AAPM Foster July 2009

The present and future role of computers in medicine
Ian Foster
Computation Institute
Argonne National Lab & University of Chicago

Credits
Thanks for support from
Chan Soon-Shiong Foundation
Department of Energy
National Institutes of Health
National Science Foundation
And for many helpful conversations, Carl Kesselman, Jonathan Silverstein, Steve Tuecke, Stephan Erberich, Steve Graham, Ravi Madduri, and Patrick Soon-Shiong

Biology is shifting from being an observational science to a quantitative molecular science
Old biology: measure one/two things in two/three conditions
High cost per measurement
Analysis straightforward as little data
Enormously difficult to work out pathways due to inadequate data
New biology: measure 10,000 things under many conditions
Low cost per measurement
Analysis no longer straightforward
Payoff can be bigger: potential to understand a complex system
Ajay Jain, UCSF

Change health care
from
an empirical, qualitative systemof silos of information
to a model of
predictive, quantitative, shared,evidence-based outcomes

The health care information technology chasm
Health care IT [is] rarely used to provide clinicians with evidence-based decision support and feedback; to support data-driven process improvement; or to link clinical care and research.
Computational Technology for Effective Health Care, NRC, 2009

Digital power =
computing x communicationxstorage x content
Moore’s law
doubles
every 18
months
disk law
doubles
x every 12
months
fiber law
doubles
xevery9
months
community law
n
x 2
where n is # people
John SeelyBrown

(Intel)

Microprocessor trends
AMD
Dual core (April 2005)
Quad core (October 2007)
Intel
Dual core (July 2005)
Quad core (December 2006)
Sun
Niagara: 8 cores * 4 threads/core (November 2005)
Niagara2: 8 cores * 8 threads/core (August 2007)
IBM POWER6
2 cores * 4 threads/core (May 2007)
Tilera 64 cores
Dan Reed, Microsoft
11

Marching towards manycore
Intel’s 80 core prototype
2-D mesh interconnect
62 W power
Tilera 64 core system
8x8 grid of cores
5 MB coherent cache
4 DDR2 controllers
2 10 GbE interfaces
IBM Cell
PowerPC and 8 cores
Dan Reed, Microsoft
12

1E+17
multi-Petaflop
Petaflop
Blue Gene/L
1E+14
Thunder
Red Storm
Earth
Blue Pacific
ASCI White, ASCI Q
SX-5
ASCI Red Option
ASCI Red
T3E
SX-4
NWT
CP-PACS
1E+11
CM-5
Paragon
T3D
Delta
SX-3/44
Doubling time = 1.5 yr.
i860 (MPPs)
VP2600/10
SX-2
CRAY-2
Y-MP8
S-810/20
X-MP4
Peak Speed (flops)
Cyber 205
X-MP2 (parallel vectors)
1E+8
CRAY-1
CDC STAR-100 (vectors)
CDC 7600
ILLIAC IV
CDC 6600 (ICs)
IBM Stretch
1E+5
IBM 7090 (transistors)
IBM 704
IBM 701
UNIVAC
ENIAC (vacuum tubes)
1E+2
1940
1950
1960
1970
1980
1990
2000
2010
Year Introduced
The evolution of the fastest supercomputer
Argonne
My laptop

The Argonne IBM BG/P

www.top500.org
1
2
3-4
>128K

Simulation of the human arterial tree on the TeraGrid
G. Karniadakis et al.

Storage costs
(PC Magazine, Oct 2, 2007)

Informationbig bang
All informationper year
100Exabytes
Uniqueinformationper year
All human documentsproduced last 40,000 years(to 1997)
12 Exabytes
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
© Stuart Card, basedon Lesk, Berkeley SIMS, Landauer, EMC

Growth of Genbank(1982-2005)
Broad Institute

More data does not always mean more knowledge
Folker Meyer, Genome Sequencing vs. Moore’s Law: Cyber Challenges for the Next Decade, CTWatch, August 2006.

The Red Queen’s race
"Well, in our country," said Alice … "you'd generally get to somewhere else — if you run very fast for a long time, as we've been doing.”
"A slow sort of country!" said the Queen. "Now, here, you see, it takes all the running you can do, to keep in the same place. If you want to get somewhere else, you must run at least twice as fast as that!"

Computing ondemand
Public PUMA knowledge base
Information about proteins analyzed against ~2 million gene sequences
Back officeanalysis on Grid
Millions of BLAST, BLOCKS, etc., onOSG and TeraGrid
Natalia Maltsev et al.

1.6 Tbps
Cost perGigabit-Mile
320 Gbps
Moore’sLaw
2.5 Gbps
50 Mbps
1984
1994
1998
2000
1993
1998
2002
Opticalnetworkingbreakthrough!
Revolution
Capacity increase and new economics
Nortel

Optical switches
Lucent

Empiricism
Theory
Simulation
Data
New ways of knowing
300 BCE
1700
1950
1990
Enhanced by the power of collaboration

Quantitative medicine is the key to reducing healthcare costs and improving healthcare outcomes
Patients with same diagnosis

Quantitative medicine is the key to reducing healthcare costs and improving healthcare outcomes
Non-responderstoxic responders
Non-toxic responders
Patients with same diagnosis
Misdiagnosed

Leukemia and Lymphoma
After Mara Aspinall, GenzymeGenetics; Felix W. Frueh, FDA

Currently, 17% of Burkitt's Lymphoma are incorrectly diagnosed as Diffuse Large B Cell Lymphoma
Classic
Burkitt’sLymphoma
Atypical
Burkitt’sLymphoma
Diffuse Large
B Cell Lymphoma
Louis Staudt, National Cancer Institute

Classic Burkitt Lymphoma Cure Rate
Diagnosis
Atypical Burkitt Lymphoma Cure Rate
Treatment
Patients’ Actual Disease
Burkitt’sLymphoma
60 %
80 %
Intensive chemotherapy
Burkitt’sLymphoma
Diffuse Large B Cell Lymphoma
0 %
15 %
CHOPregimen
Burkitt’sLymphoma

Survival estimates for patients with Burkitt's Lymphoma
Best treatment for Burkitt’s Lymphoma
Best treatment for Diffuse Large B Cell Lymphoma
Dave et al, NEJM, June 8, 2006.

Burkitt’s
Lymphoma
Diffuse Large
B-cell Lymphoma
Classic Atypical
Louis Staudt, National Cancer Institute

Imaging biomarkers: Diffusion Tensor Imaging and brain injury
Kraus et al., Brain (2007), 130, 2508-2519

Enabling quantitative medicine
Collect a lot of patient data
Analyze data to infer effective treatments
Identify personalized treatment plans
Clinical practice
Basic research
Clinical trials

Challenges
Increasing volumes of data, types of data: genomics, blood proteins, imaging, …
New science and treatments are hidden in the data, not the biology (biomarkers)
Too much for the individual physician or researcher to absorb
… have to pay attention to cognitive support … computer-based tools and systems that offer clinicians and patients assistance for thinking about and solving problems related to specific instances of health care.
NRC Report on Computational Technology for Effective Health Care: Immediate Steps and Strategic Directions, 2009

Bridging silos to enable quantitative medicine
Basic research
ongoing investigative studies
Outcomes, tissue bank
screening tests
pathways
library
Clinical practice
Clinical trials
trial subjects, outcomes

Addressing urban health needs

Important characteristics
We must integrate systems that may not have worked together before
These are human systems, with differing goals, incentives, capabilities
All components are dynamic—change is the norm, not the exception
Processes are evolving rapidly too
We are not building something simple like a bridge or an airline reservation system

Healthcare is acomplex adaptive system
A complex adaptive system is a collection of individual agents that have the freedom to act in ways that are not always predictable and whose actions are interconnected such that one agent’s actions changes the context for other agents.
Crossing the Quality Chasm, IOM, 2001; pp 312-13

Non-linear and dynamic
Agents are independent and intelligent
Goals and behaviors often in conflict
Self-organization through adaptation and learning
No single point(s) of control
Hierarchical decomp-osition has limited value

We need to function in the zone of complexity
Low
Chaos
Zone ofcomplexity
Agreement
about
outcomes
Plan and control
High
Low
High
Certainty about outcomes
Ralph Stacey, Complexity and Creativity in Organizations, 1996

We need to function in the zone of complexity
Low
Chaos
Agreement
about
outcomes
Plan and control
High
Low
High

We call these groupingsvirtual organizations (VOs)
A set of individuals and/or institutions engaged in the controlled sharing of resources in pursuit of a common goal
But U.S. health system is marked by fragmented and inefficient VOs with insufficient mechanisms for controlled sharing
Healthcare = dynamic, overlapping VOs, linking
Patient – primary care
Sub-specialist – hospital
Pharmacy – laboratory
Insurer – …
I advocate … a model of virtual integration rather than true vertical integration … G. Halvorson, CEO Kaiser

The Grid paradigm
Principles and mechanisms for dynamic VOs
Leverage service oriented architecture (SOA)
Loose coupling of data and services
Open software,architecture
Engineering
Biomedicine
Computer science
Physics
Healthcare
Astronomy
Biology
1995 2000 2005 2010

The Grid paradigm and healthcare information integration
[Grid architecture joint work with Carl Kesselman, Steve Tuecke, Stephan Erberich, and others]
Manage who can do what
Make data usable and useful
Platform services
Name data and move it around
Make data accessible over the network
Data sources
Radiology
Medical records
Pathology
Genomics
Labs
RHIO

Enhance user cognitive processes
Security and policy
Incorporate into business processes
Transform data into knowledge
Integration
Platform services
Management
Publication
Data sources
Radiology
Medical records
Pathology
Genomics
Labs
RHIO

Cognitive support
Security and policy
Valueservices
Applications
Analysis
Integration
Platform services
Management
Publication
Data sources
Radiology
Medical records
Pathology
Genomics
Labs
RHIO

We partition the multi-faceted interoperability problem
Process interoperability
Integrate work across healthcare enterprise
Data interoperability
Syntactic: move structured data among system elements
Semantic: use information across system elements
Systems interoperability
Communicate securely, reliably among system elements
Applications
Analysis
Integration
Management
Publication

Publication:Make information accessible
Make data available in a remotely accessible, reusable manner
Leave mediation for integration layer
Gateway from local policy/protocol into wide area mechanisms (transport, security, …)

Imaging clinical trials
NeuroblastomaCancerFoundation
Childrens Oncology Group

Stephan Erberich,
Carl Kesselman, et al.

As of Oct19, 2008:
122 participants
105 services
70 data
35 analytical

Data movement in clinical trials
(Center for Health Informatics)

Community public health:Digital retinopathy screening network
(Center for Health Informatics)

Integration:Making data usable and useful
?
Adaptive approach
100%
Degree of communication
Loosely coupled approach
Rigid standards-based approach
0%
0% 100%
Degree of prior syntactic and semantic agreement

Integration via mediation

Map between models
Scoped to domain use
Multiple concurrent use
Bottom up mediation
between standards and versions
between local versions
in absence of agreement

Global Data Model
Query reformulation
Query in union of exported source schema
Query optimization
Distributed query execution
Query execution engine
Wrapper
Wrapper
Query in the
sourceschema
Alon Halevy, 2000

Analytics:Transform data into knowledge
“The overwhelming success of genetic and genomic research efforts has created an enormous backlog of data with the potential to improve the quality of patient care and cost effectiveness of treatment.”
— US Presidential Council of Advisors on Science and Technology, Personalized Medicine Themes, 2008

Published
The imagepyramid
Enrolled/ evaluated
Eligible patients
Created
Michael Vannier

Query and retrieve microarray data from a caArray data service:cagridnode.c2b2.columbia.edu:8080/wsrf/services/cagrid/CaArrayScrub
Normalize microarray data using GenePattern analytical servicenode255.broad.mit.edu:6060/wsrf/services/cagrid/PreprocessDatasetMAGEService
Hierarchical clustering using geWorkbench analytical service: cagridnode.c2b2.columbia.edu:8080/wsrf/services/cagrid/HierarchicalClusteringMage
Microarray clustering using Taverna
Workflow in/output
caGrid services
“Shim” services
others
Wei Tan et al.

Many many tasks:Identifying potential drug targets
2M+ ligands
Protein xtarget(s)
Benoit Roux et al.

6 GB
2M structures
(6 GB)
~4M x 60s x 1 cpu
~60K cpu-hrs
FRED
DOCK6
Select best ~5K
Select best ~5K
~10K x 20m x 1 cpu
~3K cpu-hrs
Amber
Select best ~500
~500 x 10hr x 100 cpu
~500K cpu-hrs
GCMC
ZINC
3-D
structures
Manually prep
DOCK6 rec file
Manually prep
FRED rec file
NAB scriptparameters
(defines flexible
residues,
#MDsteps)
NAB
Script
Template
DOCK6
Receptor
(1 per protein:
defines pocket
to bind to)
FRED
Receptor
(1 per protein:
defines pocket
to bind to)
PDB
protein
descriptions
1 protein
(1MB)
BuildNABScript
Amber prep:
2. AmberizeReceptor
4. perl: gen nabscript
NAB
Script
start
Amber Score:
1. AmberizeLigand
3. AmberizeComplex
5. RunNABScript
For 1 target:
4 million tasks500,000 cpu-hrs
(50 cpu-years)
end
report
ligands
complexes

DOCK on BG/P: ~1M tasks on 118,000 CPUs
CPU cores: 118784
Tasks: 934803
Elapsed time: 7257 sec
Compute time: 21.43 CPU years
Average task time: 667 sec
Relative Efficiency: 99.7% (from 16 to 32 racks)
Utilization:
Sustained: 99.6%
Overall: 78.3%
Time (secs)
Ioan Raicu et al.

The health care information technology chasm
Health care IT [is] rarely used to provide clinicians with evidence-based decision support and feedback; to support data-driven process improvement; or to link clinical care and research.

Six research challenges for information technology and healthcare
Patient-centered cognitive support
Modeling—an individualized virtual patient
Automation—integrated use, adaptivity
Data sharing and collaboration
Data management at scale
Automated full capture of physician-patient interactions

Functioning in the zone of complexity
Low
Chaos
Agreement
about
outcomes
Plan and control
High
Low
High

Cognitive support
Security and policy
Valueservices
Applications
Analysis
Integration
Platform services
Management
Publication
Data sources
Radiology
Medical records
Pathology
Genomics
Labs
RHIO

“People tend to overestimate the short-term impact of change, and underestimate the long-term impact.”
— Roy Amara
“The computer revolution hasn’t happened yet.”
— Alan Kay, 1997

Thank you!
Computation Institutewww.ci.uchicago.edu