e-LICO Kick-off meeting

e-LICO Kick-Off Meeting

12-13 February 2009

University of Geneva AI Laboratory

1227 Carouge/GE, Switzerland

Participants

  • University of Geneva (UNIGE): Melanie Hilario, Huyen Do, Alexandros Kalousis, Phong Nguyen, Dina Vishnyakova, Adam Woznica
  • Institut National de la Santé et de la Recherche Médicale (INSERM): Joost-Peter Schanstra, Julie Klein
  • Medicel Oy (MEDICEL): Christophe Roos, Jukka Matilainen
  • National Hellenic Research Foundation (NHRF): Aristotelis Chatziioannou, Ilias Maglogiannis
  • Rapid-I (RAPID-I): Ingo Mierswa, Simon Fischer, Sebastian Land
  • University of Manchester (UNIMAN): Robert Stevens, Steve Pettifer
  • University of Zurich: Abraham Bernstein, Jörg-Uwe Kietz, Floarea Serban

The meeting agenda covered both technical and administrative issues. Below we focus on the major technical issues concerning the major components of the e-LICO project: the infrastructure, the data mining (DM) layer, and the pilot application on the kidney and urinary pathways (KUP).  Below we summarize briefly

The e-Science Layer

  • Call for use cases:  UNIMAN stressed the need for use cases to guide the customization of the myGrid/Taverna infrastructure and the development of e-LICO-specific add-ons such as the collaborative ontology editor.
  • Training on ontology development: UNIMAN proposed to organize experimental sessions on ontology development for small groups of around 6 persons, e.g., partners and users involved in the development of the application domain ontology.

The Data Mining Layer

  • Design of the DM layer: The DM layer will be built around RapidMiner, RAPID-I's open-source DM tool suite. There are two alternative options for integrating RapidMiner into the myGrid/Taverna-based infrastructure. In loose coupling mode, the whole RapidMiner platform will be encapsulated as a single web service and will appear as a single node in any workflow developed by the e-LICO user. In tight coupling mode, each of RapidMiner's 500 DM operators will be made available as a web service in e-LICO. The user will be able to work on the DM part of the overall workflow without leaving the Taverna workflow editor/enactor.The advantages and limitations of both modes will be further investigated.
  • Development of the DM Ontology: Discussions focused on 3 ontological sources contributed by e-LICO partners from past projects or from ongoing work: MiningMart's M4 metamodel (UZH and RAPID-I), IDEA's DM ontology (UZH), and an initial hierarchy of DM concepts proposed by UNIGE.
  • Other open issues include: (1)  the need to clarify connections between structured DM, TM and IM. A first step towards this clarification will be provided by the use case to be developed for the pilot application. (2) issues such as DM process standards and scalability should be borne in mind throughout the project.

The Application Layer

  • Application requirements and specifications: The KUP use case developed at the request of UNIMAN will be generalized and expanded by INSERM into a comprehensive user specifications document.
  • KUP ontology and data base: e-LICO partners INSERM, UNIMAN and UNIGE---and perhaps MEDICEL---will participate in the EuroKUP working meeting to be held in Napflio, Greece, on 29 March 2009. They will give a presentation on building the EuroKUP ontology and data base and recruit volunteers to contribute their domain expertise. They will also invite interested EuroKUP members to participate in the session on ontology development proposed by UNIMAN.