Transcript Patient-specific medicine

Integrated e-Infrastructure for
Distributed, Data-driven, Dataintensive High Performance
Computing: Biomedical
Requirements
Peter V Coveney
Centre for Computational Science
University College London
Integrating the Strengths of the e-Research Community,
NeSC, Thursday, 10th March 2011
Contents
• Computational Biomedicine
– HIV/AIDS
– Cardiovascular medicine
– Cancer
• ICT, e-Health and the
Virtual Physiological Human
• Infrastructure support
• Shortcomings in UK infrastructure
• Major policy hurdles
•
UCL CLMS initiative
• Conclusions
UCL Projects
•
•
•
•
•
•
VPH Network of Excellence – EU (€8M); no HPC
ContraCancrum – EU (€3.4M); no HPC
VPH-Share – EU (€10.7M); no HPC
P-Medicine – EU (€13.7M); no HPC
INBIOMEDVision – EU (€2M)
MAPPER – EU (€2M); no HPC
• A new approach to Science at the Life Sciences
Interface – EPSRC (£4M) + HECToR
• Large Scale Lattice-Boltzmann Simulation of
Liquid Crystals – EPSRC (£800K) + HECToR
Patient-specific medicine
• ‘Personalised medicine’ - use the patient’s specific profile to better
manage disease or a predisposition towards a disease
• Tailoring of medical treatments based on the characteristics of an
individual patient
Why use patient-specific approaches?
• Treatments can be assessed for their effectiveness with respect to the
patient before being administered, saving the potential expense of
ineffective treatments
Patient-specific medical-simulation
• Use of genotypic and or phenotypic simulation to customise treatments
for each particular patient, where computational simulation can be used
to predict the outcome of courses of treatment and/or surgery
See: P. V. Coveney et al (eds), Interface Focus, Theme Issue on VPH Vol. 1, No. 3 Online 25th April 2011
Medical/clinical domain I: HIV/AIDS
HIV-1 Protease is a common target for HIV drug therapy
•
Enzyme of HIV responsible for
protein maturation
• Target for Anti-retroviral
Inhibitors
• Example of Structure Assisted
Drug Design
• 9 FDA inhibitors of HIV-1
protease
So what’s the problem?
• Emergence of drug resistant
mutations in protease
• Render drug ineffective
• Drug resistant mutants have
emerged for all FDA inhibitors
Monomer B
101 - 199
Glycine - 48, 148
Monomer A
1 - 99
Flaps
Saquinavir
P2 Subsite
Leucine - 90, 190
Catalytic Aspartic
Acids - 25, 125
C-terminal
N-terminal
 Integrate simulation with conventional clinical decision support systems to
refine results
Medical/clinical domain II: Grid enabled
neurosurgical imaging using simulation
The goal: to simulatelarge scale patient specific cerebral blood
flow in clinically relevant time frames
Objectives:
•To study cerebral blood flow using patient-specific image-based
models.
•To provide insights into the cerebral blood flow & anomalies.
•To develop tools and policies by means of which users can better
exploit the ability to reserve and co-reserve HPC resources.
•To develop interfaces which permit users to easily deploy and
monitor simulations across multiple computational resources.
•To visualize and steer the results of distributed simulations in real
time
Yield patient-specific information which helps plan embolisation of arterio-
venous malformations, aneurysms, etc.
M. D. Mazzeo and P. V. Coveney, Computer Physics Communications, 178, (12), 894-914, (2008). DOI: 10.1016/j.cpc.2008.02.013.
Medical/clinical domain III: ContraCancrum
Multi-level data
Multi-level Modelling
Two dedicated clinical studies in
ContraCancrum, one in glioma and
one in lung cancer (200
cases/year)
Schedule 1
Schedule 2
Schedule …
Schedule n
Multi -level Models of Cancer
Other clinical
data needed
Prediction of the best treatment schedule / schema
Clinically Oriented Translational Cancer Multilevel Modelling
http://www.contracancrum.eu
Virtual Physiological Human
• Funded under EU FP 7; ~ €250M
• 20 projects: 1 NoE, 5 IPs, 11 STREPs, 3 CAs.
“a methodological
and technological
framework that,
once established,
will enable
collaborative
investigation of
the human body
as a single
complex system
...”
Networking
NoE
VPH-Share Overview
HIV
Heart
Aneurisms
Musculoskeletal
VPH-Share will provide the organisational fabric (the infostructure), realised as a series
of services, offered in an integrated framework, to expose and to manage data,
information and tools, to enable the composition and operation of new VPH workflows
and to facilitate collaborations between the members of the VPH community.
€11M, 2011-2015, EU FP7 – Promotes cloud technologies
p-medicine
Multi-level disease modeling
Disease
Modelling at the
molecular Level
Disease
Modelling at the
cellular Level
G S G M
G
1
0
2
N A
Disease
Modelling at the
tissue/organ
Level
Multi-scale therapy
predictions/disease
evolution results
Predictive disease modeling in p-medicine will contribute to the optimization of
cancer treatment by fully exploiting the individual data of the patient.
p-medicine is focusing on Wilms tumor, breast cancer and acute
lymphoblastic leukemia
The p-medicine infrastructure supports both a generic seamless, multi-level data
integration purpose and a VPH-specific, multi-level, cancer data repository to
facilitate model validation and clinical translation through trials.
The infrastructure is scalable for any disease as long as predictive modeling is
clinically significant in one or more levels (from molecular to tissue level) and the
development of such models is feasible (i.e. there is enough understanding of the
biological mechanisms involved to develop them).
€13M, 2011-2013, EU FP7
Led by a clinical oncologist - Prof Norbert Graf!
Large scale data & computing
Models are built for use in clinical
decision support
results are needed in a timely
fashion
It is necessary to have the
possibility of seamlessly “plugging
in” resources for parallel and large
scale computing “here and now”
petascale computing is needed to
perform e.g.:
activities like drug binding affinity
determination
Blood flow through tumours
Gratis via VPH-NoE supervised VPH
Virtual Community allocations of
time on DEISA and, in future
PRACE via MAPPER, …?
Seamless access and integration of distributed, heterogeneous data
in a data warehouse repeatedly over time (≈ 200 GB / patient and time point)
07/04/2016
MAPPER: Objectives and
Challenges
MAPPER will develop computational strategies,
software and services for distributed multiscale
simulations across disciplines, exploiting existing
and evolving European e-Infrastructure.
Driven by seven exemplar multiscale applications,
MAPPER will deploy a computational science
infrastructure for distributed multiscale computing
on and across European e-Infrastructures.
By taking advantage of existing software and
services, MAPPER will deliver high quality
components aiming at large-scale, heterogeneous,
high performance multidisciplinary multiscale
computing, while maintaining ease of use and
transparency for end users.
MAPPER will advance state-of-the-art in high
performance computing on e-Infrastructures by
enabling distributed execution, across all
European e-Infrastructures, of multiscale models.
http://www.mapper-project.eu
VPH ToolKit
http://toolkit.vph-noe.eu
VPH Virtual Community on DEISA
+ euHeart in second wave, and other non-VPH EU projects
VPH was awarded 2 million standard DEISA core hours for 2009,
renewed for 2010 and 2011
• HECToR (Cray, UK)
• SARA (IBM Power 6, Netherlands)
DEISA-TeraGrid interoperability project has additional access to LRZ
VPH requires HPC and Data Integration
• Computational experiments integrated seamlessly into current clinical
practice
• Clinical decisions influenced by patient specific computations: turnaround
time for data acquisition, simulation, post-processing, visualisation, final
results and reporting.
• Fitting the computational time scale to the clinical time scale:
– Capture the clinical workflow
– Get results which will influence clinical decisions: 1 day? 1 week?
– This project - 15 to 30 minutes
• Development of procedures and software in consultation with clinicians
• Security/Access is major concern
• Need to integrate Data, Compute via Workflows
• On-demand availability of storage, networking and computational resources
Many of the projects we are involved in have
non-standard requirements with respect to
HPC service providers
•
•
•
•
•
•
•
Ability to co-reserve resources  HARC
Launch emergency simulations  SPRUCE
Consistent interfaces for federated access  AHE
Access to back end nodes: steering, visualisation
Lightpath network connections
Data integration from multiple sources  IMENSE
Support for software (ReG steering toolkit etc)
Individualized MEdiciNe Simulation
Environment  IMENSE
• Data repository – this is the key store for project data containing
all patient data, and simulation data derived from the patient data.
• Integrated web portal – this provides the central interface from
which users upload and access data sets, and analysis services.
The interface provides users with the facility to search for patient
data based on a number of criteria.
• Web Services – the web services platform implements required
data processing functions.
• Workflow environment – the workflow environment provides a
virtual experiment system, from which users can launch predefined workflows to automate moving data between the data
environment and multiple data processing services.
Coveney et al, “An e-Infrastructure Environment for Patient Specific Multiscale Modelling and Treatment”, preprint, 2011
IMENSE Environment
IMENSE Interface
Workflows
• GSEngine is a workflow orchestration engine developed
by the ViroLab project
• Can be used to orchestrate applications launched by AHE
• It allows services to be orchestrated using both point and
click and scripting interfaces
• Workflows stored in a repository and shared between
users
• Many of the aims of ViroLab similar to VPH-I projects, so
GSEngine will be useful here
Malawski et al, Future Generation Computer Systems, 26, (1), 138—146, 2010
Inside IMENSE: Integrating the components
Coveney et al, “An e-Infrastructure Environment for Patient Specific Multiscale Modelling and Treatment”, preprint, 2011
UK Infrastructural Failures

UK computing e-Infrastructure is crumbling.

Not a holding partner in PRACE.


Only one Tier-1 machine (with issues).






No Tier-0 site in the works.
HECToR has had several major failures, researchers seem
to have trouble using/trusting it, given its usage.
What’s happening next?
Tier-2 facilities are also being dismantled.
NGS core nodes being shut down!!
We cannot maintain a good level of e-Science
research without the infrastructure to support it
Relative to other countries we’re in full scale retreat!
Infrastructure in the UK is fragmented
NGS
Data
HPC
Networks
?
22
TeraGrid  eXtreme Digital (XD)
• Two sets of services:
– XES will provide a set of well-known (and standard) protocol
specifications and profiles
– CPS will support both the diversity of different services and
capabilities required by the community
• From the desktop to the largest machines!
• XD design is firmly tied to the user requirements of the science
and engineering research community.
• Presents the individual user with a common user environment
• Caters to both researchers whose computations require very
little data movement and those performing very data-intensive
computations.
• Will offer a highly capable service interface to “community user
accounts” such as science gateways
https://www.teragrid.org/web/about/xdtransition
We face major policy hurdles
• For our projects to be successful, we need
integrated compute, storage, networks and
services.
– HPC’s antediluvian policies prevent this from happening
They still have a batch job mentality!
• No coordinated allocations policies in the EU
– Need to apply for a project, then if successful apply for
compute access
 Can’t do project if compute application rejected!
Importance of connectivity



With limited national facilities, connectivity to
other countries becomes crucial.
1-10Gbit wide area networks are needed for
large simulations and data movements.
However, network provisioning is currently
extremely difficult and time-consuming.

Researchers end up having to request the links,
rather than resource providers.
Policy issues


E-science research has always required
changes in resource provider policies to thrive.
Support for advance machine and network
reservations.


Including urgent computing.
Improvements in accessibility and usability.
Support for Audited Credential Delegation.
 Interoperability between machines & infrastructures.
DEISA’s Failure to address this augurs poorly for the
future

Political issues

Streamlined procedures for UK or EU scientific
projects.

All-in proposals which, when accepted, grant
everything needed for a research project.


This includes funding for research as well as HPC resource
allocations.
More sensible service level agreements.

If a simulation uses multiple machines and one fails,
a full allocation refund should be given.
MAPPER Policy Document – copies available
Computational Life and Medical Science
The CLMS Network is 3 year initiative from September 2010
Management:
Supported by the
Provost's Strategic Fund
Director: P.V. Coveney
Dean’s Committee
Steering Committee
http://www.clms.ucl.ac.uk
CLMS Goals
CLMS brings together UCL researchers with clinicians from UCL partners
to develop shared data + compute + data transfer + application support services
 Integrated e-Infrastructure and Services
1. Expand UCL’s world-leading position in life and
biomedical sciences
2. Steering the collaboration with academic institutions:
within UCL, with UCLP and the NHS, UK-CMRI, Yale,
and others
3. Exploit initiatives in integrative biomedical systems
science from the UK Research Council, EU and others
around the world
4. Grow collaborations with industry, create business and
commercial opportunities, promote UCL IP licensing
5. Plan for the next stages of activity in computational life
and medical sciences at UCL
Conclusions
• Biomedical projects all put pressure on resource providers
to offer new services and new ways of working
• For interactive and urgent work the batch processing
model does not work
• The very conservative model adopted by HPC providers
proscribes their resources from being used in innovative
ways to do new science and engage new and different
kinds of users
• If HPC is to be exploited in computational biomedicine it
needs to be used in a way that fits in with the medical &
clinical workflow
• VPH and similar initiatives: Will only increase pressure for
non-standard services from resource providers