Minutes of Access Grid meeting, 25 Apr 2003

Attendees

EBI: Peter Rice, Tom Oinn

Manchester: Robert Stevens, Mark Greenwood, Phil Lord, Nedim Alpdemir

Newcastle: Anil Wipat, Savas Parastatidis?

Nottingham: Kevin Glover, Milena Radenkovic, Chris Greenhalgh

Southampton: Ananth Krishna, Simon Miles, Justin Ferris

Agenda

1. Discussion of RobertsISMBDemoScript (and Anil's GDScenarioReqs)
2. Types based on Chris G's observations - GravesStoryBoardDetails and SimpleSequenceTyping

Demo Features

Robert presented the main features that he would like to include in the ISMB demo, with the Graves Disease providing the basic scenario.

Running a workflow

An essential. In the longer term interacting with a workflow is required. The first stage will see an experiment consist of more than one workflow. The user interacts with myGrid to browse the results of one workflow and select some of its outputs as part of the input to the next workflow.

Visualisation of data in MIR

An essential. In the Graves Disease scenario there is the particular requirement to display a gene, its introns, exons, and the SNPs so that the user can select a particular SNP for primer design. Artemis was mentioned as a candidate, offering good visualisation, but might be difficult to fit within NetBeans in the time available. A BioJava^? sequence viewer was a suggested alternative, as Kevin and Chris already have some experience with BioJava^?.

There NetBeans interface will provide a visualisation of provenance, with a folder for each workflow enactment instance, the inputs and outputs being children, and the workflow provenance log content.

Notification

The basic notification infrastructure is available, but publishers and subscribers need to be built and tested. A first prototype for MIR notification is under development. The notification of a new DataThing^? being added to a WorkContext^? of interest covered many of the possible examples for the demo. There is considerable ongoing work on a notification subscriber that fits the NetBeans interface. Issues include UI for notifications, intrusiveness, and how notifications relate to the MIR, and topic selection. Mena, Nedim and Simon agreed to continue detailed discussions.

Discovery

Chris Wroe is working on using Pedro as a tool for annotating bioinformatics data with semantic concepts. Of particular interest is the annotation of the registry views. The registry with views is now running (in memory and not as a web service). The EBI can generate metadata; Martin has tools for EMBOSS services. There do not appear to be any conceptual problems, but the details still require working out.

It was noted that Phil's semantic find service requires notifications from the registry views at the time when a service is registered. (The computation involved in reasoning means that it is impractical to leave this to when a client is trying to find a service.) One proposal was that the find service should be a special kind of view. Again the details need to be worked out.

Personalisation

Robert had sent a list of alternative views over the MIR: person centric, project centric, etc. At the moment there is one view, the project/work context view and this closely tied with the underlying data model. Nottingham do not have the resource to develop multiple views within the timescale for ISMB.

Types

There was a discussion on types. Chris G. had identifies a number of concerns about the relationship between the web service world and the ontological description world. See GravesStoryBoardDetails and SimpleSequenceTyping for details

There have been some discussions on exploiting the ACD type system used in the EMBOSS tools. Chris W. has some ideas and is going to talk with Chris G. and others at Nottingham.

Other Issues Discussed

Populating the MIR

For a credible demo, the MIR needs to be populated with data from Graves Disease, and other projects. This includes annotating the data with semantic concepts.

Workflow Language

There was discussion about whether we would be using WSFL or Sculf (from the taverna project) as the workflow description language for the demo. The current position is to use both. The WSFL enactment engine is more tried and tested, but the WSFL language is no longer being used and there are no open source graphical editors for WSFL that we can adopt. A XSculf (XML version of Scufl) enactment engine, based on the same core, is available. Scufl/XScufl has a graphical visualisation available, and a Talisman application for generating EMBOSS workflows that includes graphical visualisation. It is also expected to be better received by biologists and bioinformaticians.

Peter Li was due to be at the EBI for the "emboss, soaplab, talisman" course and the beginning of the following week. Peter, Tom, Darren, Justin, Matthew and Mark are actively working on the workflows and what combination of WSFL and Scufl/XScufl is sensible in the ISMB timescale.

Tom's ISMB talk

This is on Talisman. It will include emboss, soaplab, enactor (including Scufl) and talisman itself. Potentially also any other myGrid components that work with Talisman by the beginning of June.

Follow up

• GDScenarioReqsWithWP6Annotation
• ClassificationWizardNotes

-- MarkGreenwood - 30 Apr 2003