Meeting  for 06-13 Workflows
Present: Bill, Rebecca, Richard

Richard - going over the excel sheet  from before
    - would be useful to go over the  differences between knowledge retrieval WF
- Kepler seems  to be much more used in the literature for things that wouldn't be in  myExp, as that is populated by developers and bioinformatics alone

Bill:
Hypotheses:
1)    Workflows are becoming more complex and powerful over time:
a.      as demonstrated by increases in numbers of components and dataflow  links
b.     as demonstrated by increased  branching
c.      as demonstrated by increased  numbers of sub-workflows (embedded workflows)
d.      proportion of workflows that perform simple data acquisition vs those  that perform numerous processing steps 
2)   Most  workflows perform simple, but repetitive data acquisition tasks as  opposed to complex operations
3)   Workflows  become more complex as one gains more experience (i.e., number of  previous workflows created by that individual)
4)   Workflow  re-use (downloads) is proportional to the complexity of tasks performed  by the workflow.
5)   Workfow re-use is proportional to  the sufficiency (comprehensiveness) of the documentation (i.e.,  metadata) 
6)    

Understanding workflows:
1)    numbers of data sources within a workflow
2)    incorporation of QA/QC
3)   number of components
4)    number of workflow links
5)   date created
6)    number of downloads
7)   number of sub-workflows
8)    use of specific data sources
9)   use of  specific models embedded in workflow
10)  number of  workflows created by an individual
11)   discipline covered by workflow
12)  number of  workflows  created by individual users (what is the shape of the  curve?  Long tail?) 
13)                   
 
 Richard:

   Organise in terms of 2,3 tiers of information 
      1: high level - author,  when, how many, nodes, workflow links, goal
        2: of nodes, % of data  aquisition, % of data sources, type of nodes, models?
            3: sufficiency of  the metadata (semantic and natural language description), plays into  workflow reuse
  Time ten random workflows
  
  The goal is to verify/ look  into the hypothesis above. (and then go on to more)
  
  additional hypotheses?
   - ways to automate? (for  the email to David)
      - also, suggestions for  other ways to analyse these
      
      1. revise spreadsheet to  be organised by tiers
       2. Get to grips with  SPARQL as much as possible
      3. write up for the  approach (methods, teir 1/2/3, selection processes, null hypothesis,  etc...)
      4. use for a basis to  contact David, and as a mini-proposal for what to do for the summer.  Send out when done. Significant time restraints in what we can do -  justification for myExperiment, using tiered approach to do that. (2-5  pages probably) 
      5. Keep track of a list  of bioinformatics/ confusing ones
      
      London time 1600,  Mountain Time 900, Pacific Time 800, Tuesday