#810 assay harmonization - outputs of assay

[jzheng]

141 out of 175 assays specified outputs.
List of assays with specified outputs (shown in the ‘output’ worksheet):
https://docs.google.com/spreadsheets/d/1wg2Gg2OvnkX84lVeqRO0ie-_tSvtEe4Pq0rtnKsveas/edit#gid=1934198426
The general outputs are usually ‘measurement datum’ while the less broad outputs are either specific data items or data items about some entity (generally GO biological process or molecular function). One example of the latter is ‘binding assay,’ which has ‘binding datum’ as the output. Also, the assays that have outputs of ‘informational content entity’ need to be updated because it has been agreed upon that all outputs are data items since the output is something we believe to be true.

Two kinds of outputs are used in IEDB assays:
(1) has_specified_output some X [X is subClassOf 'data item']
(2) has_specified_output some ('is about' some X) [X is GO biological process]

Assays other than IEDB assays used the similar expressions. For (2), it is broder than GO biological process, can be molecular entity, cell, organism, etc. In addition, in (2) it would be specify it is the 'data item'. Otherwise, it infers the output is ICE.

[bjoernpeters]

Call 9/12
Agreed that the output of an assay is always a data item, as previously discussed, as the information genereated is considered (approximately) truthful (previous tracker item 733)

Assays typically produce intermediate information that is routinely transformed to the desired final information entity, for example, in the case of a 'microarray gene expression assay', there is intermediate data produced about optical readouts that are subsequently transformed into gene expression values. The information captured as the output of the assay should be the final data item. If there is a desire to model the intermediate steps, the assay can be broken down into parts + data transformations. (see also #https://sourceforge.net/p/obi/obi-terms/816/)

There should always be an 'is about' relationship between the data item and something that is being measured, such as that the assay is about a binding process.

If the output has a specific form, subclasses of data item can be used, such as 'half life of binding datum', which is one way of charactereizing a binding process

When an assay is general, such as a 'Chip-Seq assay', or a 'binding assay', we should not try to enumerate all possible biological applications it can be used for in the types of data items generated. Rather, the data item should be generic, and if there is a desire to model specifically e.g. Transcription factor binding site identification by Chip-Seq, then that specific assay should be created with the more specific data item as an output.

There might be outliers in the current assay list (such as performing a clinical assessment) which fall out of the discussion above; they should be revisited separately, and in the diagnosis case a tracker items exists (https://sourceforge.net/p/obi/obi-terms/637/).

[bjoernpeters]

Ticket 155 was related to this (raw vs. processed data). We might want to be able to call out the 'raw data' / original data' given by an assay. The 'original' data from a given assay is the specified output of that assay, e.g. DNA sequences are the ouptut data of a DNA sequencing assay, not intermediate steps (such as images), and not processed versions thereof (such as a phylogenetic tree).