Assay axiomatization is complex and frequently over-specified

[cmungall]

I see there is an assay paper being written, I hope the following comments are useful for future evolution of assays in OBI, in being able to rapidly add new assays based on user requests, and make OBI easier for non-ontologists to understand.

I find many of the assay axioms are complex and difficult for humans to understand, and logical definitions may be overspecified or not specified at all.

Example:

http://purl.obolibrary.org/obo/OBI_0000424 transcription profiling assay

Def: An assay that determines gene expression and transcription activity using ribonucleic acids collected from a material entity.

Equivalent to: assay and (((realizes some (evaluant role and (role of some RNA extract))) and (has_specified_input some (RNA extract and (has role some evaluant role))))) and (((realizes some (analyte role and (inheres in some ribonucleic acid))) and (has_specified_input some ribonucleic acid))) and (has_specified_input some (specimen and (has part some ribonucleic acid))) and (has_specified_output some measurement datum) and (has_specified_output some is about some gene expression) and (achieves_planned_objective some transcription profiling identification objective)

This is very difficult for a human to understand and mentally reason over, and to determine if the entailed models are the intended ones (e.g. coreference issues).

The text definition does not seem to align with the logical one. See S11 "Match text and logical definitions" in Seppälä et al summarized here: https://douroucouli.wordpress.com/2019/07/08/ontotip-write-simple-concise-clear-operational-textual-definitions/

It is also over-specified. See https://douroucouli.wordpress.com/2019/07/29/ontotip-dont-over-specify-owl-definitions/

Consider a simpler DP, e.g. X assay = assay and has-objective some X objective; then add the other clauses as additional separate necessary conditions. Overspecifying logical definitions limits the ability of OWL reasoners to classify subclasses or individuals.

Another example:

http://purl.obolibrary.org/obo/OBI_0000445 mass measurement assay

An assay that determines the mass of an evaluant

This has no logical def, but has 3 subclass axioms to:

1. assay
2. has_specified_output some mass measurement datum
3. (realizes some (evaluant role and (role of some material entity))) and (has_specified_input some (material entity and (has role some evaluant role)))

I'm not sure what the principles are for when a logical def is present vs when it is not. It seems there is a clear logical def here - are there other assays that produce a mass measurement datum which are not mass measurement assays?

I'm puzzled by the 3rd subclass axiom. Why is it asserted? What information does it add? Do all assays not have material entities as evaluants? I am reasonably familiar with OBI and I know OWL, this confuses me, the average user may be particularly confused.