#808 assay harmonization - inputs of assay

[jzheng]

Around half of the OBI assay terms (92 out of 175) defined a specified input. These inputs are usually the analytes or evaluants of an assay. Some assays also specified devices or reagents used in the assays.

List of assays with defined inputs (e.g. input materials with evaluant or analyte role) (shown in the ‘inputs’ worksheet):
https://docs.google.com/spreadsheets/d/1wg2Gg2OvnkX84lVeqRO0ie-tSvtEe4Pq0rtnKsveas/edit?usp=sharing

The input material of some assays are the same but represented differently.
For example, three different representations were used when RNA was analyzed in an assay.
(1) has_specified_input some ‘ribonucleic acid’
http://purl.obolibrary.org/obo/OBI_0001177 RNA sequencing
(2) has_specified_input some (specimen and ('has part' some ‘ribonucleic acid’))
http://purl.obolibrary.org/obo/OBI_0001235 transcription profiling by tiling array assay
(3) has_specified_input some ‘RNA extract’
http://purl.obolibrary.org/obo/OBI_0001674 cap analysis of gene expression
It needs to be a rule to represent it consistently throughout all the assays.

Evaluant role of input material
As for the evaluant, it was present in 29 out of 175 assays. This was represented in three different ways:
(1) realizes some (evaluant role and (role of some X)) [12 assays]
http://purl.obolibrary.org/obo/OBI_0000418 measuring glucose concentration in blood serum
(2) realizes some (evaluant role and (inheres in some X)) [2 assays]
http://purl.obolibrary.org/obo/OBI_0000860 northern blot analysis
(3) has_specified_input some (X and (has role some evaluant role)) [16 assays]
http://purl.obolibrary.org/obo/OBI_0000626 DNA sequencing

Most of the assays use (1) and (3) expression. Only two assays, ‘northern blot analysis’ and ‘southern blot analysis’, use expression (2).

IEDB assay is using following logical axioms:
(has_specified_input some X) and (realizes some ('evaluant role' and ('inheres in' some X))

If OBI would like to take IEDB pattern, evaluant role will use logic axiom (2) and input material should have evaluant role.

Regarding the three different expressions of RNA, the RNA rather than 'RNA extract' or 'specimen has_part RNA' should be used.

Analyte role of input material
20 out of 175 assays specified the analyte role. There were two different logical expressions used to specify the analyte role:
(1) ‘realizes some (analyte role and (role of some X))’ [18 assays]
http://purl.obolibrary.org/obo/OBI_0000435 genotyping assay
(2) ‘realizes some (analyte role and (inheres in some X))’ [2 assays]
http://purl.obolibrary.org/obo/OBI_0000860 northern blot analysis

IEDB assay is using following logical axioms:
(has_specified_input some X) and (realizes some ('analyte role' and ('inheres in' some X))

When consider using IEDB assay logical axioms, some special cases need to look at:
1. http://purl.obolibrary.org/obo/OBI_0001248 ChIP-chip assay
Evaluant is DNA and protein complex, Shall we use two logical axioms to specify each of it has evaluant role?
In addition, DNA also has analyte role. So, same material can play both analyte and evaluant role, right?

1. http://purl.obolibrary.org/obo/OBI_0000787 human antithrombin-III (AT-III) in blood assay
   Current analyte role inheres_in 'scattered molecular aggregate' and ('has grain' some antithrombin-III). We need to change to analyte role inheres_in antithrombin-III and antithrombin-III is input of the assay, right? The input of this assay is 'blood serum specimen'. In this case, shall we have axiom that antithrombin-III is part_of 'blood serum specimen'?

.

[jzheng]

Issues with inputs of following two assays:

http://purl.obolibrary.org/obo/OBI_0001001 Edinburgh handedness assay
has_specified_input value 'Edingburgh handedness inventory'
'Edingburgh handedness inventory' is instance of questionnaire

Is it an assay? If so, it means assay may not only take specimen as input.

http://purl.obolibrary.org/obo/OBI_0001683 chromosome organization assay by fluorescence in-situ hybridization
(in-situ hybridization and (has_specified_input some 'binding constant') ....)
'binding constant' is a data item, is it something wrong here?

[bjoernpeters]

This was discussed on the group call on 7/18.

1) 'has_specified_input' use

Decision As a general rule, whenever possible, for everey inputs to an assay, it should be further specified what the role or function of that input is in the assay. This includes evaluant, reagent, device, analyte. In the three examples:
"(1) has_specified_input some ‘ribonucleic acid’
http://purl.obolibrary.org/obo/OBI_0001177 RNA sequencing"

Here we should state: "RNA sequencing has_specified input some (evaluant and has part some 'RNA')"

2) Use of 'inheres in' vs. 'role of' relationships:

Both ways are saying nearly the same, we just need to agree on one vs. the other. There is a potential argument to be made to use more specific relationships, as that increases reasoning power. There is the opposite argument to be made that increased reasoning power will then also be needed for query execution times in a large triple store (or pre-reasoning with the associated explosion of triples).

Decision: We will for now stick to 'role of'. We will revisit this decision once we have a sufficient number of assays modeled so that we can quantify the impact of one vs. the other.

3) The general case is the use of 'realizes some role (Y and has role X)' vs. 'has specified input (X and has role Y).

Decision: The 'has_specified_input' relationship needs to be implemented, as the current OBI papers and 'normal people' understand it best. However the 'realizes' relationship is necessary to accurately represent what is meant in BFO / OWL. Thus the decision is to implement both

(has_specified_input some (X and has role evaluant role))
and
(realizes some ('evaluant role' and ('role of' some X))

This is a modified implementation of what the IEDB has done so far.

Here as with all decisions, we will revisit the modeling pattern based on performance of queries in triple stores and accuracy of reasoning.

[Chris Stoeckert]

OBI JUly 25, 2016 call: returning to question of RNA sequencing. Proposal was:
"Here we should state: "RNA sequencing has_specified input some (evaluant and has part some 'RNA')"
New proposal and current plan:
"RNA sequencing has_specified input some ('RNA extract' and (has_role some evaluant role))"

[Chris Stoeckert]

OBI call July 25, 2016
http://purl.obolibrary.org/obo/OBI_0001001 Edinburgh handedness assay
add an additional axiom: has_specified_input some (Homo sapiens and has role some evaluant role)