Table 4 Our guidelines for fairer and comprehensive benchmarks of collaborative-filtering-based drug repurposing models. MF: matrix factorization. NN: neural network. GB: graph-based.
From: Comprehensive evaluation of pure and hybrid collaborative filtering in drug repurposing
Topic | Questions | Our recommendation |
|---|---|---|
Evaluation of models | RQ1. Which metric should the model optimize for? | NS-AUC |
RQ2. Which dataset should the model be evaluated on? | PREDICT (private) or DNdataset | |
Future models | RQ3. Should a method be pair- or matrix-oriented? | Pair-oriented |
RQ4. Which type of algorithms (MF, NN, GB) is the most promising? | Graph-based |
