologically relevant classifiers, or
  biomarkers of 1) desirable actions, 2)
  undesirable but tolerable drug actions, 3)
  healthy and fully compensatory responses to
  exposure, 4) intolerable drug actions leading to
  irreversible outcomes? how are they
  biologically relevant?

29
Modifying Traditional Drug Development
Related Target Hit
Lead Compounds
FDA Approval
Animal Trials
Clinical Trials
Absorption
Metabolism
Excretion
Toxicity
30
Likely Pattern of Incorporation and Assessment
of Microarray Data in Decision Making along the
Medical Product Development Pipeline.
General relationship between 1) the number of
microarray endpoints likely to be measured, 2)
the degree of precision in each measurement that
would likely be expected, and 3) the level of
analytical and biological validation achieved by
the microarray platform, as a function of the
application stage of product development.
31
What CDER is doing to address these issues

• Building the internal capacity infrastructure
• Establish core expertise within the CDER
  laboratory
• initial approach arrays, cDNA scanner, the ILSI
  collaboration
• enabled leveraging Affymetrix GeneChip system
  and research agreement leveraged Rosetta
  research agreement
• Office of Science Grant from Commissioner RNA
  standards development initiative (all medical FDA
  Centers NCTR).
• Expand core expertise for CDER review enhancement
• CDER Nonclinical Pharmacogenomics Subcommittee
  established - focus on regulatory decision making
  practices, procedures, and policies
• leveraged Iconix DrugMatrix database - expanded
  and focused reviewer/researcher training

32
? 0.984 (intersection)
? 0.936 (union)
33
? 0.972 (intersection)
? 0.873 (union)
34
What if.a reviewer sees increased expression of
an oncogene in a product submission package?
Diphenhydramine
Mannitol
Aspirin
72 -oncogenes
White 9 Human and Rodent Genotoxic
Carcinogens Blue 5 Human and Rodent
Nongenotoxic Carcinogens Yellow 14 Rodent
Noncarcinogens
35
Efforts have been intiated by CDER, NCTR, CDRH,
CBER with NIST, NIEHS, and Microarray
Stakeholders (Affymetrix, Rosetta/Agilent,
Iconix/Amersham) to Develop Standards useful for
evaluating platform features..

• No manufacturing defects
• Insignificant platform lot-to-lot variability
• Assess integrity of feature location
• Unambiguous consensus sequence annotation
• Lack of cross-contamination in tiled probe
  features

36
.and for evaluating experimental performance

• Quality (integrity /purity) of starting sample
• Quality of processed (labeled/amplified) sample
• Hybridization performance (probe sensitivity,
  specificity)
• Image scanning limitations (backgrd/slope/saturati
  on)
• Transformation process into rough measured data
  (bckgrd/slope/saturation)
• Normalization/scaling to an analytical value
  worthy of comparison
• Data selection and analysis procedures to focus
  biological thinking (false positive/false
  negative minimization)
• Biological conclusions that are independent of
  platform and represent biological truth

37
Goals of CDER PTCC Nonclinical PG Subcommittee

• Develop standards and procedures for submission,
  review, integration of PG data.
• Develop internal consensus regarding added value,
  best interpretations, and regulatory review
  implications.
• Develop Center expertise to effect an appropriate
  infrastructure for PG data review and
  integration.
• Develop initiatives to keep abreast of latest
  developments in PG.
• Interface with other CDER review discliplines
  (including CDER IT groups).
• Provide forum for communication with the
  regulated industry and other stakeholders.

38
What CDER is doing to address these issues

• Establishing a network to assimilate reasonable
  consensus
• Develop mechanisms to communicate and deliver
  needs
• NIST RNA Standards workshop and initiative
• NCTR collaborations ongoing (several biological,
  database, statistical, ref standards)
• FDA Intercenter Communications group (consumer
  white paper)
• National Academy of Sciences Committee -
  Government Liaison
• NIH/NIEHS MIAME-Tox Interchange
• Engage external expertise (non-collaborative)
• Pharm Tox Subcommittee to CDERs Advisory
  Committee for Pharmaceutical Science- focused
  expert group - 6/10/03 Meeting
• Practice mock data submission scheduled

39
CDER shares the vision that applications of
pharmacogenomic technologies will improve.
(1) recognition (easier, quicker, or
more accurate identification of efficacy,
efficacy potential, toxicities, and toxicity
potential) (2) understanding (modes or
mechanisms of drug actions), (3) assessment
(significance and relevance of findings to
humans) .of
Pharmaceuticals
40
CDER has a responsibility to enable and not force
nor impede evolution/vision
Related Target Hit
Lead Compounds
FDA Approval
Animal Trials
Clinical Trials
Animal Trials
Absorption
Metabolism
Excretion
Toxicity