WSBioZard: A Wizard for Composing Bioinformatics Web Services

1
WS-BioZard A Wizard for Composing Bioinformatics
Web Services

• Zhiming Wang, John Miller, Jessica Kissinger, Rui
  Wang, Doug Brewer, Cristina Aurreochea
• Computer Science Department
• Genetics Department
• University of Georgia

2
Web Service Composition

• Many data and knowledge resources as well as
  analysis tools are now available via
• Downloadable stand-alone applications
• Web pages/forms
• Web services
• Web Service Composition
• Goal provide a convenient way to combine
  services
• Standards-driven approach
• Growing number of bioinformatics Web services
• Complex task needs to compose multiple Web
  services

3
Benefits of Web Service Composition

• No need to download/install/maintain compared
  with stand-alone applications
• No need to copy/paste/access the same Web page
  again and again compared with using Web
  pages/forms
• Easier to change a Web service during the
  composition through drag and drop in a design
  tool.
• Easier to develop (depending on the quality the
  design tool)?
• Easier to reuse, since the resultant process can
  itself be invoked as a standard Web service

4
Designing Tools for Web Service Composition

• Which Composition Language WS-BPEL, WS-CDL, XPDL
• Challenges of BPEL Composition
• Non-BPEL Systems and Their Problems
• Our Approach
• Semi-automatic composition to produce BPEL
• BPEL should run in multiple execution engines
• Add semantics to aid both service discovery and
  composition
• Provide support for data mediation
• WS-BioZardPrototype Implementation
• Service discovery tool
• Simplified BPEL editor with Wizard
• WSDL generation, BPEL deployment and invocation

5
WS-BPEL Web Service Business Process Execution
Language

• The most popular standard for composing Web
  services
• Facilitates the integration of distributed
  software systems
• Executable XML specification
• BPEL process can be invoked as an ordinary Web
  service
• Support from industry many free and commercial
  tools
• Several BPEL model API's/implementations based on
  Java
• ActiveBPEL, NetBeans and Eclipse BPEL
• Easy to reuse compared with self implementation
• Several BPEL execution engines
• Apache ODE, ActiveBPEL and Sun Glassfish

6
Challenges of BPEL Composition

• Service discovery not integrated with composition
• Little assistance with data handling (no data
  mediation)
• Tremendous diversity in Bioinformatics Web
  Service data format and terms, lacking semantics
  -- makes data handling and service composition
  hard
• Complex GUI designer/editor which can be hard to
  use
• Reflects all the complexity of BPEL to the GUI--gt
  complex control flow and data flow
• The user manually defines almost all the process
  when using GUI design tools ActiveBPEL, NetBeans
  BPEL, Eclipse BPEL
• Limited or no ability to generate the process
  WSDL
• Annoyance of having to make deployment descriptors

7
Non-BPEL Systems and Their Problems

• Taverna -- uses XScufl rather than XPDL or BPEL
• Easier to use -- straightforward control flow
• Automatically handle looping to some degree
• No data mediation
• Uses BeanShell language to handle data extra
  effort
• Non-Industry standard composition that cannot
  execute outside of Taverna
• Bio-Moby
• Facilitates service discovery
• No composition language delegate to Taverna
• Documentation could be improved
• Currently best combo in bioinformatics

8
Design Goals for WS-BioZard

• Lower the complexity of composing Bioinformatics
  Web services
• Use semi-automatic composition
• Add semantics to Bioinformatics Web Services
• To aid service discovery and composition
• To aid data mediation
• Support and evaluate multiple forms of data
  mediation
• Provide an intuitive GUI with a multi-faceted
  wizard to assist Bioinformatics Web service
  discovery, composition, deployment and invocation

9
Approaches to Web Service Composition

• Automatic composition
• Requires complete formal representations of the
  domain knowledge and the task's initial and goal
  states
• Requires preconditions and effects to be fully
  specified for correct operation
• May be the ideal future solution, but not mature
  enough for practical use
• Manual composition using BPEL designer/editor
• Tedious, time consuming and error prone
• End user must handle everything in detail
• Semi-automatic composition - an intermediate
  solution
• An interactive approach with greater assistance
  from the tool
• Reduced learning curve for end users

10
Benefits of Adding Semantics

• Adding semantics to Web services can provide
• Better service composition
• Better data mediation
• Better service discovery
• Semantic Web Services (SWS) Projects/Standards
• DAML-S/OWL-S First Major Project
• WSMO Large European Initiative
• METEOR-S Lightweight Approach
• SAWDSL First W3C Standard for SWS
• Semantic Annotations for WSDL

11
Data Mediation (DM)

• Converts data format/type from one Web service's
  output to the next Web service's input
• DM is an important task, but can be a very
  difficult task (especially to automate)
• Needed because of data heterogeneity, which is
  prevalent in the biological domain
• Structural and semantic heterogeneity
• Many theoretical approaches to DM, yet no widely
  available practical implementations
• Two research efforts to provide DM utilizing
• SAWSDL modelReference, lowering/liftingSchemaMapp
  ing

12
The METEOR-S Approach to DM

• Combining modelReference, liftingSchemaMapping,
  loweringSchemaMapping and XSLT/XQuery as the
  basis for data mediation
• Service provider needs to write XSLT or XQuery
  for each annotation
• Add a component to the run time system
• Generic Web service for data mediation
• Hard to express generic data types for WS
• SOAP Engine (e.g., Axis2) Plugin
• Complex, low-level coding
• SOAP Engine dependence
• Not all BPEL engines use Axis2

13
WS-BioZard Approach to DM

• A practical solution based on a bottom up
  strategy
• Requires SAWSDL Annotations
• modelReference pointing to an ontology, e.g.,
• Sequence Ontology (SO)?
• Utilizes a Data Mediation Algorithm
• Automatic for Identical, Equivalent or Superclass
  Types
• Popup Wizard when input values are missing
• Rejection when matching is insufficient
• Automatic XPath Generation

14
SAWSDL Annotation
ltxsdcomplexType name"inputType"gtltxsdsequencegt lt
xsdelement name"locations" type"xsdstring" s
awsdlmodelReference"http//purl.org/obo/owl/sequ
encelocation_text"gtlt/xsdelementgt ltxsdelement
name"optionalParameters" type"xsdstring" saw
sdlmodelReference"http//purl.org/obo/owl/sequen
ceOPTIONAL"gtlt/xsdelementgt ltxsdelement
name"extension" type"xsdint" sawsdlmodelRef
erence"http//purl.org/obo/owl/sequenceextent"gtlt
/xsdelementgt
ltxsdelement name"wuBlastResponse"gt ltxsdcomplexT
ypegt ltxsdsequencegt ltxsdelement name"locations"
type"xsdstring" sawsdlmodelReference"http//p
url.org/obo/owl/sequencelocation_text"
/gt lt/xsdsequencegt lt/xsdcomplexTypegt lt/xsdelemen
tgt
GeneByLocation WS
15
Data Mediation Algorithm
For case 1, WS-BioZard will automatically
generate the BPEL ltassigngt/ltcopygt elements. For
case 2, the data mediation wizard will popup and
assist the user to produce expressions for
missing values. WS-BioZard ability to
automatically generate XPath expressions greatly
simplifies the composition task for the user.
16
Automatic XPath Generation
ltassign name"invoke102Assign"gt ltcopygt
ltfromgtltliteralgtProductlt/literalgtlt/fromgt
lttogt invoke103inputVar.parameters/ns102in/ns102
optionalParameterslt/togt lt/copygt ltcopygt
ltfromgtltliteralgt0lt/literalgtlt/fromgt lttogt
invoke103inputVar.parameters/ns102in/ns102exten
sionlt/togt lt/copygt ltcopygt ltfromgtwuBlast102outputV
ar.parameters/ns101locationslt/fromgt lttogtinvoke10
3inputVar.parameters/ns102in/ns102locationslt/togt
lt/copygt lt/assigngt
17
Architecture of WS-BioZard
18
Comprehensive FrameworkWS-BioZard

• SAWSDL Annotation Tool - Radiant
• Semantic Service Discovery Tool Lumina 2
• Simplified BPEL Process Designer/Editor
• Auto-generate many BPEL elements
• assign, copy, invoke, variable, namespace,
  receive, reply, assign, partnerlink
• Designer can focus on application logic
• not get lost in details
• Support for practical data mediation

19
Lumina 2 A Service Discovery Tool
20
Complexity of WS-BPEL 2.0
21
BPEL Process Designer/Editor
22
Wizard for Data Mediation
23
A Composition Example
24
(No Transcript)
25
Comparison between Semi-Automatic Composition
Approaches
26
Conclusions

• Lower the complexity of BPEL composition through
  semi-automatic composition
• Simplified Control Flow
• fewer activities, abstracted activities,
  auto-generation of details
• Simplified Data Flow
• data mediation and assistance from wizard
• Comprehensive framework WS-BioZard features
• Practical use of semantics (via SAWSDL) to aid in
  service discovery, composition and data mediation
• Integrated service discovery tool
• Simplified BPEL designer/editor with a
  multi-faceted wizard
• Automatic WSDL generation and BPEL deployment and
  invocation

27
Current Solutions and Their Problems

• ActiveBPEL -- manual approach based on BPEL
• Complex control flow and data flow
• No assistance for data mediation -- using XSLT,
  XPath or XQuery
• Manually define everything
• Taverna -- based on XScufl
• Much easier than ActiveBPEL
• Straightforward control flow
• Automatically handle looping to some degree
• Little assistance for data mediation
• Using BeanShell language to handle data mediation
• Non-Industry standard composition that cannot
  execute outside of Taverna
• Bio-Moby
• Easier data flow
• No central editor and no composition language
• Little documentation

28
ActiveBPEL
29
NetBeans
30
Taverna