Transcript No Slide Title

The Paradigm Shift from Traditional
to Virtual
• Stephen K. Durham, PhD
• Department of Lead Safety
Assessment
Factors Influencing Change
• Technology
– Combinatorial Chemistry
– High-throughput Screens
– Computational Power
– Genomic revolution
• Escalating Costs
The Changing Paradigm
Traditional
(Sequential)
Current
(Parallel)
MTS
HTS
Potency
Potency
Selectivity
Specificity
Selectivity
Specificity
Functional Activity
ADME/Pharmaceutics
Safety
Functional Activity
DEVELOPMENT
Future
(Knowledge-Based)
Computational
Design and
Screening of
Virtual
Libraries
In Vitro Confirmation
What Are the Key Toxicological Liabilities
Affecting Drug Development?
•
•
•
•
•
•
Genotoxicity
Carcinogenicity
Teratogenicity
Liver Toxicity
Extrahepatic Toxicity
P450 Induction
Why Do We Want to Find Out the
Liabilities Early?
Studies Required for an NDA:
•
•
•
•
•
•
•
•
•
•
•
Genotoxicity Studies (in vitro and in vivo).
Single-Dose Studies in Mice and Rats.
Two-Week, One-Month or Three-Month Studies in Rats and Dogs.
Six-Month Study in Rats.
Chronic (6 – 12 Month) Study in Dogs.
Segment I, II, and III Reproductive Toxicity Studies in Rats and/or Rabbits.
Palatability and 3-month Range-Finding Studies for Carcinogenicity Studies.
Carcinogenicity Studies in Mice and Rats.
Local Tolerance Study in Rabbits.
Antigenicity Study in Guinea Pigs.
Others as needed.
How Do We Address Safety Issues
Until Virtual is a Reality?
Tiered Multivariate Analysis
9
No of Compounds
8
In Vitro Studies
7
6
In Vivo Studies
5
Computational Analyses
4
3
2
1
0
0
5
10
15
Stage of Development
20
25
In Silico Predictive Toxicity
Computational programs ultimately fulfill the
requirement for determining liabilities at the early
stages of discovery
Mutagenicity
N
N
N
N
O
Carcinogenicity
O
Reproductive
Toxicity
In Silico Predictive Toxicity
ADVANTAGES
STRUCTUREBASED
RULES-BASED
DISADVANTAGES
TOPKAT
“Best-fit” data set and
non-proprietary FDA
database
Closed system
No metabolite prediction
MULTICASE
Proprietary FDA data
Open system
Metabolite prediction
Dated operating system
Different modules not
well-integrated
DEREK
Rules upgradeable
Clear, concise
explanation of
structural liability
Difficult to identify new
rules
Metabolite prediction not
fully-implemented
Rigid rule interpretation
No negative controls
Approaches to Analysis
STRUCTUREBASED
Collects
molecular
fragments and
descriptors
Calculates values of chemical
descriptors
Compares to known compounds
Reports probability of being
member of toxic class in a
multifactorial statistical analysis
Identifies potential structural
liabilities
RULES-BASED
Inspects
molecule for
known structural
liabilities
Prepares a summary report of
findings
Identifies specific structural
liabilities
Gives references
Typical TOPKAT Output
Typical Multicase Output
Typical DEREK Output
Size Does Matter
• Large Pharma Advantages
– Robust Institutional Dataset
– Extensive Logistical Resources
• Biotech Advantages
– Flexible and Agile
– Risk Tolerant
– Strong Academic Ties
“Quid pro quo”
Internal Evaluation Protocol
• Comparative computational toxicological
evaluation using a pharmaceutical data set
• Analysis of compounds not existing in training
dataset (MCASE/ TOPKAT)
• Include BMS “institutional” data
• Compliance for robustness and chemical diversity
Acceptability Criteria for
Computational Analysis
• 85% Concordance
• Require low false negatives (high specificity)
• Willing to accept false positives followed by rapid
in vitro verification
“Still looking for Utopia”
Post-computational Verification:
Acceptability Criteria for In Vitro Analysis
•
•
•
•
High concordance
Require low false negatives and false positives
Small compound requirements
Moderate through-put with rapid results
“Reliable in vitro assays are necessary to confirm
computational predictions”
The Changing Paradigm
MUTAGENICITY
CARCINOGENICITY
TERATOGENICITY
TRADITIONAL
Ames:
1 Month
Segment II
Assays:
7 Months
PARADIGM
SHIFT
DNA
Damage
Assay:
Hours
Rodent
Carcinogenicity
Program:
4 Years
Cell
Transformation
Assay:
Weeks
Cell
Differentiation
Assay:
Days
Emerging Technologies
Usefulness
Limitations
Transcriptome
Many endpoints
Knowledge of human
genome
May explain mechanism
Proteome
Closer to function
Coordinate analysis with
expression-profiling
Metabonome
Medium to high throughput
Simple sample preparation
Identify organ-specific
changes
Kinetic animal or cell culture
studies
Expensive
Very early stage of validation
Small current learning set
Need to study many known
toxic compounds
Need to correlate with
functional changes
Technology not amenable to
high throughput
Acknowledgements
• Genetic Toxicology, Drug Safety Evaluation
– Andrew Henwood
– Larry Yotti
• Lead Safety Assessment
– Oliver Flint
– Greg Pearl
• Structural Biology and Modeling
– Deborah Loughney
– Jonathan Mason
– Roy Vaz