In silico Disease Models:

1

• In silico Disease Models
• Applications in Drug Discovery and Development
• Seth Michelson, Ph.D.
• VP, In Silico R D
• Davis, California
• Jan 16, 2004

2
Agenda

• Drug Discovery and Development
• Clinically relevant systems biology
• Clinically predictive biosimulation The
  PhysioLab platform
• Its structure
• Its capabilities
• Top Down Modeling
• An example of from target validation The size of
  dust
• Validating the models and exploiting the
  technology
• Modeling Predictability
• The PhysioLab Platform as an enabling technology
  
• The Patient Population Variability,
  Observability, and Clinical trial Design

3
Drug Discovery and DevelopmentPharma 101
4
An Overview of the Processes around Discovery R
D

• Target Identification
• Target Validation
• Target Selection
• Lead Identification/ Lead Development
• Lead Optimization

Genomics, serendipity, basic research, fast
follower
Gene expression, knockouts, in-vitro assays, in
silico driven bioassays

• HTS/uHTS
• Combinatorial
• Chemistry
• In vivo efficacy

SAR
Predictive Toxicology ??
5
Drug Development ProcessAttrition of Potential
NCEs
Total time - 10-15 years 800,000,000
6
Target Identification and Validation

• Where Do Targets Come from?
• Genomics and Expression Data
• Which tissues under what conditions?
• Smart People
• Ideas, good science, seredipity and 22
• Sometimes 2 2 5
• Fast Follower Strategy
• How do they decide which target is worth
  pursuing? Who is they?
• How do they validate it once they do?
• Which scientists are involved and what do they do?

7
Target Selction and Screening

• What are HTS and uHTS?
• Screening is the testing of a compound library
  against a well defined assay meant to describe
  the activity of status of a target
• HTS is an automated version of the screen 30,000
  to 40,000 compounds per day
• Infrastructure
• Inventory management and control
• Data Capture, Analysis, and Management
• Cell Cultures and plate preparations
• Assay development, automation, and
  miniaturization (cost per well)
• uHTS HTS in spades (150,000 to 200,000
  compounds per day)
• Infrastructure
• Assay Development and new technologies
• You dont screen to screen, you screen to follow
  up. How?
• Hints to Chemistry on where to begin
• Which scientists are involved and what do they do?

8
Target Selction and Screening

• What is Lead Development and Optimization about?
• Making a compound look like a medicine
• Multivariate criteria
• Set by Regulatory Oversight
• Safe
• Safety Pharm non-GLP
• Toxicology GLP Regulatory controls
• Effective
• Secondary In vitro screening
• Tertiary In vivo screening
• Set by Marketing pressures
• Dosing and scheduling
• Cost of goods
• Cost of manufacture process chemistry
• Medicinal Chemistry begins its magic
• Which scientists are involved and what do they do?

9
Associating Chemical Character with Biological
Activity
10
Segmenting Chemical Character Space Based on
Similarity of Structure
Chemical Character
Biological Activity

• Assay 2
• Data Set 1
• Data Set 2
• Data Set M

11
Distance and Perceived Distance Projection
into a Lower Dimensional Space
D1 ltlt D2
12
The Key to Ultimate Success is Understanding the
Biological Space More Completely
Chemical Character
Biological Activity
13
Target Selction and Screening

• Clinical Trials The ultimate proof
• Proving a compound is a medicine (GxP criteria
  and Regulatory Affairs oversight)
• Similar Multivariate criteria
• Set by Regulatory Oversight and reporting
  criteria
• Safe
• Adverse events profiles
• Toxicology GCP Regulatory controls
• Phase I trials normal healthy volunteers
• Effective
• Phase II first signs of efficacy in a patient
  population
• Phase III the first, true, large scale proofs
  of principal
• Informed Consent, Ethical constraints, trial
  design and the science it is meant to uncover
• Set by Marketing pressures
• Dosing and scheduling
• Cost of goods
• Cost of manufacture process chemistry
• Packaging and shelf life constraints
• Clincial Sciences and research begins

14
Phase I Clinical Trials

• Typically small number of patients (10s)
• Depending on drug, may be tested on healthy
  volunteers
• Safety and preliminary efficacy assessment
• Blood levels measured for dose ranging
• Surrogate markers also assessed

15
Phase II clinical trials

• Larger number of patients (50-100)
• Frequently absolute efficacy measurements
• First real indication of efficacy
• Usually used to select final formulation and
  dose(s)

16
Phase III clinical trials

• Usually large (100s-1000s) depending on drug
  and indication
• Usually multicenter
• FDA requires 2 well controlled trials as proof
  of efficacy
• Efficacy frequently measured against standard of
  care (as opposed to placebo)
• New Drug Application

17
Agenda

• Drug Discovery and Development
• Clinically relevant systems biology
• Clinically predictive biosimulation The
  PhysioLab platform
• Its structure
• Its capabilities
• Top Down Modeling
• An example of from target validation The size of
  dust
• Validating the models and exploiting the
  technology their predictions
• Modeling Predictability
• The PhysioLab Platform as an enabling technology
  
• The Patient Population Variability,
  Observability, and Clinical trial Design

18
Systems biology
50 of compounds Fail in Phase II
Bridging this gap means bringing these data into
a clinically relevant context
Genomics threatens to increase not only the
overall associated RD costs but also the average
cost per new chemical entity (NCE) or
drug Lehman Brothers, McKinsey Co.
19
Clinically relevant systems biology
20
Only at the patient level can systems biology
account for all the factors governing clinical
response

• Clinically Relevant
• Novel Therapy
• Biomarker
• Diagnostic

21
Agenda

• Drug Discovery and Development
• Clinically relevant systems biology
• Clinically predictive biosimulation The
  PhysioLab platform
• Its structure
• Its capabilities
• Top Down Modeling
• An example of from target validation The size of
  dust
• Validating the models and exploiting the
  technology their predictions
• Modeling Predictability
• The PhysioLab Platform as an enabling technology
  
• The Patient Population Variability,
  Observability, and Clinical trial Design

22
Biosimulation as supported by the Entelos
PhysioLab Platform
23
Innovative Technology Platformand
Multi-Disciplinary Methodology(How Entelos
translates biology into mathematics)
24
Hypotheses exploration drives decision-making
throughout the R D pipeline

• Knowledge Gaps
• Genetic and environmental factors of disease
  pathophysiology
• Unknown pathways

• Relative contributions of redundant pathways
• Compensating effects of feedback mechanisms
• Nonlinearities in signal transduction and
  integration (e.g., saturation, synergies)
• Kinetics of signalling and effector mechanisms
  (e.g., delays, persistence)

Consistent with Data (where appropriate)
Understand the Implications of Competing
Hypotheses
25
Making hypotheses exploration more systematic

• Knowledge Gaps
• Genetic and environmental factors of disease
  pathophysiology
• Unknown pathways
• Relative contributions of redundant pathways
• Compensating effects of feedback mechanisms
• Nonlinearities in signal transduction and
  integration (e.g., saturation, synergies)
• Kinetics of signalling and effector mechanisms
  (e.g., delays, persistence)

Data
Consistent with Data (where appropriate)
Understand the Implications of Competing
Hypotheses
26
Hypotheses for the progression of disease are
represented as perturbations of homeostasis
Health
Treatment
Disease
Healthy Homeostasis
Mild Disease
Physiologic State
Severe Disease
Time
27
Novel Systems Biology ApproachBehavior-Driven,
Top-Down Modeling
Functional Metabolic Pathway
Free-living Conditions
Behavior Complexity
Fasting
Metabolic Pathways
Model Detail
Genes
28
Novel Systems Biology ApproachBehavior-Driven,
Top-Down Modeling
Functional Metabolic Pathway
Free-living Conditions
Behavior Complexity
Fasting
Metabolic Pathways
Model Detail
Collect Additional Data
Sensitive Pathway
Genes
Protein Target
29
Agenda

• Drug Discovery and Development
• Clinically relevant systems biology
• Clinically predictive biosimulation The
  PhysioLab platform
• Its structure
• Its capabilities
• Top Down Modeling
• An example of from target validation The size of
  dust
• Validating the models and exploiting the
  technology their predictions
• Modeling Predictability
• The PhysioLab Platform as an enabling technology
  
• The Patient Population Variability,
  Observability, and Clinical trial Design

30
Applying the PhysioLab Platform to Target
Validation
Product Realization
Proof of Concept
Lead Optimization
Target Selection Validation

• PhysioLab platform Asthma
• Partner situation
• Novel gene available through strategic alliance
  with genomics partner
• Little data to assess details of gene function
• Urgent need for approach to prioritize
  opportunities from genomics investment
• Partner proprietary data
• Novel gene sequence
• Homologue sequence
• Limited expression data in healthy and diseased
  tissue
• Collaboration objective
• Validate protein(s) for which this gene codes as
  candidate drug targets
• Evaluate within common platform of PhysioLab in
  anticipation of large-scale gene
  evaluation/target prioritization collaboration

31
The Gene Push
32
The Gene Push
Target Validation Plan
Cellular Pathway Hypotheses
33
Agenda

• Drug Discovery and Development
• Clinically relevant systems biology
• Clinically predictive biosimulation The
  PhysioLab platform
• Its structure
• Its capabilities
• Top Down Modeling
• An example of from target validation The size of
  dust
• Validating the models and exploiting the
  technology their predictions
• Modeling Predictability
• The PhysioLab Platform as an enabling technology
  
• The Patient Population Variability,
  Observability, and Clinical trial Design

34
Modeling Disease from a Mathematical Point of
View
Cascade of biological path-ways linking target to
clinical endpoints
target
35
What should a mathematical model be able to do?

• Generate comparable results to those derived
  experimentally and make clinically relevant
  predictions

• Include most, if not all, important measurable
  physiological variables

• Include most, if not all, experimental/clinical
  protocols

• Provide enough detail to represent interventions
  at key steps in pathways quantitatively

• Help design better experiments and clinical
  trials

36
What does it mean to say a model is correct?
37
Constraining the model to the range of potential
solutions

• Solutions
• consistent with
• Expert judgment
• Engineering criteria, e.g., stability

38
Adding data constrains the space further
Solutions that match data set 1

• Solutions
• consistent with
• Expert judgment
• Engineering criteria, e.g., stability
• And Data Set 1

39
. . . and so on
Solutions that match data set 1

• Solutions
• consistent with
• Expert judgment
• Engineering criteria, e.g., stability
• And Data sets 1 2

Solutions that match data set 2
40
But what happens when the data and the model do
not agree?
Solutions that match data set 1
Solutions that match data set 3 Uh oh!
Model lacking depth or incomplete data

• Solutions
• consistent with
• Expert judgment
• Engineering criteria, e.g., stability
• And Data sets 1 2

Solutions that match data set 2
41
We refine the model
Solutions that match data set 1
Solutions that match data set 3
Refine or Expand Model

• Solutions
• consistent with
• Expert judgement
• Engineering criteria, e.g., stability
• And data sets 1, 2, 3

Solutions that match data set 2
42
Conceptual classification of knowledge
Pathways
PhysioLab Representation
P T S
Knowledge Gaps
Knowledge Nuggets
43
Exploiting That Classification

• PhysioLab Representation
• Systematic methodologies for exploring
• clinical relevance of
• Targets
• Biomarkers
• Diagnostics
• Dynamics

Pathways
P T S

• New Model Development
• More Definition
• Added Biological Structure
• More Pts.

Knowledge Gaps
Knowledge Nuggets
44
Systematic Interrogations
Pathways or Pathway Hypotheses
45
Drilling into interesting segments of the
interrogation matrix
Wet Lab Validation
In Silico Development Exploration
Data Analysis and Visualization
Clinical Data Collection
Knowledge Products

• Feedback
• New model development
• Refined dynamics

46
Where do the Data Come From?
Patient-Level Physiologic Pathway Network

• Snapshots of these dynamics are supplied by data
  from
• Gene Expression Arrays
• Protein Expression Technologies
• Bioinformatics Support Databases

47
Deriving The Appropriate Information from
Expression Data Requires Context
Control
Treated
Cell Biology Physiology
Cell Biology Physiology
Gene Expression
Gene Expression
48
A multivariate statistics pattern recognition
problem
IN CONTEXT For each of K conditions (e.g.
exptl protocols)
49
A multivariate statistics pattern recognition
problem
For each of K conditions (e.g. exptl protocols)
50
The value a theoretical infrastructure brings to
these interrogations is the context
51
Scenario 1 Driving the In Vitro Vector
Measure these physiological effects based on
guidance from Physiolab / CytoLab analysis and
dynamics
To this cell type (Choice of cell type derived
from systematic interrogation of the PhysioLab)
Do this assay
Under these Conditions
And do the math . . .
52
Scenario 2 Driving Experimental Strategy
Measure these physiological effects based on
guidance from Physiolab / CytoLab analysis and
dynamics
And for each cell type identified as key to the
disease pathophysiological pathway
For each input cue determined to be clinically
relevant in the PhysioLab
And from these data establish conditions that
drive confirmatory experimentation
53
Agenda

• Drug Discovery and Development
• Clinically relevant systems biology
• Clinically predictive biosimulation The
  PhysioLab platform
• Its structure
• Its capabilities
• Top Down Modeling
• An example of from target validation The size of
  dust
• Validating the models and exploiting the
  technology their predictions
• Modeling Predictability
• The PhysioLab Platform as an enabling technology
  
• The Patient Population Variability,
  Observability, and Clinical trial Design

54
(No Transcript)
55
Which One?
56
An Issue of Observability
Modeled In silico
Clinical Readout
Modeled In silico
Directly Observable
Indirectly Observable
57
Modeling and Experimentation Systematically
formulating and testing hypotheses throughout
the R D pipeline
58

• Thank You
• www.entelos.com