Hello, i have doubts regarding target prediction

Doubt :

SwissTargetPrediction gives targets based on

1. Uses 2D and 3D chemical similarity to known bioactive compounds in ChEMBL.

2. Predicts likely targets based on similarity to ligands with known activity.

And binding db gives targets based on

1. Reports measured binding data (Ki, IC50, etc.) from published experiments.

2. Data from articles and patents, focused on protein–ligand interactions.

Which one is preferable? As research on phytochemicals have less data, then which method should be selected? Or both methods are selected for knowing targets for my phytochemicals compound??