BioThings PFOCR issues

[colleenXu]

Two things I've noticed:

(1) queries like https://biothings.transltr.io/pfocr/query?q=associatedWith.mentions.genes.ncbigene:5879 return no hits even when there are records with that value (example). Maybe it's because there's a misspelling in the metadata/fields entry - which says associatedWith.mentions.genes.bcbigene...

(2) For the PFOCR flavors, I'm not sure how to specify them in the OpenAPI spec / x-bte annotation. They're not regular query-level parameters like fields/size/jmespath/etc....can they be? (see previous Slack convo) @newgene

Refs:

• Create 3 APIs for PFOCR "flavors" #148
• Investigate PFOCR options (strict, synonyms, all) for BTE use biothings_explorer#778

[colleenXu]

@newgene @everaldorodrigo I think the first point is easy to fix - and a bug that needs addressing...

I think the second point may need some more discussion

[everaldorodrigo]

@colleenXu the first point fix was released to the CI, Test, and Production environments.

[everaldorodrigo]

@colleenXu, according to Openapi specs we could modify the endpoint to accept a new parameter called flavor. Ex:

• flavor=all
• flavor=strict
• flavor=synonymous

We also could keep both behaviors. The current one and the new one using a parameter called flavor.

What are your thoughts?

[newgene]

I would just use flavor=all|strict|synonymous, don't think we need to keep both.

[colleenXu]

I think a new parameter like flavor sounds perfect!

[everaldorodrigo]

@newgene @colleenXu the flavor=all|strict|synonymous was deployed to the CI environment. The default value for flavor parameter is all

Examples for testing:

• https://biothings.ci.transltr.io/pfocr/query?q=associatedWith.mentions.genes.ncbigene:5879&flavor=strict
• https://biothings.ci.transltr.io/pfocr/query?q=associatedWith.mentions.genes.ncbigene:5879&flavor=synonyms
• https://biothings.ci.transltr.io/pfocr/query?q=associatedWith.mentions.genes.ncbigene:5879&flavor=all
• https://biothings.ci.transltr.io/pfocr/query?q=associatedWith.mentions.genes.ncbigene:5879

[colleenXu]

I have two observations for you/Alex Pico to check on. Otherwise, I think this looks fine and ready to deploy to all instances!

Two observations:

• It looks like only the gene list differs for the 3 flavors (not chemicals or diseases). Is this correct @AlexanderPico?
• The document order was different for the different flavors. I don't think this is important for Translator use, but I want to point it out in case you think it's a problem.

To check behavior, I took a few individual document/records and compared the mention counts. For example, I checked PMC6883319__gr8...

• setting flavor to all looks the same as not setting flavor: 8 chemicals, 1 disease, 152 genes
• vs setting flavor to synonyms: 69 genes (other entity counts didn't change)
• vs setting flavor to strict: 32 genes (other entity counts didn't change)

[colleenXu]

And a note to myself: for now, Andrew said I don't need to add this parameter's information to the BioThing API's SmartAPI annotation. We're currently using the default (all).

[AlexanderPico]

It looks like only the gene list differs for the 3 flavors (not chemicals or diseases). Is this correct @AlexanderPico?

Correct

[colleenXu]

@everaldorodrigo @ctrl-schaff @newgene

Your thoughts on my second observation?

[newgene]

order does not matter across different flavor parameter values, as they are different set of hits.

[colleenXu]

So from my perspective, this seems good and ready to deploy to all instances.

[colleenXu]

Let's use the tags to track deployment? Based on the Slack convos, it looks like this is on Test now...

[colleenXu]

@everaldorodrigo Are the PFOCR flavors already deployed to Prod? Does that mean this issue can be closed?

[everaldorodrigo]

Hi @colleenXu, yes they were deployed to Prod.