generate domain/range validation report for SEMMEDDB

[sierra-moxon]

Using the files: https://lhncbc.nlm.nih.gov/ii/tools/MetaMap/documentation/SemanticTypesAndGroups.html
and
https://files.slack.com/files-pri/TSCGQ3XGB-F054B54G9TP/download/semmedver43_2023_r_predication.116080.metaedges.txt?origin_team=TSCGQ3XGB

produce an output file with the following header:
"SEMMEDDB_Subject", "SEMMEDDB_Predicate", "SEMMEDDB_Object", "BiolinkSubject", "BiolinkPredicate", "BiolinkObject", "Valid_Edge?", "Component Missing"

The mapping for subject, predicate, object in Biolink should be made independently, with a second check for whether this combination is a valid edge in biolink.

Biolink edge is valid when the code can find valid biolink classes for subject and object and a valid biolink association slot (which is also a child of related_to, aka: a biolink predicate), for the predicate, AND the domain and range constraints on the biolink predicate encompass the passed subject and object respectively.

For example:

SEMMEDDB_Subject        SEMMEDDB_Predicate      SEMMEDDB_Object BiolinkSubject  BiolinkPredicate        BiolinkObject   Valid_Edge?     Component Missing
dsyn    PROCESS_OF      humn    biolink:Disease biolink:occurs_in               False   humn
fndg    PROCESS_OF      humn    biolink:DiseaseOrPhenotypicFeature      biolink:occurs_in               False   humn
neop    PROCESS_OF      humn    biolink:Disease biolink:occurs_in               False   humn
topp    TREATS  podg    biolink:Procedure       biolink:treats  biolink:Cohort  False   None
patf    PROCESS_OF      humn    biolink:PathologicalProcess     biolink:occurs_in               False   humn
topp    USES    phsu    biolink:Procedure       biolink:has_input       biolink:Drug    False   None
topp    TREATS  dsyn    biolink:Procedure       biolink:treats  biolink:Disease False   None
cell    LOCATION_OF     aapp    biolink:Cell    biolink:location_of     biolink:Polypeptide     True    None
mobd    PROCESS_OF      humn    biolink:Disease biolink:occurs_in               False   humn
bpoc    PART_OF mamm    biolink:GrossAnatomicalStructure        biolink:part_of         False   mamm
sosy    PROCESS_OF      humn    biolink:PhenotypicFeature       biolink:occurs_in               False   humn

for the first row - the output is saying this isn't a valid Biolink edge because there is no mapping in Biolink for STY:Txxx humn (I imagine this is "human").

for this row:

topp    TREATS  podg    biolink:Procedure       biolink:treats  biolink:Cohort  False   None

we're saying that we can match everything in the row (there are no missing components), but its not a valid edge in biolink. For the treats predicate, the domain and range are as follows:

domain: chemical or drug or treatment
range: disease or phenotypic feature

[sierra-moxon]

code to produce this report is here: github.com/sierra-moxon/predicates-analysis/

first draft output for this is here:
semmeddb_biolink_triples.tar.gz

[sierra-moxon]

per feedback from Andrew, added back in counts, reordered columns:
semmeddb_biolink_triples.tar.gz

[sierra-moxon]

next iteration, per feedback from Matt:
semmeddb_biolink_triples.tsv.tar.gz

[sierra-moxon]

report was presented and accepted; relay session to follow but for now this work is done. closing.