Towards a Clinically Relevant

1

• Towards a Clinically Relevant
• Systems Biology Model
• for the
• Human Immune System
• Mark Halling-Brown
• ISMB 09/11/2009

2
Overview

• Introduction to Complexity of Immune System
• Introduction to ImmunoGrid
• Agent-based cellular automata C-IMMSIM
• Improvements to C-IMMSIM (in progress)
• Data Repository
• Introduction of molecular level data (in
  progress)

http//www.immunogrid.org
3
Immune System Characteristics
Extremely Complex System

• Multi-levelled (lines of defence)

4
Layers of the Immune System
Systems
Organs
Cells
Intra- cellular
5
Immune System Characteristics
Extremely Complex System

• Multi-levelled (lines of defence)
• Diverse (molecules, cells)

6
Cells of the Immune System
Stem cell
Myeloid progenitor
Lymphoid progenitor
Monocyte
B cell
T cell
Eosinophil
Neutrophil
Natural killer cell
Mast cell
Basophil
HUGE NUMBERS OF CLONES
Plasma cell
Memory B cell
Helper T cell
Killer T cell
Dendritic cell
Macrophage
ENORMOUS REPETOIRE
V. Brusic 2007
7
Immune System Characteristics
Extremely Complex System

• Multi-levelled (lines of defence)
• Diverse (molecules, cells)
• Temporal (seconds to years)

8
Phases of Adaptive Immunity
9
Immune System Characteristics
Extremely Complex System

• Multi-levelled (lines of defence)
• Diverse (molecules, cells)
• Temporal (seconds to years)
• Molecular Interactions (Antigen Recognition)

10
Molecular Detail
T cell Binding
Proteasomal Cleavage
Cathepsin Cleavage
TAP Transporter Binding
MHC class II Peptide Binding
MHC class I Peptide Binding
11
Immune System Characteristics
Extremely Complex System

• Multi-levelled (lines of defence)
• Diverse (molecules, cells)
• Temporal (seconds to years)
• Molecular Interactions (Antigen Recognition)

Key modelling issue Complexity versus
simplification
Focussing
12
ImmunoGrid
Towards a Clinically Relevant Systems Biology
Model for the Human Immune System
CINECA (project coordinator) CNR (National
Research Council) University of
Bologna University of Catania
CNRS France (Centre for national research)
University of Queensland
Birkbeck
Denmark Technical University
http//www.immunogrid.org
Interuniversity Consortium of Northeastern Italy
for Automatic Computing
13
ImmunoGrid Aims

• Develop a virtual human immune system

• Simulate immune processes at natural scale,
• connecting molecular level interactions with
  system
• level models

• Provide tools for applications in clinical
  immunology,
• the design of vaccines and immunotherapies
• - Viral HIV (antiretroviral HIV therapy),
  Influenza
• - Immunotherapies for cancer SimTriplex
  vaccine
• - Optimisation of vaccine schedules

SimTriplex Prof Pier Luigi Lollini Prof Santo
Motta
14
The starting point C-ImmSim
Developed by Dr Filippo Castiglione at Italian
National Research Council (CNR) C-IMMSIM is an
agent-based, Lattice Gas (diffusion dynamics)
model for immune response, various cellular
entities, various molecular entities and various
cytokines.
2D simulation Virus diffusing on a grid
2D simulation Bacteria growing on a grid
15
Conway's Games of Life
16
Standard model entities
17
Current Uses SimTriplex
Triplex vaccine is an immunopreventive cellular
vaccine for mammary carcinogenesis Produced an
almost complete protection from mammary
carcinogenesis in HER-2/neu transgenic female
mice using a chronic vaccine schedule.
Simulation are being run to predict the
outcome of varying the vaccine schedule
18
Current Uses SimTriplex
19
Main tasks for ImmunoGrid
The focus is to improve and extend the simulator
to more accurately capture real behaviour

• 1) Scale up the simulations and move to 3D space
  (scaling)

20
Main tasks for ImmunoGrid
The focus is to improve and extend the simulator
to more accurately capture real behaviour

• 1) Scale up the simulations and move to 3D space
  (scaling)
• 2) Introduce molecular level detail into the
  simulations (complexity)
• 3) Increase the number of molecule types
  (complexity)

21
Current Molecular Affinity
Pinteract f(match)
Molecular affinity
0000 1111 0101 1010
1111 0000 1010 0101
22
Introducing new representation

• Represent interfaces with amino acids instead of
  bit strings
• Introduce immunoinformatics tools to access
  binding affinities instead of Hamming distance

23
The Problems

• The problems presented to Birkbeck are
• Increasing the complexity and scale of
  simulations will require an increase in the
  number and power of machines we can access
• Integration of molecular level detail and
  introduction of new and improved entities
  requires data gathering and storage.

24
GRID Implementation
Application Hosting Environment Prof. Peter
Coveney
Recent developments have made this approach
possible
RSL - Resource Specification Language
JSDL - Job Submission Description Language
NJS - Network Job Supervisor
25
The Problems

• The problems presented to Birkbeck are
• Increasing the complexity and scale of
  simulations will require an increase in the
  number and power of machines we can access
• Integration of molecular level detail and
  introduction of new and improved entities
  requires data gathering and storage.

26
Introducing Molecular Detail

• Solve the system for each cell and time step
  during the simulation

Real Predicted Data
2) Precompute possible outcomes before the
simulation
Real Predicted Data
Sdasdas Dasdasdas Fsdfsdfsd Gfgfdgfdgfdgfd Sdasdas
dasdasdasdas Dfsdfsdfsd Fdsfsdfsd Dfsdfsdfsdf fds
fsd
27
Data Repository
A database repository is required to hold
parameters data in order to aid the
initialisation of simulations and storage of the
results

• Organ details
• Sizes, volumes
• Cell
• General details
• Repertoires
• Lifecycles
• Concentrations
• Flow rates

• Molecules
• Chemokines
• Cytokines
• HLA
• Alleles
• Population frequencies
• Binding preferences
• Binding half lives
• Results
• .

28
Data Repository
29
Data Repository
30
Where to get data from?
Adrian Shepherd Renata Kabiljo
DTU
ANRI
31
Acknowledgements

• Birkbeck
• David Moss
• Adrian Shepherd
• Clare Sansom
• Matt Davies
• Renata Kabiljo
• Collaborators
• Paul Travers, Anthony Nolan Research Institute
• Darren Flower, Jenner Institute for Vaccine
  Design
• Mark Coles, University of York

• ImmunoGrid Partners
• CINECA Elda Rossi
• Brisbane Vladimir Brusic
• CNR Filippo Castiglione
• CNRS Marie-Paule Lefranc
• DTU Soren Brunak
• Bologna Pierre-Luigi Lollini
• Catania Santo Motta

http//www.immunogrid.org
32
Data Type Examples
Cell Organ Species Protein Peptide Complex Recepto
r MHC Group
part of binds to from type of interacts
with found in presented by kind of
ph Temperature Pressure Age Sex Infected EBV
33
Current Uses SimTriplex
Triplex vaccine is an immunopreventive cellular
vaccine for mammary carcinogenesis Produced an
almost complete protection from mammary
carcinogenesis in HER-2/neu transgenic female
mice using a chronic vaccine schedule.
Simulation are being run to predict the
outcome of varying the vaccine schedule
34
Current Uses SimTriplex
35
Vaccination Schedule
The best vaccination schedule for Triplex
vaccine protocol that minimizes the number of
vaccinations without reducing tumor prevention
efficacy in comparison to the Chronic protocol.
36
Optimal Vaccination Schedule
Redundant?
Insufficient
37
Recruitment of Mice

• Completed recruitment of 8 mice per experimental
  group in Chronic, Genetic, Heuristic, Early,
  ongoing for Untreated (4 mice so far)
• Early vaccination protocol already concluded,
  other protocols ongoing

38
Interim Conclusions

• First preclinical test of simulator-designed
  vaccination protocol is ongoing
• Vaccines are inducing the (expected) strong
  antibody response
• Survival results will be available in 6-12 months

39
Where to get data from?
Text mining
Processing literature (HTML/PDF) in order to
extract useful/relevant information We have
developed a service to extract paragraphs and
tables from a PDF document
40
Where to get data from?
Data will be extracted from other
databases/services CBS IMGT NCBI
dbMHC AlleleFrequencies .
Other Databases/ services
41
Where to get data from?
Not much fun, but specific items of data can be
found by trawling literature manually
User contributions
ANRI is providing us with quantities of cell
concentration data lots more
The approach we want to take will allow for
registered users to add new parameters to the
repository
Manually gather
Personal Contacts
42
Parallelisation
Assignment of lattice sections among processors
43
Parallelisation
Assignment of immune entities among processors
44
Parallelisation
Each lattice section is treated in a
decentralised fashion by every processor in the
simulation
45
Parallelisation
Inter-domain communication Entities can migrate
from one lattice section to another
46
Why model the Immune System?

• Experimentation can sometimes have limitations
• number of experiments
• time-scale (e.g. HIV infection)
• ethical concerns
• Computer models can complement experiments by
  focusing attention on where experimentation is
  required
• Feedback from experiments is essential to build
  the models

47
First version of C-ImmSim models cells in 2D space
C-ImmSim
48
GRID Definition
Grid computing is distributed computing
performed transparently across multiple
administrative domains Peter Coveney
49
GRID Rationale

• Distribute simulations
• Allow exploration of parameter space
• Simulate allelic variation in populations
• Access our consortiums geographically
  distributed resources
• Access to large machines for computationally
  expensive simulations

One of our partners (CINECA) is providing access
to a super computer
50
Parallelisation of Simulations

• Parallelisation of simulator
• 2. Farming of independent simulations

Important distinction between parallelisation
and Grid
51
Examples of Technology to Access
52
Layers of the Immune System
Systems
Organs
Cells
Intra- cellular
53
Data Repository
Simulation Request
54
Data Repository
Unified Web Service Interface
Simulation Request
55
Data Repository
Data Repository
Experimental Predicted Data
Unified Web Service Interface
Workflow Web Service
Simulation Request
56
Data Repository
Data Repository
Experimental Predicted Data
Immunoinformatics Services
Unified Web Service Interface
Workflow Web Service
NetChop
Simulation Request
Bepipred
NetMHC
..
Prof Søren Brunak, DTU is providing many
immunoinformatic services
57
Data Repository
Data Repository
Experimental Predicted Data
Unified Web Service Interface
Immuno
Workflow Web Service
Simulation Request
NN
Motif
SVM
Matrices
NN
SVM
58
Data Repository
Data Repository
Experimental Predicted Data
Unified Web Service Interface
Workflow Web Service
Simulation Request
C-ImmSim Simulation Starts
59
Data Repository
Data Repository
Experimental Predicted Data
Unified Web Service Interface
Workflow Web Service
Simulation Request
C-ImmSim Simulation Starts
60
IMS Multilayered architecture

• 1st Layer skin and physiological conditions (pH,
  temperature)
• 2nd Layer Innate
• (scavenger cells clean pathogens and debris)
• 3rd Layer Adaptive (Antigen recognition)