Fig. 1
From: Dyport: dynamic importance-based biomedical hypothesis generation benchmarking technique
Summary of the HG benchmarking approach. We start with collecting data from Curated DBs and Medline, then process it: records from Curated DBs go through parsing, cleaning and ID mapping, MEDLINE records are fed into SemRep system, which performs NER and concept normalization. After that we obtain a list of UMLS CUI associations with attached PMIDs and timestamps (TS). This data is then used to construct a dynamic network G, which is used to calculate the importance measure I for edges in the network. At the end, edges \(e \in G\) with their corresponding importance scores \(I_t(e)\) are added to the benchmark dataset