The Semantic Web - This time... its Personal

“Shopping for data should be as easy

as shopping for shoes!!”

Carole Goble

The Semantic WebThis time… it’s personal!

Mark Wilkinson, PI Bioinformatics, Heart + Lung Institute @ St. Paul’s HospitalVancouver, BC, Canada

My Lab

Engineering &

Research

Programmers&

Students

Coolness!&

Study of Coolness!

Middleware

Robert Stevens

Shouldn’t be

seen!

DEMO

...to give you incentive to listen to the rest of the presentation

Show me patients with elevated creatinine along with their latest BUN and creatinine levels

PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#> PREFIX patients: <http://sadiframework.org/ontologies/patients.owl#> PREFIX pred: <http://sadiframework.org/ontologies/predicates.owl#> SELECT ?patient ?bun ?creatFROM <http://sadiframework.org/ontologies/patients.rdf>WHERE {

?patient rdf:type patients:ElevatedCreatininePatient .?patient pred:latestBUN ?bun . ?patient pred:latestCreatinine ?creat .

}

VOILA!

There was no database...

There was no warehouse...

There was no patient data anywhere annotated as

“Elevated Creatinine Patient”

How did I answer a question where the required data didn’t exist?

(...no, I didn’t just make it up LOL!)

In the beginning was the problem...

The Problem

The Problem

The Holy Grail:(circa 2002)

Align the promoters of all serine threonine kinases involved exclusively in the regulation of cell sorting during wound healing in blood vessels.

Retrieve and align 2000nt 5' from every serine/threonine kinase in Mus musculus expressed exclusively in the tunica [I | M |A] whose expression increases 5X or more within 5 hours of wounding but is not activated during the normal development of blood vessels, and is <40% homologous in the active site to kinases known to be involved in cell-cycle regulation in any other species.

The Problem

The Problem

The Solution??

Why not?

Heart

Heart

You don’t know

what I know!

You don’t think

how I think!

No Personalization!

So… what can we do?

What we need

How do we make Bruce’s knowledge machine-readable?

Ontologies!

Two Problems with that…

#1 Ontology Spectrum

Catalog/ID

SelectedLogical

Constraints(disjointness,

inverse, …)

Terms/glossary

Thesauri“narrowerterm”relation

Formalis-a

Frames(Properties)

Informalis-a

Formalinstance

Value Restrs.

GeneralLogical

constraints

Originally from AAAI 1999- Ontologies Panel by Gruninger, Lehmann, McGuinness, Uschold, Welty; – updated by McGuinness.Description in: www.ksl.stanford.edu/people/dlm/papers/ontologies-come-of-age-abstract.html

#1 Ontology Spectrum

Catalog/ID

SelectedLogical

Constraints(disjointness,

inverse, …)

Terms/glossary

Thesauri“narrowerterm”relation

Formalis-a

Frames(Properties)

Informalis-a

Formalinstance

Value Restrs.

GeneralLogical

constraints

#1 Ontology Spectrum

Catalog/ID

SelectedLogical

Constraints(disjointness,

inverse, …)

Terms/glossary

Thesauri“narrowerterm”relation

Formalis-a

Frames(Properties)

Informalis-a

Formalinstance

Value Restrs.

GeneralLogical

constraints

WHY?

#1 Ontology Spectrum

Catalog/ID

SelectedLogical

Constraints(disjointness,

inverse, …)

Terms/glossary

Thesauri“narrowerterm”relation

Formalis-a

Frames(Properties)

Informalis-a

Formalinstance

Value Restrs.

GeneralLogical

constraints

WHY?

Because I say so!

Because it fulfils XYZ

Because I say so?!?

That’s not very... Personal...

No room for disagreement

No place for scientific discourse

It is what it is...Because I say so!

Clay Shirky

“Ontology is over-rated”

http://www.shirky.com/writings/ontology_overrated.html

If we’re going to personalize the Semantic Web

We must change the way we create and use

classification systems(“shelves”)

“Get rid of the shelf”

- Clay Shirkey

You don’t know

what I know!

So don’t tell ME

how data should be interpreted!

How do we get rid of the shelf?

This is going to hurt...

Web Services

vs.

Semantic Web

Web Servicesare not “connected” to

the Semantic Web

Why?

Web ServicesXML + XML Schema

Semantic WebRDF + OWL

Web ServicesPOST of SOAP-XML

Semantic WebGET of RDF-XML

Web ServicesNo (rigorous) semantics

Semantic WebRich, flexible semantics

Web Services&

Semantic Web

Fundamentally and deeply different Web technologies!

>1000 X more data!

Accessing these databases and

analytical algorithms “transparently”,

based on an individual researcher’s

ideas, beliefs, and preferences

will help us personalize

medical research

Mark Butler (2003) Is the semantic web hype? Hewlett Packard laboratories presentation at MMU, 2003-03-12

Semantic Web?(my definition)

An information system where machines can receive information from one source, re-interpret it, and correctly use it for a purpose that the source had

not anticipated.

Re-interpretation

Correct re-use

Both are critical to

the personalization

of research

Building a

personalized Semantic Web…

Step-by-step…

Founding partner

Semantic Automated Discovery and Integrationhttp://sadiframework.org

(open source)MicrosoftResearch

“best-practices” for Semantic Web Service provision

standards-compliant

Lightweight(only 2 “rules”)

Rules come from observations:

SADI Observation #1:

Web Services in Bioinformatics create implicit biological relationships

between their input and output

SADI Observation #1:

SADI Best Practice #1

Make the implicit explicit…

A Web Service should create “triples” linking the input data to the output data, thus explicitly describing the semantic

relationship between them

SADI Best Practice #1

This is what bioinformatics Web Services implicitly do anyway!

Easy to implement this as a best-practice

SADI Observation #2:HTTP GET and POST

GET guarantees the response relates to the request URI

in a very precise and predictable way

POST does not…

SADI Observation #2:GET and POST

That’s why Web Services have a fundamentally different behaviour than the Semantic Web

SADI Observation #2:GET and POST

We can fix that!

(without breaking any existing rules or standards!)

SADI Best Practice #2

SUBJECT URI of the output graph (triples) is the same

as the SUBJECT URI of the input graph (triples)

(the output is “about” the input... Now explicitly!)

Consequence

The “Semantics” of our interaction with the Web Service are now

explicit and identical to the “Semantics” of GET

SADI Web Service Interfaces

Service Interfaces defined by two OWL classes:

SADI Web Service Interfaces

OWL Class #1: My Input Class

SADI Web Service Interfaces

OWL Class #2: My Output Class

SADI Web Service Interfaces

My Service consumes OWL Individuals of Class #1

and returns OWL Individuals of Class #2

…but the URI of those two individuals is the same!(see best practice #2)

How do we discover services?

Since input and output are about the same “thing”,

we can automatically determine

what a service does

by comparing

the Input and Output OWL classes

Automatically index services in a registry based

on what properties (predicates) Services add to

their respective input data

How do we discover services?

EXAMPLE

Input Data: BRCA1 rdf:type Gene ID

Output Data: BRCA1 hasDNASequence AGCTTAGCCA…

Registry Index: Service provides “hasDNASequence” property to Gene IDs

Now we can answer questions like

“what is the DNA sequence of BRCA1?”

Discover a SADI Web Service that generates the

DNA Sequence property for gene identifiers

Okay, enough tech gobbledygook

What will this do for ME?

Demo #1

Imagine there is a “virtual database” containing all of the data from all of the databases,together with the output of

every conceivable analysis

How do we query that database?

“SHARE”

Semantic Health And Research Environment

SADI client application

What pathways does UniProt protein P47989 belong to?

PREFIX pred: <http://sadiframework.org/ontologies/predicates.owl#>PREFIX ont: <http://ontology.dumontierlab.com/>PREFIX uniprot: <http://lsrn.org/UniProt:>SELECT ?gene ?pathway WHERE {

uniprot:P47989 pred:isEncodedBy ?gene . ?gene ont:isParticipantIn ?pathway .

}

Recapwhat we just saw

A standard SPARQL query was entered into SHARE, a SADI-aware query engine

Recapwhat we just saw

The query was interpreted to extract the properties being queried and these were passed to SADI for Web Service discovery

Recapwhat we just saw

SADI searched-for, found, and accessed all databases and/or analytical tools capable

of generating those properties

Recapwhat we just saw

We posed, and answered a complex database query

WITHOUT A DATABASE

(in fact, the data didn’t even have to exist...)

The Holy Grail:

Align the promoters of all serine threonine kinases involved exclusively in the regulation of cell sorting during wound healing in blood vessels.

Retrieve and align 2000nt 5' from every serine/threonine kinase in Mus musculus expressed exclusively in the tunica [I | M |A] whose expression increases 5X or more within 5 hours of wounding but is not activated during the normal development of blood vessels, and is <40% homologous in the active site to kinases known to be involved in cell-cycle regulation in any other species.

Cool!

…but I’m supposed to be personalizing research…

Let’s make this a little more personal by bringing in Ontologies

My Definition of Ontology (for this talk)

Ontologies explicitly define the things that exist in “the world”

based on what properties each kind of thing must have

Ontology Spectrum

Catalog/ID

SelectedLogical

Constraints(disjointness,

inverse, …)

Terms/glossary

Thesauri“narrowerterm”relation

Formalis-a

Frames(Properties)

Informalis-a

Formalinstance

Value Restrs.

GeneralLogical

constraints

Demo #2Discover instances of OWL classes

from data that doesn’t exist…

Data exhibits “late binding”

Late binding:

“purpose and meaning” of the data is

not determined untilthe moment it is required

Benefitof late binding

Data is amenable toconstant re-interpretation

...MY interpretation

How?

DO NOT PRE-CLASSIFY DATA

Just hang properties on it

Ontologies are in the “Frames” area of the Ontology spectrum, and therefore can leverage SADI and be

“executed” as workflows

???

Did you just say

“execute ontologies as workflows?!”

Show me patients with elevated creatinine along with their latest BUN and creatinine levels

PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#> PREFIX patients: <http://sadiframework.org/ontologies/patients.owl#> PREFIX pred: <http://sadiframework.org/ontologies/predicates.owl#> SELECT ?patient ?bun ?creatFROM <http://sadiframework.org/ontologies/patients.rdf>WHERE {

?patient rdf:type patients:ElevatedCreatininePatient .?patient pred:latestBUN ?bun . ?patient pred:latestCreatinine ?creat .

}

Start burrowing through the OWL class find that we need aregression model OWL class

Regression models have features like slopes and intercepts… and so onThe class is completely decomposed until a set of required Services are discoveredcapable of creating all these necessary properties

Successful decomposition of the OWL class to discover the need for a LinearRegression Web Service, and so on

VOILA!

Current Research Project Benjamin Vandervalk

There are many ways to resolve these ontologies and queries into workflows

My student Benjamin is currently studying optimization of this query resolution strategy

Interesting counterpart to DBMS query resolution because there are no indices for Web Services, and other issues (e.g. Service speed) are an important factor.

The Holy Grail:

Align the promoters of all serine threonine kinases involved exclusively in the regulation of cell sorting during wound healing in blood vessels.

Retrieve and align 2000nt 5' from every serine/threonine kinase in Mus musculus expressed exclusively in the tunica [I | M |A] whose expression increases 5X or more within 5 hours of wounding but is not activated during the normal development of blood vessels, and is <40% homologous in the active site to kinases known to be involved in cell-cycle regulation in any other species.

OWL Class restrictions converted into workflows

SPARQL queries converted into workflows

Reasoning happening in parallel with query execution

Data fulfilling OWL models is discovered,

or generated through running analytical tools

SADI and CardioSHARE

I still don’t seehow this is“Personal”

??

SELECT ?patient ?bun ?creatFROM <http://sadiframework.org/ontologies/patients.rdf>WHERE {

?patient rdf:type patients:ElevatedCreatininePatient .?patient pred:latestBUN ?bun . ?patient pred:latestCreatinine ?creat .

}

I created a small ontologydescribing my definition of

an Elevated Creatinine Patient

… it was MY ontology!

I can re-use it

I can modify it as I change my world-view

I can publish it for others to use

Others can modify it to fit THEIR world-view

My personal world-view is being dynamically resolved againstglobal data and knowledge

…but it’s bigger than that…

“Elevated Creatinine Patient”

I made that up! It came out of my head!

What’s another word for a world-view that you make-up?

Hypothesis

Current Research Project

We believe that ontologies and hypotheses are, in some ways, the same “thing”…

…simply assertions about individuals that may or may not exist

Current Research ProjectSoroush Samadian

e-Copy Clinical Outcomes investigations to determine if Outcomes hypotheses can be modeled as OWL ontologies and automatically resolved by SHARE

Recap

SADI Semantic Web Services generate triples; the predicates of those triples are indexed... Period.

For a given query, determine which properties are available, and which need to be

discovered/generated

Find services that generate the properties we need

Semantic Web

An information system where machines can receive information from one source, re-interpret it, and correctly use it for a purpose that the source had

not anticipated.

My Purpose!!

What SADI + SHARE supports

Re-interpretation

We constantly compare the collection of properties, gathered from third-parties worldwide, to whatever world-model (query/ontology) we wish to view it

through.

MY world model

Novel re-use

There is no way for the provider to dictate how their data should be used, or how it should be interpreted. They simply add

their properties into the “data cloud” and those properties are used in whatever

way is appropriate for ME.

What SADI + SHARE supports

And all this because SADI simply requires

that the input URI

is the same

as the output URI

Important “wins”

Data remains distributed

no warehouse!

Data is not “exposed” as a SPARQL endpoint

greater provider-control over computational resources

Semi-automated SADI service writing and deployment

Taverna

Semantically-guided SADI service discovery and pipelining

SADI Plug-ins

Where do we go from here?

Consent

Who defines my consent?

IRB

(in Canada REB)

Where’s MY opinionin that process?

New Research Project - iConsent

I define an ontology describing the conditions under which I would allow my bio/medical data to be used

Type of study, funding source, type of information

Highly granular, highly detailedHighly PERSONAL

New Research Project - iConsent

Researchers approach the database with an “instance” of their study (e.g. an instance of the OBI)

Semantic negotiation between my ontology and their study parameters determines which aspects of my data they are

allowed access to (if any)

New Research Project - iConsent

For Patient:

Personalized access to personal information

Much more granular than current consent models

New Research Project - iConsent

For Researcher:

Access to more data from willing patients

Automated profiling of data you weren’t allowed access to – detection of sample bias!

Simple and Open WINS!Join us!

We have recently received funding from CANARIEto assist and train service providers

in deploying their own SADI Semantic Web Services

Come join us – we’re having a lot of fun!!

http://sadiframework.orghttp://twitter.com/sadiframework

Fin

This presentation available on SlideShare: keywords ‘sadi’, ‘wilkinson’

C r e d i t s

B e n j a m i n V a n d e r V a l k M S c S t u d e n t B i o i n f o r m a t i c s T r a i n i n g P r o g r a m m e

L u k e M c C a r t h y L e a d P r o g r a m m e r , S A D I

S o r o u s h S a m a d i a n P h D S t u d e n tB i o i n f o r m a t i c s T r a i n i n g P r o g r a m m e

Microsoft Research