gaurav
commented
8 months ago At the moment, we have:
- Two Redis databases for conflation: gene_protein_db and chemical_drug_db. Each of these consists of identifiers mapped to a list of identifiers (e.g. in chemical_drug_db, RXCUI:1486320 is mapped to
"[\"PUBCHEM.COMPOUND:16666\", \"RXCUI:393549\", \"RXCUI:1046245\", \"RXCUI:1046531\", \"RXCUI:1235314\", \"RXCUI:1245396\", \"RXCUI:1486320\", \"RXCUI:1795938\", \"RXCUI:1863585\", \"RXCUI:1994323\", \"RXCUI:1999886\", \"RXCUI:1999887\", \"RXCUI:1999888\", \"RXCUI:1999889\", \"RXCUI:2102940\", \"RXCUI:2105551\", \"RXCUI:2566881\", \"RXCUI:2568111\", \"UMLS:C2698944\"]\n") - Two different flags for activating these conflations:
conflate(for gene/protein conflation) anddrug_chemical_conflate(for drug/chemical conflation)
We probably want to roll these up into:
- A single conflation database that is divided up between different conflations. The dumb way of doing this in Redis would be with a prefix (e.g.
chemical_drug_db|RXCUI:1486320), and since we only have ~9k DrugChemical conflations and 8.9 million GeneProtein conflations, I don't think it would be unreasonable to put them all into a single place, although of course lookups might be slower. - A single flag (e.g.
conflations) to which a list of conflations to apply could be provided.
Currently we have gene/protein conflation.
The ability to handle other types of conflation is partially implemented. Specifically, the backend parts are probably reasonably close, but we don't have a very good parameter scheme for specifying different types of conflation independently.