1

Modelling-in-the-Large

Anthony Finkelstein

London Software Systems

Subtitles:

How Requirements Engineering can curecancer …

How Requirements Engineering caneliminate diabetes …

How Requirements Engineering cansolve the problems of climate change …

2

What is a Keynote Talk?

An excuse to be ‘programmatic’. Toask questions without offeringanswers. To put forward outrageousgeneralisations from narrowfoundations.

Three initial excursionsrelated to my major themes

But basically just thingsI want to say!

3

As a consequence of the growth ofCS as a discipline we largelyabandoned direct teaching ofsoftware development to engineersand scientists

1

They learn to program(mostly through experience)

But they do not learnsoftware engineering

We have focused on advancingresearch in software engineering andwe have neglected our ‘service’ role

2

We can (and should) contributeas practitioners. What matters isthat science advances as a whole

and not just our little corner

4

We have a professional responsibilityto contribute to the solution of bigand meaningful problems

3

Disease, poverty,environment, crime, security, energy,

democracy …

Two thematic threads

(a) Requirements Engineering used to‘understand science’

(b) Requirements Engineering used in ‘supporting science’

5

Cancer Research:• improved ‘high throughput’

techniques from molecularbiology

• large scale trials data• growth in scientific literature• new imaging technologies

a

Has led to a rapid growth of dataresources … and some ‘services’

a

HeterogeneousDistributed

Locally managedand of variable quality

6

aIntegratethe data

Build a portalConstruct a

platform

But what exactly does this mean?What is data integration? How isthis data to be accessed? How is itto be used?

– There is a general sense thatthere is ‘something’ of significantvalue in the fused data. But whatand how to use it is unknown.

a

7

The science has created newpossibilities that requires genuinelynew types of infrastructure and waysof working (processes) in order torealise them

a

What is needed …A way of understanding how

these new resources mightbe used

A way of understanding howscientists work with complexdata resources in sync withexperimental and clinicalwork

A plan for how to build thesupport necessary for futurecancer research

a

8

In other words … RequirementsEngineering is needed to help to build anew set of ‘system metaphors’ and‘concept of operation’ for infrastructure tosupport cancer research

Scenario and use case analysisGoal-oriented analysisModellingSystematic elicitationObservational methods

a

What we have been doingNCRI Informatics Initiative[in conjunction with NIH caBIG]

Funded by

a

9

• The centrality of the ‘investigation’

• The need for ‘lightweight’ data transformationsupport

• The fuzzy boundary between ‘external’ and‘internal’ data

• The complexities of scientific ‘data sharing’

• The importance of the ‘community’

• The move away from a static understanding ofintegration

• Hidden data management problems(versioning)

What we have ‘discovered’ …

a

Vision …Through Requirements Engineeringwe might effect a radical change inthe way that cancer scientists workwith their data and make a smallcontribution to finding a ‘cure forcancer’

a

10

An influential paper …Computational Thinking, JeanetteWing, CACM, March 2006, 49, 3

b

b "Computational thinking is using abstractionand decomposition when attacking a largecomplex task or designing a large complexsystem. It is separation of concerns. It ischoosing an appropriate representation for aproblem or modeling the relevant aspects ofa problem to make it tractable. It is usinginvariants to describe a system’s behaviorsuccinctly and declaratively. It is having theconfidence we can safely use, modify, andinfluence a large complex system withoutunderstanding its every detail. "

11

A ‘reductionist’ scientific agenda hasbeen highly successful … now weneed to complement this with theability to use the knowledge we havegained to understand ‘complexsystems’

b A quick switch of focus

Example: Biomedicine

b

12

The Molecular Revolution

A revolution that has reshaped the life sciences

We now understand:– the DNA sequence of many genes, up to

whole genomes– the mechanics of much of RNA synthesis– the genetic code for specifying amino acids

so that the backbone of a protein can bedirectly predicted

– the means by which one gene can generatemany RNAs and therefore proteins

b

– how DNA sequences, called promoters,determine which genes are expressed

– how DNA binding proteins, calledtranscription factors, modify gene expression

Knocking-out and over-expressing genes andRNAs have revealed how particular genescontribute to certain biological processes.

b

13

The focus now is turning to wholebiological systems: the heart, thecardiovascular system, the brain, theliver and to ‘complex diseases’ suchas cancer …

systems biology

b

To succeed it is necessary to combineinformation from the many rich areas ofbiological information. Alongside the genome,our knowledge about genes, we place theproteome, metabolome, and physiome, ourinformation about proteins, metabolicprocesses, and physiology

So as to build an integrated physiology ofwhole systems

b

14

The prize to be attained isimmense!– From ‘in-silico’ drug

design and drug testing– To individualised

medicine that will takeinto account physiologyand genetic profile

Systems biology has thepotential to have aprofound impact onhealthcare and beyond.

b

Even if we had a catalogue of all thegene sequences, how they aretranslated to make proteins, whichprotein can interact with which, andthe way in which the protein backbones fold, it is not possible to putthem into a functionally meaningfulframework

b

15

Cataloguing is not Understanding

–All proteins are post-translationallymodified. These additions influencethe actual shape of proteins

– Just because two proteins caninteract, it does not mean that theydo so in real cells

–Many functionally important, smallmolecules are synthesized bymetabolism

b

Modelling

A bottom-up, ‘data-driven’ strategy, willnot work — it is not possible to build anunderstanding of biological systems froman understanding of the componentsalone

What other approaches might be tried?

b

16

Use experimental information tobuild models at different biologicalscales, integrating these models tocreate an ‘orchestrated’ assemblageof models ranging from grossmodels of physiological functionthrough to detailed models thatbuild directly on molecular data

b

• Matching modelling schemes to the systemicphenomena that are of interest (structural andbehavioural, discrete and continuous)

• Integrating heterogeneous modelling schemeswith different structuring schemes

• Synthesising the results of analysis

• Correlating the models with ongoingexperimentation

What does this mean …b

17

• Supporting the construction and evolution ofcomplex models

• Supporting global collaboration around themodels

• Supporting ‘contested science’ - conflict,inconsistency

• Supporting the model-experiment loop

What does this mean …

I think this looks familiar!

b

• From gene to cell, from cell to organ, fromorgan to whole body physiology

– And ultimately to individuals

– Compare this with the current situation inmedical informatics

But the scale is beyond anything familiarb

18

This is a real Grand Challenge for softwareengineers and specifically for RequirementsEngineering

And it applies also to climate modelling …(and many others)

b

What I (actually we) have done:A framework for selecting modelling schemesA metamodel for systems biologyA model parameter repositoryA prototype middleware for integrating

heterogeneous modelling platformsA unified ontology framework for systems

biology modelsA new versioning and impact analysis tool for

systems biologists

(the first) modular, integrative, scale-crossing, hybrid model of liver glucose homeostasis

b

19

By way of a conclusion …

We have much to contribute

We should look outward rather thaninward