Questions arising form the SQL-to-ARC view definitions

[Zalfsten]

SQL-to-ARC is based on a set of view definitions, defined in the file 'docs/sql_to_arc_database_views.md'.

These views are based on the ARC ISA XLSX specification. The initial attempt was to adopt these specs 1:1. Unfortunately this is not possible, because a Excel-Sheets definition cannot always directly be mapped to a database layout (e.g. the Excel-Sheets have implicit relationships just by being defined in the same file). Also the XLSX spec defines entities that are not known to the ARCtrl library. The same holds in the other direction: the ARCtrl library (and also the ARCitect) have features that are not reflected within in the XLSX spec.

This issue collects all discrepancies and questions.

• In ARCtrl many entities (OntologySource, Investigation, Publication, Contact, Study, Assay) and/or the relationships between them can be assigned one or more comments. This is not covered by the XLSX spec.
• ARCtrl offers the field 'ORCID' for a Contacts, the XLSX spec does not.
• The assay technology platform can be an ontology reference within ARCtrl, but is a simple string in the XLSX spec.
• The XLSX spec defines a bunch of Study properties that are not available in ARCtrl:
  • study design descriptors
  • study factors
  • study protocols
  • study assays
• Concerning 'study protocols': According to my understanding, protocols are encoded within the headers of annotation tables. Annotation tables are also part of the XLSX spec. So are there two ways to define study protocols in the XLSX spec? Maybe due to historical reasons? Or are these distinct concepts? I assume that the annotation table column type 'protocol_ref' is not needed, if there are no explicit 'study protocol' definitions, is this corect? What about the column types 'protocol_description', 'protocol_type' and 'protocol_uri'?
• Concerning 'study assays': in the XLSX spec they are defined identically to the entity assays. The only difference seems to be that 'study assays' belong to studies, whereas 'assays' belong to the investigation. This does not seem to map to the ARCtrl approach, where assays always belong to the investigation but can be registered within an arbitrary number of studies.
• For annotation table input/output columns ARCitect (probably also ARCtrl) allows to either specify an ontology reference or a string value. The XLSX spec only allows an ontology reference.