The TCDC and the SRI, with input from the O&O team, the UI team, and the ARS, converged on the plan below for incorporating clinical_evidence into O&O and the f()-score. This issue is to track progress on implementation and resolve any unforeseen issues that may emerge.
O&O Clinical_Evidence factor and SRI Answer Appraiser ('official' name)
Each clinical KP lead creates a KGX file with drug-disease pairs plus five edge attributes (log_odds_ratio, log_odds_ratio_95_ci, total_sample_size, infores:id, biolink:supporting_data_source [URL for wiki page])
JSON-Lines version preferable, not tsv, if possible
TCDC Max/Kara creates merged KGX from the three clinical KP KGX files and adds a fourth edge with weighted average of log_odds_ratio for each drug/chemical-disease pair
Impose cap on score: MIN(log_odds_ratio, CAP=10)
Normalize [0,1] using arctan function
TCDC Merged KGX file to Abrar for incorporation into SRI Answer Appraiser
ARS gets results set 1 & calls BTE Annotation Service, then SRI Answer Appraiser (sequential calls)
SRI Answer Appraiser receives request & finds drug/chemical-disease pairs in each result, then looks at merged KGX file to determine if drug/chemical-disease pair is present
If present, adds four edges to result
Update: Edges will be added post-September release; score only pre-September, leveraging wiki page to explain score to
users
The TCDC and the SRI, with input from the O&O team, the UI team, and the ARS, converged on the plan below for incorporating clinical_evidence into O&O and the f()-score. This issue is to track progress on implementation and resolve any unforeseen issues that may emerge.
O&O Clinical_Evidence factor and SRI Answer Appraiser ('official' name)
TCDCMax/Kara creates merged KGX from the three clinical KP KGX files and adds a fourth edge with weighted average of log_odds_ratio for each drug/chemical-disease pairTCDCMerged KGX file to Abrar for incorporation into SRI Answer AppraiserUpdate: Edges will be added post-September release; score only pre-September, leveraging wiki page to explain score to users