Provide a list of _patterns_ associations with LOINC terms

[matentzn]

After todays great meeting, @ShahimEssaid provided a good idea of how we can save some time in our quest for identifying different patterns.

Shahim, can you create a table for us that has three columns, one with the LOINC code (full assay), and one with the pattern (the logical definition entirely independent from the actual fillers)? Something like hasComponent some X and hasAnalyte some Y.). The goal is to group the LOINC terms by pattern and use that information to drive our own identification and prioritisation of patternisation.

cc @drseb

[matentzn]

@ShahimEssaid @putmantime we are ready to resume this! Can you provide us with this list perhaps?

[ShahimEssaid]

@matentzn I haven't had much time recently to work on this but I''ll be getting to this this week or next. I'm not sure I understand the table yet.

col1: LOINC code as a string col2: an example: "hasComponent some X and hasAnalyte some Y" col3: ?

Our current tooling is LinkML based and I'm not yet sure about how to configure the dumping of the instances into such a table but we'll figure it out. I'm new to LinkML and I've only learned how to dump to OWL so far.

[matentzn]

I think col1, col2 is a great first pass! Maybe add the LOINC code name. We can then proceed to figure out how to map hasComponent some X and hasAnalyte some Y to DOSDP patterns. Thank you!

[matentzn]

@ShahimEssaid did you get a chance to look at this yet?

[ShahimEssaid]

A draft is here: https://drive.google.com/file/d/1UWNVcN06L9qaMX8FR_uMZW7bfjX8OvpC/view?usp=drive_link

Let me know what you think and I can adjust. It has all the LOINC codes, their names and classes, and the corresponding component and analyte entries.

[matentzn]

Fantastic @ShahimEssaid . I was hoping for something else, but this may be even better - depending of what column C mean

LOINC CODE | LOINC NAME | ? | ? | ? | COMPONENT | ? | ? | ? | ? | ? | ? -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- 1-8 | Acyclovir [Susceptibility] | ABXBACT | 1 | LP14522-4 | COMPONENT | Acyclovir | Acyclovir | LP14522-4 | COMPONENT | Acyclovir | Acyclovir Can you fill in the column headers so I can understand them better?

[ShahimEssaid]

I just noticed that I accidentally saved a resorted version of that file and the headers got moved to row 50797. I saved a new file here: https://drive.google.com/file/d/1B2HBpyKj9wwBQPUsz0ZFNe-rzs07s9Xt

The headers are mostly self explanatory, and are based on the release files' header names, but here are a couple of useful pointers:

• loinc_class: https://loinc.org/kb/users-guide/classes/
• loinc_class_type: 1=Laboratory class; 2=Clinical class; 3=Claims attachments; 4=Surveys and see field "CLASSTYPE" here: https://loinc.org/kb/users-guide/loinc-database-structure/
• The two columns filled with "COMPONENT" turned out to not be that useful but they represent the "PartTypeName" that is documented in the release file: AccessoryFiles/PartFile/PartFileReadMe.txt

[matentzn]

Thank you @ShahimEssaid. Given this table, how would I be able to determine which LOINC terms would be part of the same "pattern", i.e. is associated to the same kind of logical axiom template?

[joeflack4]

Just linking this here; I think it may be related:

• https://github.com/loinc/comp-loinc/issues/37#issuecomment-2286744388: Shows Unique combinations of decomposed analyte properties (patterns) by analyzing the supplementary part model (LoincPartLink_Supplementary.csv).