Page 1

1
Urgent and Critical Problems

• Patrick LincolnSRI International
• June 10, 2008

2
Three Talking Points Some Problems

• Terrifying software present
• Symbolic Systems Biology Formal Methods For
  Biology
• The worlds most urgent, critical problem

3
Verification and Validation At Risk

• VV is losing the race against the growth in
  complexity of software
• Formal analysis of software using old approaches
  is slow and costly
• Lack of effective technologies to automate
  system-level analysis prevents routine use on
  systems of industrial scale
• State space explosion exponential in the size of
  the system
• Test coverage of complex interconnected systems
  0
• Requires new approaches to VV
• Requires new approaches to combine evidence from
  multiple analyses of multiple components

4
Mil-Spec Mashups

• Kids these days are programming with huge hunks
  of dodgy code (services) composed with
  disreputable glue
• Mashup of google maps, craigslist, epinions,
  newsfeeds,natural language translations, image
  databases, video
• This is how your future programmers are trained
• Could we provide a framework to enable
  appropriate use of mashups of untrusted,
  uncontrolled services in critical applications?
• Time-critical, Safety-critical, Mission-critical
• Basis for skepticism Mountains of bad code no
  one fully understands
• Basis for optimism hope springs eternal

5
Whats a Mashup?

• A quick-and-dirty way to throw together some new
  web service from several component services.
• 10 minutes vs 10 months
• The only way kids know how to program
• Example mashup generation frameworks
• Intel Mash Maker
• Google Mashup Editor
• LiquidApps
• Microsoft Popfly
• Serena Mashup Editor
• Yahoo pipes

6
Total Mashups on programmableweb.com
7
Twitter Word Association

• Twitter is a microblog site
• 140 character max per post
• Is extremely popular with your future students /
  employees

Twitter of attention (time spent on
twitter.com as percentage of all time spent
online)
8
Problem

• Not all of these are 100 bug-free

9
We Might Need Higher Assurance

• Formal Methods to the Rescue!

10
Way Forward

• Get the requirements right
• What do we have to care about?
• Compositional frameworks
• Reliable development from unreliable components
• Include evidence of assurance as first-class
  citizen in development proc
• Combining evidence from multiple analysis tools
  (toolbus)
• Service-Oriented Architecture everywhere
• Rigorous specification of top-level service
• VV for critical aspects of framework
• VV of certain critical component services
• Zero assurance for most components
• High assurance for critical pieces and framework
• High assurance for overall system

11
Mil-Spec Mashups The Network Is The Computer

• SOA view of software

Standard View
Mashup View
12
Mil-Spec Mashups Service-Oriented Architecture

• Future systems will mostly be built this way
• Whether we like it or not
• Programming environments become service
  composition frameworks
• More and more software will be built at a
  higher-level of abstraction as composition of
  functionalities and services
• While this useful for building systems and rapid
  prototyping, it creates risks of leaky
  abstractions where lower levels of bugs impact
  system reliability
• Google Mashup Editor
• LiquidApps
• Microsoft Popfly
• Yahoo pipes

13
Mil-Spec Mashups Whats missing? Assurance

• Assurance for services and composition
• We need to support High-Assurance SOA with new
  directed research
• Diversity of service provision
• Redundancy management
• Composable assurance of software components
• Highly automated analysis test generation of
  key properties of subsystem

14
Mil-spec Mashups Key Lever Automated
Abstraction

• View of system to analyze a property
• As simple as possible, but no simpler
• Automation is vital for complex systems
• Hybrid abstraction and analysis
• Complete analysis of simple properties of complex
  systems
• Hybrid SAL, CHARON, CHECKMATE,Hyper, HyTech,
  Simulink/Stateflow

15
Mil-spec Mashups Restating the Problem

• You have to bet your life / company / country
• on systems quickly slapped together from services
  and
• components built by untrusted sources by
  untrained staff
• What do you do?

16
PHD Modeling

• SRI PHD Data Processing and Modeling Team

PHD Kickoff 28-29 September, 2007
17
What Are We Trying to Do?

• Help address threats to men and women in uniform
• Our enemies subvert neutral (if annoying)
  infrastructure, including airborne
  infrastructure, and cause grievous harm on
  warfighters and civilians and compromise
  operational readiness

18
Challenges for Reasoning About Biological Systems

• Biological processes are complex
• Dynamics that range over huge timescales
• Microseconds to years
• Nonlinear systems
• Spatial scales over 12 orders of magnitude
• Single protein to cell, cell to whole organism
• Oceans of experimental biological data being
  generated
• Hurricanes of noise, experimental error, and
  natural variation
• What can we do?

19
Eschericia coli Metabolism Karp et al
20
A Metabolic reaction in more detail
21
Computational Point of View

• View cells as if they are performing
  computations, just like chips
• Pathways might be understood as large scale
  computational networks

Digital circuits
Human or Disease Pathways
22
Symbolic Systems Biology

• Proposed Definition
• The qualitative and quantitative study of
  biological processes as integrated systems rather
  than as isolated parts

23
Rigorous Abstract Methods Are Needed To

• Accommodate conventional types of discrete
  reasoning based on experimentation
• Unambiguously define a model and allowable
  reasoning steps
• Combine data-driven and model-driven refinements
• Provide predictive power for generating testable
  hypotheses

24
Compositionality

• Subsystems that
• Exhibit identifiable interfaces
• Are modifiable independently
• Enable abstract modeling
• When compositionality holds, it is very useful
• Leverage point for abstraction
• Enables decomposition analysis and understanding
  tasks
• Enables reasoning at much larger scales

25
Are Biological Systems Compositional?Can we
reason about them as if they are?

• Genetic code
• Translation into amino acids
• Protein domains
• Operon structure
• Bilipid layer membranes
• Organelles
• Signaling pathways Eg motifs, modules (Arkin)
• Metabolic pathways
• Cells Eg delta notch signaling (Tomlin)
• Organs
• Individuals

26
Example Problems in Diet Planning

• Given a qualitative metabolic map
• What are all the minimal sets of nutrients?
• Further, assume the networkis optimal for growth
• What are the rate constantsgoverning each
  reaction inthe metabolic network?

27
Example Problem PHD

• Predict Health and Disease
• From a small sample (blood, saliva, breath), from
  a patient, determine
• If they are sick
• What they are sick with
• How long they have been sick
• If they are infectious
• If they will succumb to significant symptoms in
  the near future
• Initial study on Rhino virus in humans, and 8
  diseases in vitro suggests we can (!)

28
Tools Used

• Principal Component Analysis (PCA)
• Independent Component Analysis (ICA)
• Canonical Correlation Analysis (CCA)
• Hierarchical clustering
• Pathway Logic (Symbolic pathway analysis)
• BioCyc (metabolic pathways)
• Hybrid Analysis Laboratory (HAL)
  (continuoussymbolic)
• Hidden Markov Model (HMM)
• Dynamic Bayesian Nets (DBN)
• Support Vector Machines (SVM)
• Other standard statistical tools

29
Symbolic Systems BiologyA Large Collaborative
Effort
Adam Arkin (LBL/HHMI) David Dill (Stanford
CS) Rasha Hamameiah (WRAIR) Ronojoy Ghosh
(Stanford) John Doyle (CalTech Control) Drew Endy
(MIT BioE) Geoff Ginsberg (Duke) Al Hiro
(UMich) Michael Hucka (CalTech SBW) Jose Meseguer
(U.Il. CS) Marti Jett (WRAIR) Herbert Suaro
(Keck) Claire Tomlin (Stanford Aero) Denise Wolfe
(UCB B.sub)

• Annalisa Dandrea (SRI Immunology)
• Steven Eker (SRI Algorithms)
• Tom Garvey (SRI Reasoning)
• Maryann Greco (SRI Proteoemics)
• Peter Karp (SRI BioCyc)
• Tom Kilduff (SRI Neuro)
• Merrill Knapp (SRI Cancer)
• Keith Laderoute (SRI Cancer)
• Patrick Lincoln (SRI)
• Anne Rascle (SRI Immunology)
• John Rushby (SRI formal methods)
• Natarajan Shankar (SRI formal methods)
• Mark-Oliver Stehr (SRI Rewriting)
• Carolyn Talcott (SRI Rewriting)
• Ashish Tiwari (SRI SAL)
• Larry Toll (SRI Hormone Discovery)

Green is Bio Blue is Chem Purple is
Control Yellow is Math Black is Comp Sci
30
Quiz

• What is the worlds most urgent and critical
  problem?

31
Answer

• The worlds most urgent and critical problem is
• Improving our collective ability to solve
  urgent and critical problems

32
Answer

• The worlds most urgent and critical problem is
• Improving our collective ability to solve
  urgent and critical problems
• Doug Engelbart, March 1951 (paraphrased)

33
Thanks

• To all the many collaborators,advisors,
  educators
• To Funding Agencies
• SRI corporate investment
• NSF
• DARPA BioCOMP
• DARPA PHD
• To you
• For listening
• Your tax dollars