Transcript No Slide Title

Application of Toxicology
Databases in Drug Development
(Estimating potential toxicity)
Joseph F. Contrera, Ph.D.
Director, Regulatory Research and Analysis
FDA Center for Drug Evaluation and Research
(CDER), Office of Testing and Research
[email protected]
THE REVOLUTION IN PHARMACEUTICAL
DEVELOPMENT
Combinatorial Chemistry
High Through-Put Screening
The Human Genome
The Rapidly Increasing Number and
Diversity of Potential New Products
The Limitations of Current
Toxicology Screening Methods
Increasing Demands on Regulatory
Processes
The Need for Rapid and Effective Screening
Methods to Identify and Prioritize
Potential Toxicity
• For lead selection of the products of high
through-put technology
• To more efficiently assess the potential
hazard of substances especially when
limited experimental evidence is available
• As a rational basis for decisions on the
nature and degree of testing
• Reduce animal testing
Toxicology Studies: Promise
• There are 6 major categories of toxicology
studies: genotoxicity, acute toxicity, chronic
toxicity, reproductive and developmental
toxicity and carcinogenicity
• The design of studies in these categories is
relatively standardized to meet regulatory
requirements
• Post-GLP (Good Lab Practices;1978) studies
and reviews are a potentially rich resource of
good quality toxicology data
Information Applications
Toxicology Databases
• Regulatory decision support
• Retrospective analysis
• Product development
• Guidance development; improving and
•
updating regulatory standards
Identifying relationships between animal
toxicology and human adverse events
CDER Toxicology Databases Contributed to
International Conference on Harmonization
(ICH) Guidances for Pharmaceuticals
• ICH S1B: Testing for Carcinogenicity of
Pharmaceuticals
• ICH S1C: Dose Selection for Carcinogenicity
Studies of Pharmaceuticals
• ICH S1CR: Use of Limit Dose in Dose Selection
for Carcinogenicity Studies
• ICH S4;S4B: Duration of Chronic Toxicity
Testing in Animals
Information Applications
Computational Toxicology; SAR; E-Tox
• Structure activity analysis (SAR) and
•
•
•
predictive modeling for regulatory decision
support
Lead selection in drug development
Estimating and prioritizing potential hazard
when data is limited
Hypothesis generation, identifying data
gaps; prioritizing research
Computational Toxicology; E-Tox
The application of computer technology
to analyze, model and predict
toxicological activity
E-ADME
The application of computer technology to
analyze, model and predict absorption,
distribution, metabolism and excretion
Current Database Needs and Issues
• Critical need for uniform compound
identification; problems with multiple drug
names, codes, CAS numbers for same active
ingredient
• Better search and retrieval capability within
and across databases
• Chemical structure similarity search and
clustering capability
• Data entry, quality and compatibility issues
• Proprietary issues; Data sharing
Major FDA/CDER
Carcinogenicity Database Fields
• Drug name
• *Molfile digital
chemical structure
• 2D structure
• Administrative code
(NDA, IND number)
• Clinical indication(s)
• Pharmacological or
chemical class
•
•
•
•
•
•
•
Species, strain
Sex
Route
Doses
Duration of dosing
Tumor site, type
Tumor incidence
Using Chemical Structure (Molfile) as a
Key Field to Link Databases and
Expand Search Capabilities
Compound
Names
Molfile
“core”structure
fingerprint
Key Field
Structural Similarity
Searching, Cluster
Analysis
(ISIS-Base)
Compound
Structure
SAR/E-Tox
MCASE structural
alerts
FDA CDER TOXICOLOGY KNOWLEDGE BASE
For Decision Support and Discovery
Chemical Structure
Similarity Searching
(MDL Isis-Base)
Clinical
*ADR
AERS
*Clinical Post-Marketing
Adverse Drug Reaction
Adverse Event Reporting
Systems Databases
Chemical
Structure Based
Substance Inventory
(MOLFILE)
Computational
Toxicology
E-Tox
Pharm/Tox
Study
Summaries
Toxicology
Data Bases
E-Reviews
Freedom of
Information
Files
A Knowledge Base is the Combination
of Databases and Computational
Methods to
Discover Meaningful Relationships
The CDER Toxicology Knowledge
Base is a Prototype for an FDA
Knowledge Base
Estimating Potential Toxicity
E-Tox/SAR
Modeling
Molecular
Descriptors
Biological
Descriptors
Weight of
Evidence
Factors
Major Structure-Activity (SAR) Based
Predictive Models
• Expert Rule Based Methods
• Prior expert knowledge and mechanistic
hypotheses required
• Derek; Oncologic
• Statistical/Correlative Methods
• Little prior knowledge required. Computer
generated patterns and relationships from a
statistical analysis of a data set
• MCASE; Topkat
Representative Molecular Descriptors
• 2D molecular structure based clustering
• 2D molecular substructure clustering;
molecular fragmentation
• 3D rigid and flexible molecular
configuration clustering
• Physical chemical parameters, eg. Log P;
homolumo constants; electrotopographic
properties
Modeling Biological Descriptors
Major Sources of Error
• Inadequate size of control data set
• Inadequate representation of molecular
diversity (coverage)
• Over simplification, poor use of biological data
• Unbalanced representation of biological
activity
• Inadequate validation of predictive models due
to lack of studies not included in the control
data set
The Representation of Molecular Diversity
The Size and Diversity of Control Data Set
• Coverage: The FDA rodent carcinogenicity
data base contains more than 1000 compounds
that include both pharmaceuticals and nonpharmaceuticals
• Balanced representation: Approximately equal
number of positive and negative studies in the
FDA carcinogenicity database
• Validation: Availability of a large pool of new
studies improves the validation process
The Representation of Biological Activity
Two Year Rodent Carcinogenicity Studies
•
•
•
•
•
Male and female dose groups
Male and female untreated control groups
50+ animals/sex/group (400+ total)
40+ organ/tissue pathology analyses/animal
Relatively high spontaneous age related
background tumor rate
• Relatively high probability of some treatment
related findings
• Sensitivity/Specificity Issues
The Representation of Biological Activity
Modeling Rodent Carcinogenicity Studies
•
•
•
•
Four Study Cells
Male and Female Rats
Male and Female Mice
Each study cell can be considered an
independent study
• More than one positive study cell is
necessary to corroborate a positive
finding
The Representation of Biological Activity
Weight of Evidence and Data Quality
• Separate evaluation/modeling of male and female
rat and mouse study results (4 study cells)
• More positive cells=greater potency and
confidence
• A biologically relevant molecular descriptor is
one that is linked to positive findings in at least
two study cells
• The greater the number of compounds containing
a molecular descriptor associated with
carcinogenicity in the database, the greater the
degree of confidence in the finding
Assignment of Carcinogenic Potency
Compounds that induce trans-species tumors present the
highest degree of risk because they adversely alter mechanisms
that are conserved across species.
Tennant, Mutat. Res. (1993) 286, 111-118.
TUMOR FINDINGS
Trans-species, multiple site
(++++Potent)
Single/trans-gender, multiple site
(+++Potent)
Trans-species single site
(++Potent)
Trans-gender, single site
(+Weak)
Single gender, single site
(Equivocal)
No findings
POTENCY
( log units)
70-79
50-69
40-49
30-39
20-29
10-19
THE FDA-CDER INFORMATION CYCLE
Submission
Approval
Review
Drug R &
D
Applications
R&D
Decision Support
Guidances
E-Tox
Institutional Memory
IND Reviews
NDA Reviews
Proprietary Data
Non-proprietary
Nonproprietary
Databases
Proprietary
Databases
From Pharma 2005: An Industrial Revolution in R&D
Pricewaterhouse Coopers
Now
Primary
Science
Labs/Patients
Secondary
Science eR&D
in-silico
computers
Transition
Primary
Science
Secondary
Science
Future
Exp. Science
eR&D
computers
Confirmatory
Science
Labs/Patients