Open andrewsu opened 3 years ago
final result will depend on the fix of #185
Rebooting this issue as a rotation project for Hanbing. Let's start with cell 20 in https://github.com/SuLab/DMDB_Analysis/blob/main/1_code/1_basic_dmdb_analysis.ipynb, which lists the common metapaths used in DrugMechDB. Starting with one of the more common metapaths, let's create a corresponding TRAPI query and see how many examples BTE (and/or other Translator tools) can retrieve.
DrugMechDB (https://github.com/SuLab/DrugMechDB) is a database of drug mechanisms expressed as paths (or small graphs) through a knowledge graph. Let's take each edge in DrugMechDB, formulate a TRAPI query, and see if BTE can find it. Output will be some sort of report that describes what percentage of DrugMechDB edges are found, stratified by edge type.
Mike created an initial analysis in this repo https://github.com/SuLab/DMDB_Analysis/tree/main/1_code and in particular this notebook https://github.com/SuLab/DMDB_Analysis/blob/main/1_code/4_dmdb_bte_trapi_analysis.ipynb. Would be good to finish/finalize this.