1st (RSBI) ISA-Tab Workshop

1
1st (RSBI) ISA-Tab Workshop Scope and Outcome

• Tackle today's need for exchange of multi-omics
  experiments
• Evaluate the ISA-TAB straw-man (incomplete)
  proposal
• - It builds on this existing paradigm of the
  MAGE-TAB
• It shares the same motivation for the use of
  spreadsheet (versus the complex, but expressive
  MAGE-ML)
• Current use cases
• Submission (import) and download (export)
  tabular format
• Note not human created but assisted by a tool
  (e.g. forms)
• Also, as exchange format while the suites of
  FuGE-ML based modules are under development
• At that stage, ISA-TAB can finding utility as a
  user-friendly presentation layer for the
  XML-based formats (via an XSL transformation)
• Work plan on how to move forward
• The ISA-TAB may not fulfil everybody needs
• Hard to make one measure to fit all
• What need to be modified or changed, how to test
  it, etc.
• To be finalized with interested parties

2
Technical Challenges To Be Solved (1)

• Semicolon-separated fields into a tabular format
  (e.g. Person Role) Phil J
• Formatting issues
• Problem when sources are two different
  CVs/Ontology
• Section dividers Michael M
• Is an empty line the best way? Roles to be
  defined (e.g. Protocol)
• Or should each section be in different tabs,
  e.g. for sorting?
• CVs/Ontology and keywords/comment (free text)
  Weida T, Philippe RS
• Which fields need CVs/ontology vs free text
  (e.g. Factor must be a CV)
• For CVs/Ont what about having term, id and
  reference? Chris T
• Ensure that id (how?) columns are the only one
  carried across tabs Allyson L
• E.g. in both Study and Assay tab
  CharacteristicMaterialType is repeated

3
Technical Challenges - To Be Solved (2)

• Need for qualifiers to be CV/Ontology
  Philippe RS
• Allowing free text is problematic?
• Restricting FactorValue Michael M
• Is there ever case where FactorValue is neither
  Characteristics or Parameter?
• There may be a problem with how ISA-TAB currently
  deals with protocols and protocol parameters
  (inherited from MAGE-TAB)
• Parameters and characteristics have the same
  weight
• Either we capture all parameters/characteristics
  in the protocol section, or we have it only in
  the study/assay sections
• What if the contact is not a person? Chris T
• Example Person X is quitting, so wants to fill
  in a job title or role instead
• Do we need to worry about this, or is it simply
  not worth modelling?

4
Technical Challenges - To Be Solved (3)

• Reference system between ISA-TABs Study and
  (SDTM) SEND and CDISC Michael M
• Also reference between ISA-TABs Studies when
  e.g. samples are reused
• Referencing out to other Protocol Formats Weida
  T
• In the Reference tab it would be nice to
  reference other formats of studies that aren't
  applicable to us directly
• Examples of references would include SDTM, SEND,
  and CDISC
• Possibility of renaming Sample Name to
  Specimen Name Jen F
• Alternatively, could use CharMaterialType
  specimen
• Work for an CV/ontology (to be discussed in OBI)
• Normalization Method should it be required?
  Stefan W
• Normalized data isnt worth much without knowing
  the normalization procedure
• Put in normalization as a data in data out
  protocol?

5
Questions with (possible) Answers

• When a field is critical to one system but not to
  another - Weida T
• Fully optional, marked as MI-compliant (choice
  is up to each system) and absolutely mandatory
  (for EBI systems)
• The latter are the linking names to
  successfully understand/parse it
• Current (and suggested) use cases - Phil J
• Submission (import) exchange (in the absence of
  all the FuGE-based modules)
• Download (export)
• Sorting possibly
• Some (referenced) fields are redundant - Stefan W
• In Study, however, it is when there are multiple
  study you need to know who did which one
• High level classification for Assay Weida T,
  Phil J
• Pairing of endpoint (gene expression) and
  technology type (DNA microarray)

6
Questions with (possible) Answers

• IDs Michael M
• For each ISA-TAB submission human generated
  files (internally unique if), software generated
  (globally unique ids, e.g. LSID)
• Improve I/S/A definitions - Michael M
• Move Ontology Source Ref up in then Reference
  file - Michael M
• General comments about the separation of a study
  into multiple studies
• It is left to the submitter, for their
  convenience
• Need to be made clear, also the use of
  Investigation to group related Studies

7
1st ISA-Tab Workshop Outcome and What Next (1)

• What have we achieved these two days?
• Initial evaluation of the the ISA-TAB
  straw-man proposal
• Listed technical issues that need to be
  addressed
• Started creating a new example
• We should have enough info now to develop a sound
  plan
• When can we meet next and what the goal would be?
• Around May?
• Finalize the draft specification example (to
  be published)
• What do we need to do before then?
• Create examples in the various domains (using
  published experiments)
• Identify leading persons, responsibilities
  timelines et
• Dissemination plans and documentations
• Create webpage, list etc (google) Philippe
• Refine document (iterative) Philippe, Weida,
  Susanna
• Technical challenges

8
1st ISA-Tab Workshop What next (2)

• Biological domains
• Dawn genomics/metagenomics
• Norman environmental
• Jennifer/Philippe toxicology
• Jennifer clinical
• Stefan cell assay
• Identify the papers create Studies in your
  domains and with the help of the leaders below-
  work on creating Assays
• Technological domains
• Dawn sequencing
• Phil proteomics
• Nigel metabolomics
• Tim transcriptomics/PCR
• Stefan/Philippe cell assay (non-omics)
• Jennifer histopathology
• Define the required column headers Assist
  biological domain leaders above and work to
  define fields in the your Assay tabs

9
1st ISA-Tab Workshop What next (2)

• Implementation (ideas-plan)
• Phil/Marco
• Excel (some versions are GLP) and open office,
  etc.
• Initial design of BioMAP-Harvest/ISA-tab-Harvest
  (based on PrideHarvest)
• Conversion to FuGE and reverse (XSL
  transformation) (ideas-plan)
• Allyson/Philippe/Marco
• Timelines/deliverables (end of each month)
• JAN Examples/papers identified (from biological
  groups) webpage up
• FEB Define columns headers in each technology
  domains defined
• MAR Technical challenges addressed Study tabs
  filled
• APR Assay tabs filled check/review examples
• (Iterative) refine the document
• (mid?) MAY Meeting
• - Spec, examples ready, ideas on implementations
  and FuGE conversions