Transcript ImmunoGrid – The virtual human immune system project

ImmunoGrid – The Virtual Human
Immune System Project
Andrew Emerson, High Performance Systems,
CINECA, Italy
ImmunoGrid: Definition
“…a 3 year project funded by the European Union which
will establish an infrastructure for the simulation of the
immune system that integrates processes at molecular,
cellular and organ levels.”
To be designed for applications that support clinical
outcomes such as design of vaccines and
immunotherapies and optimization of immunization
protocols.”
The result will be a “simulator” which will allow us to
model the human immune system on the natural scale.
http://www.immunogrid.org
2
The Consortium
 CINECA, Bologna.
 University of Queensland.
 CNR, Rome.
 CNRS, Montpellier.
 Technical University of
Denmark
 Birkbeck College, University
of London.
 Dept of Experimental
Pathology,
University of Bologna.
 University of Catania,
Catania.
http://www.immunogrid.org
Expertise
System modelling
Molecular level
modelling
Immune system
standardized
concepts
Grid technology,
high performance
computing
Clinical/Pre-clinical
aspects
(users of the system)
Immunoinformatics
and databases
3
The Human Immune System
• A complex and adaptive learning system which has evolved to
defend an individual against foreign invaders.
• Operates at multiple levels: from molecule to cell, organ and
organism
• Some cells exist in many forms which differ in their specificities:
 >1013 MHC class I haplotypes
 107-1015 different T-cell receptors
 1012 B-cell clonotypes in each individual
 1011 possible linear MHC-binding epitopes composed of nine
amino acids
 >>1011 different conformational epitopes
combinatorial
 >109 combinatorial antibodies
complexity
http://www.immunogrid.org
4
Immunomics: Successes and failures
Vaccines have been instrumental in controlling
many diseases
 Eradication of smallpox
 Near eradication of polio
But many diseases are still poorly protected
against
 e.g. failure of the BCG vaccine against TB in
some communities
http://www.immunogrid.org
5
Why model the Immune System ?
In particular why on the natural scale ?
• Computer models complement and replace actual
testing or experiments
• Experimentation is expensive and has limitations:
 some experiments cannot be performed
 number of experiments that can be performed
 time-scale (e.g. HIV infection)
 ethical concerns (e.g. there are strict rules as what
experiments can be performed in humans)
http://www.immunogrid.org
6
Models
We use agent-based models derived from the
ImmSimm program of Celada-Seiden:
HIV-1 infection
C-ImmSimm
(Castiglione et al)
ImmSimm (Celada et al.)
SimTriplex (Catania
Mouse model)
describes the immune
system entities with their
different biological states
and the interactions
between different entities
(Pappalardo et al.)
http://www.immunogrid.org
cancer immunoprevention
vaccines
8
Modelling the immune system
Self-peptides
Antigens
Thymus
non-self
Th, CTL
Virus, bacteria, …
Thymocytes
Positive/Negative
selection
B
Th
T
CTL
MA
DC
Ag
Bone marrow
All cells
B, MA, DC, …
Simulation space
(secondary organ)
http://www.immunogrid.org
9
Modelling the Immune System
These models have proved to be very successful in reproducing
the progress of AIDS or tumour progression in vivo experiments
of HER2/neu transgenic mice.
They have also been extended by the consortium since the start
of the project, e.g. to allow chemotaxis (i.e. diffusion along a
chemical gradient) or 3D simulation spaces.
Infection site
Lymph node
Lymph channel
http://www.immunogrid.org
10
Modelling the Immune System –
computational requirements
Simple cases can run on a pc in mins but our project aims
to go much further:
 timestep from 8h → mins (for allergies)
 timesteps from 103 → 108
 no. of entities from 104 → 1015
 increase simulation time from 1 yr (mouse) → many
yrs (e.g. human)
 more complex simulation spaces (e.g. representing
real organs)
 from 1 to many individuals for population analysis
The aim is to reach the natural scale of the human immune system.
http://www.immunogrid.org
11
Motivation for the Grid
Need for high computational power to match the
combinatorial complexity and natural scale of the
human immune system.
Comment: The programs
have been parallelised,
why not just use a parallel
computer ?
http://www.immunogrid.org
12
Motivation for the Grid
• Distribute simulations
 Allow exploration of parameter space
• Access the consortium’s geographically
distributed resources
• Access to large machines for
computationally expensive simulations
• Not limit ourselves to one solution. e.g
super computer
 Cannot guarantee access to super
computer
• Different levels of usage, some not relevant
for a supercomputer.
http://www.immunogrid.org
13
Motivation for Grid
We anticipate different levels of access to the simulator
which will vary the computational requirements of the
simulator.
Educational
Small, computationally inexpensive simulations.
Complex GRID solution unnecessary. Web
services more appropriate.
Standard
Larger, computationally expensive simulations.
GRID solution required to distribute simulations
around available resources.
Developer
Large, computationally expensive simulations.
GRID solution required. Access to super
computer desirable.
http://www.immunogrid.org
14
Grid infrastructure - concepts
What our grid will not do:
• No attempt to distribute a single simulation
over different grid nodes (too difficult)
• No attempt to design new middleware
 we will use existing solutions
• No attempt to install the same middleware
on each node
 partners’ resources too diverse and
may violate local policies.
http://www.immunogrid.org
15
Grid infrastructure - concepts
What our grid must do:
• Able to integrate seamlessly the available
resources, modifying as little as possible local
infrastructures and policies.
• Must conform to standards.
• Allow the different types of usage, including access
to supercomputers and other Grids.
• Single point of access and easy-to-use graphical
interface (essential for educational purposes)
Our grid will be a kind of “virtual organisation” and user-driven
http://www.immunogrid.org
16
Grid implementation
We are still in the design phase but we intend to utilise
and amalgamate multiple different GRID solutions
GridSAM
AHE
This approach will provide flexibility and
redundancy
http://www.immunogrid.org
18
Grid Implementation - resources
Resource Type
Technology to Access
Groups’ individual PCs
Web service/GridSAM
Groups’ clusters
Web service/GridSAM/UNICORE
CINECA (Linux + IBM SP
clusters)
UNICORE/EnginFrame
National Grid Service
Globus/GridSAM
DEISA
UNICORE/DESHL/GridSAM
TeraGrid
Globus
http://www.immunogrid.org
19
Proposed GRID Implementation
Web
Interface
Job
launcher
CINECA
e.g Birkbeck college
AHE Client
UNICORE
AHE Server
JSDL
GridSAM
FORK
Web
Service
GridSAM
GridSAM
FORK
Group
Local
GridSAM
GATEWAY
RSL
NGS
GATEWAY
NJS
Group
Cluster
DEISA
CINECA
GLOBUS
http://www.immunogrid.org
20
Interfaces
Web interface will provide access to simulations.
One interface will serve each level of access.
Educational
General interface
Standard
Job Launcher
Developer
Job Launcher could act as a simple brokering agent
http://www.immunogrid.org
21
Application Hosting Environment
Web Interface
Job launcher
AHE Client
UNICORE
AHE Server
JSDL
GridSAM
FORK
Web Service
GridSAM
GridSAM
FORK
Group Local
GridSAM
GATEWAY
RSL
NGS
GATEWAY
NJS
Group
Cluster
DEISA
CINECA
GLOBUS
http://www.immunogrid.org
22
Application Hosting Environment
http://www.immunogrid.org
23
UNICORE
Web Interface
Job launcher
AHE Client
UNICORE
AHE Server
JSDL
GridSAM
FORK
Web Service
GridSAM
FORK
Group Local
GridSAM
GridSAM
FORK
Group Local
GATEWAY
GATEWAY
DEISA
CINECA
RSL
NGS
GLOBUS
UNiform Interface to COmputing Resources
http://www.immunogrid.org
24
UNICORE
 Allows access to:
 DEISA
 CINECA (Linux and IBM SP clusters)
 OGSA middleware (WSDL & SOAP based web
services)
http://www.immunogrid.org
25
Summary
We are now entering a critical phase in the project
as work begins on realising the principal objective,
i.e. the virtual human immune system simulator.
The details of the Grid implementation is still being
discussed but the overall strategy is clear.
Please see www.immunogrid.org for information
and if you wish to try out educational versions of
the simulators. Feedback is very welcome!
http://www.immunogrid.org
26
Acknowledgments
All partners and associate members of the
ImmunoGrid consortium.
The European Union for financial support.
http://www.immunogrid.org
27